Sie sind auf Seite 1von 62

Prediction of the Impact of Increasing Frequency of Bushfire on the Water Resources of the Forested Upland Catchments of the Murray

Basin
A scoping study of model options and linkages for a whole ecosystem model

TECHNICAL REPORT 09/1 August 2009

Produced for the Murray-Darling Basin Authority by:

Page i

Final Report August 2009 ISBN: 978-0-9806387-0-7 Monash Sustainability Institute, 2009 Authors: Terence Chan John Langford Ralph Mac Nally Philip Wallis Contributors: David Abramson Patrick Baker Jason Beringer Nick Bond Jo Brown Tim Cavagnaro Colin Enticott Dave Griggs Christian Jakob Phil Jordan Nick Marsh Kirsten Shelly Mike Stewardson Ross Thompson Monash Sustainability Institute Building 74, Clayton Campus Wellington Road, Clayton Monash University VIC 3800 Australia Tel: +61 3 990 59323 Fax number +61 3 990 59348 Email: enquiries@msi.monash.edu.au Web: www.monash.edu.au/research/sustainability-institute/ DISCLAIMER: Monash University disclaims all liability for any error, loss or consequence which may arise from you relying on any information in this publication. Page ii

Contents
Summary ...................................................................................................................................................... 1 1. Introduction .......................................................................................................................................... 2 1.1. Objectives of this scoping study .................................................................................................... 3 1.2. Our approach ................................................................................................................................. 4 2. Modelling capacity ............................................................................................................................... 5 2.1. Climate and weather models ....................................................................................................... 10 2.1.1. SEACI .................................................................................................................................. 10 2.1.2. ACCESS .............................................................................................................................. 10 2.1.3. Summary ............................................................................................................................. 10 2.2. Hydrology models ........................................................................................................................ 12 2.2.1. Simple rainfall-runoff models ............................................................................................... 12 2.2.2. Physically based hydrological process models ................................................................... 13 2.2.3. Modular systems.................................................................................................................. 13 2.2.4. Summary ............................................................................................................................. 14 2.3. Biogeochemistry models ............................................................................................................. 15 2.3.1. Export models ...................................................................................................................... 15 2.3.2. Biogeochemical cycling models ........................................................................................... 16 2.3.3. Summary ............................................................................................................................. 16 2.4. Vegetation models ....................................................................................................................... 18 2.4.1. Vegetation distribution models ............................................................................................ 18 2.4.2. Stand growth models ........................................................................................................... 18 2.4.3. Summary ............................................................................................................................. 19 2.5. Terrestrial biodiversity models ..................................................................................................... 20 2.5.1. Summary ............................................................................................................................. 20 2.6. Aquatic biodiversity models ......................................................................................................... 21 2.6.1. Process-based ecosystem models ...................................................................................... 21 2.6.2. Aquatic biodiversity models ................................................................................................. 21 2.6.3. Population models ............................................................................................................... 22 2.6.4. Summary ............................................................................................................................. 22 3. Linking models together ................................................................................................................... 23 3.1. Modelling systems ....................................................................................................................... 23 3.1.1. The Nimrod toolkit................................................................................................................ 23 3.1.2. Kepler .................................................................................................................................. 24 3.1.3. Interactive Component Modelling System ........................................................................... 24 3.1.4. Catchment Management Framework .................................................................................. 24 3.1.5. Ecological Modeller ............................................................................................................. 25 3.2. Data management systems ......................................................................................................... 25 3.2.1. National Data Grid Demonstrator Project (formerly PEMS) ................................................ 25 4. A conceptual framework for linking models ................................................................................... 26 4.1. Processes represented ................................................................................................................ 27 4.1.1. Climate and weather ............................................................................................................ 27 4.1.2. Hydrology............................................................................................................................. 27 4.1.3. Biogeochemistry .................................................................................................................. 28 4.1.4. Vegetation............................................................................................................................ 28 4.1.5. Terrestrial Biodiversity ......................................................................................................... 29 4.1.6. Aquatic Biodiversity ............................................................................................................. 29 4.2. Limitations.................................................................................................................................... 29 4.3. Issues of scale ............................................................................................................................. 30 4.4. Issues of uncertainty .................................................................................................................... 31 4.5. Issues of model integration .......................................................................................................... 32 5. Program of work ................................................................................................................................. 33 5.1. Recommended approach ............................................................................................................ 33 5.1.1. Define specific research questions and goals ..................................................................... 33 5.1.2. Select a case study catchment. .......................................................................................... 34 5.1.3. Identify data requirements ................................................................................................... 35 5.1.4. Obtain access to component models .................................................................................. 35 5.1.5. Implement grid workflows .................................................................................................... 36 5.1.6. Calibration and parameter optimization ............................................................................... 36 5.1.7. Validation, analysis and scenarios ...................................................................................... 36 5.1.8. Iterative model development ............................................................................................... 36 6. Conclusions ........................................................................................................................................ 37 Page iii

7. Glossary .............................................................................................................................................. 40 8. References .......................................................................................................................................... 41 Appendix 1 Detailed model comparison tables ................................................................................... 47

Page iv

Summary
Comprehensive understanding of climate change and its consequences for water resources and quality at a river basin or large catchment scale is vital to planning the future management of the Murray Darling Basin. Changes in Australias climate are causing increased uncertainty about the likely impacts at a river basin scale, of events such as repeated bushfires, on whole ecosystems and the consequences for water resource availability and quality. Climate change will simultaneously impact on the vegetation, biogeochemistry, terrestrial and aquatic ecology and the hydrology of catchments. A new approach is therefore required that takes a whole ecosystem view in understanding and predicting the impacts of climate change on large catchments over long time periods. In this report we outline how a whole ecosystem model can be assembled from existing component models, including vegetation dynamics, hydrology and biogeochemistry process models, into a single virtual model. Such a model could be driven by stochastic weather projections derived from downscaled global climate models to make predictions about the effects of climate change on ecosystems and consequently on available water resources. Model outputs, including predictions of habitat distribution, could be used to drive statistical models of terrestrial and aquatic biodiversity. This approach would be capable of being used to assess the whole ecosystem consequences of major impacts, such as large-scale bushfires, for water resources across a large catchment. This report presents the first steps in thinking about model choice within each component, and the use of the latest computing techniques to link component models for the purpose of building a whole ecosystem model. We identified six components that were considered necessary to describe a whole ecosystem, including: climate and weather (data as driving inputs); hydrology; biogeochemistry; vegetation dynamics; terrestrial biodiversity; and aquatic biodiversity.

An overview of modelling is given for each component, as well as an assessment of the required process representations. The review of component models was limited to an assessment of modelling options, as time constraints precluded a more in-depth analysis. We do not present any recommendations for choosing a specific model as this will depend on a more in-depth analysis of the aims of the modelling task. The approach to linking each of the modelling components outlined in this report utilises grid workflows technology that can link different software models together and stream data into and out of each component. This approach has some key advantages over other modular modelling systems, in that it can link models written in different code together by wrapping them in scripts that control inputs, outputs and parameters. However, while grid workflows have been demonstrated in a range of high performance computing applications, the technology has not been applied to an ecological modelling task of this magnitude and would require further development. The development of a linked modelling system that can represent whole ecosystems over long time periods in order to inform catchment-scale natural resources management is a challenging task; but one that is feasible and has the potential to redefine the way that natural systems are understood. This approach would be highly valuable to the agencies that manage natural resources on a river basin scale.

Page 1

1. Introduction
Since 1997, the forested upland catchments of the Murray Basin have experienced a shift to a drier and hotter weather pattern (Murphy and Timbal 2008). In turn, these changed climate and weather patterns have created ideal conditions for the spread of major bushfires (CSIRO 2007; Howe et al. 2005). Widespread fires have occurred in the 2002/03, 2006/07 and 2009 fire seasons covering most of the upland catchments feeding the Murray River (Figure 1). Indeed some areas have been burnt more than once within a short period of time with potentially profound implications for the vegetation cover on these catchments, and consequently the water resources derived from them.

Figure 1 Bushfire impacted areas in south-eastern Australia (inset: annual average rainfall in the Murray-Darling Basin) (Sources: DSE, Geosciences Australia, Bureau of Meteorology).

Page 2

Repeated fires in the wet eucalypt forests could eliminate mountain and alpine ash forests if highintensity fires recur before the re-growth is old enough to seed (McCarthy et al. 1999). Major fires in both the 2002/03 and 2006/07 fire seasons have severely impacted succession vegetation where repeated burns have occurred. In addition to fire frequency, a hotter and drier climate will progressively diminish the habitat of the wet eucalypt forests and they will contract. The dry mixed species eucalypt forests in the upland catchments will also suffer from the consequences of a hotter, drier climate. The mature trees could be weakened by the repeated fires and loss of nutrients from the ecosystem with potentially significant consequences for the succession vegetation, and its hydrological balance and the availability of water resources. In a world of changing climate, simply modelling the hydrology of these forests as they age after infrequent bushfires will no longer be sufficient to predict the impact on water resources. Previous studies of bushfire impacts on water yield and water quality in south-eastern Australia have focused separately on water quality (Feikema et al. 2008) or on hydrological variables and the potential effects of fire on catchment evapotranspiration and streamflow (Lane et al. 2007). Integrating and extending these studies and looking at the impact of repeated events is essential, given the potential for increasing fire risk from climate change (Howe et al. 2005). Modelling whole ecosystems as one integrated system presents a number of significant advantages for predicting the ecological effects of a range of climate change scenarios over large catchments for long time periods. It will be necessary to consider time periods of over 100 years to describe the changes in vegetation as the climate changes and the impact of repeated bushfires accumulates. A whole system approach is also essential to understanding the effects of climate change and consequent increased frequency and severity of bushfires on catchment hydrology aquatic ecosystems and river water quality. Reduced water flowing to aquatic ecosystems in the Murray River, as a result of overallocation of water resources and extremely dry conditions, have already resulted in algal blooms, black water events and acid drainage (Baker et al. 2000; Hall et al. 2006; Howitt et al. 2007). Climate change and more frequent bushfires will exacerbate these problems unless steps are taken. A whole system ecological model will allow future scenarios to be developed to inform future river basin planning on the likely availability of water resources and the management of water quality and river health.

1.1. Objectives of this scoping study


To scope the feasibility and work involved in building a whole ecosystem model suitable for prediction of large-scale impacts (in this case, bushfires in the Victorian uplands) on water quality, water yields and aquatic ecology over large catchments in the southern Murray Darling Basin for long time periods. To review the availability and utility of component models that describe a whole ecosystem (comprising six modules: hydrology, biogeochemistry, climate and weather, terrestrial biodiversity and aquatic biodiversity), and consequently define gaps in current modelling capability. To review the feasibility of interconnecting component models by leveraging the existing code and using contemporary approaches to computer workflows and model coupling. Grid Workflow systems enable the coupling of component models into a single virtual model. To build a conceptual model that shows which model elements are typically present in each component (based on the six modules listed above) and to identify where linkages can be made between common elements.

Page 3

1.2. Our approach


In this report, we present a feasibility study of the program of work necessary to build a whole ecosystem model capable of predicting the implications of climate change and an increasing frequency of fire on water quality, water yield and aquatic ecology of the Murray Basin. In addition, we review and report on the availability of models covering the six components of a whole ecosystem model, the gaps in modelling capacity, and a review of the feasibility of interconnecting the models. We review ecosystem component models within a framework that we believe is necessary to assess whole-of-ecosystem consequences of impacts, such as large-scale bushfires. These components include vegetation models, biogeochemistry models, climate and weather models, hydrological models and aquatic and terrestrial biodiversity models. We generally focus on models which attempt to dynamically represent and simulate real physical, chemical, biological and ecosystem processes, as these will be required to provide projections of impacts under conditions not previously observed. Note, however, that the models considered are often hybrids, including empirical descriptions where needed for simplification or because of data limitations. It should also be noted that where the current state of understanding is limited (and particularly where this intersects with large natural variability, e.g. in biodiversity), and process-based models are not available or feasible, statistical models are also considered. We next consider the technologies available to link these component models together into a single virtual model that allows an integrated assessment of catchment-scale impacts (e.g. large-scale bushfires) without the need to re-write a single modelling framework from scratch. We also report on the capabilities of grid workflow systems to manage computational load across computing grids. Finally, we present a conceptual framework for linking each essential component model. We show how these connect together conceptually, as well as the program of work required to link them computationally. From this analysis emerges an understanding of model shortcomings and opportunities to better integrate each component and finally to reach conclusions about the feasibility of developing a linked model of a catchment ecosystem.

Page 4

2. Modelling capacity
The capacity to model ecosystem processes varies significantly across ecological disciplines. Models have only recently begun to cross traditional disciplinary boundaries, but are still most readily classified according to ecosystem components. We believe a minimum of six components are required to describe the interactions necessary to create a whole ecosystem model of a large catchment capable of modelling over long time scales. These components are listed below and are visually represented in an integrated framework.
Vegetation dynamics module
Biogeochemistry module Climate and weather module

Hydrology module Aquatic biodiversity module Terrestrial biodiversity module

1. Climate and weather 2. Hydrology 3. Biogeochemistry 4. Vegetation dynamics 5. Terrestrial biodiversity

Model coupling

6. Aquatic biodiversity

Data storage

Statistical data mining

Parameter optimisation

To our knowledge, no whole ecosystem model exists that adequately combines the elements of these six ecosystem components. To construct a new model that includes all of these elements would be both an enormous investment of resources and would only duplicate the existing capacity to model within each component. The modelling challenge is therefore to link existing models from each component into a single virtual model that can be used to assess whole ecosystem impacts. In this section (and associated tables in Appendix 1), we assess the modelling capacity in each of the six ecosystem components listed above. This assessment includes querying model availability, capability (i.e. what variables and processes are represented), quality (e.g. peer review, documentation availability), spatial and temporal resolution and data requirements. In evaluating each model under consideration we have attempted to cover some main points, such as whether a model has been validated by reproduction and/or approximation of observed results, the degree to which transparent testing and reporting of models has occurred, and whether a detailed description of model structure and parameters exists (Jakeman et al. 2006). We identified a large number of models in some modules and it was not possible to review them all. A shortlist of models for each component was compiled for more detailed review by an expert panel, according to personal experience, known rigour, plausibility according to previous applications and peer review, local and/or regional application within southeastern Australia and availability of personnel with relevant experience. The detailed criteria considered for each shortlist are presented in Appendix 1. The following table contains a summary some of the models assessed within each component (Error! Reference source not found.).

Page 5

Model Acronym / Model Name Global climate model currently under development for the 5 assessment of the IPCC.
th

Purpose CSIRO; BOM

Developer / Owner

CLIMATE & WEATHER

HYDROLOGY

Page 6
Simulate and predict mesoscale and regional-scale atmospheric circulation and predict weather. National Center for Atmospheric Research (NCAR), (Lo et al. 2008) Downscales the IPCC model outputs for high resolution regional (catchment-scale) hydrologic assessment (rainfall, temperature, evaporation). CSIRO; MDBA; BOM; DCC Flexible model for both quantity and quality of water from (nonurban) catchments to receiving waters. Uses SIMHYD or AWBM for the hydrological component. eWater CRC, (Argent et al. 2009); Replacement for E2 Daily conceptual rainfall-runoff model (Chiew et al. 2002; Kandel et al. 2005) Monthly time-step water balance modelling, feeding into a flow and salinity model from Hume dam to the Murray mouth. Catchment water balance model with hourly/daily rainfall-runoff simulation. MDBA (Boughton 2004) Modelling catchment hydrology, particularly after vegetation impacts. Developed and applied locally. (Peel et al. 2003; Watson 1999; Watson et al. 1999) Hydrological model that helps to define the surface and subsurface movement of water and nutrients in a catchment, and evaluate the impact of different farming systems and land management strategies on vegetative growth and productivity, stream quality, streamflows and groundwater. DPI Victoria; (Littleboy et al. 1992; Weeks et al. 2008)

ACCESS

Australian Community Climate Change Earth System Simulator

MM5 WRF

The Fifth-Generation NCAR / Penn State Mesoscale Model (MM5) which became the Weather Research & Forecasting Model

Table 1 Summary of component models

SEACI downscaling of IPCC 4th Assessment models

Intergovernmental Panel on Climate Change models modified by South East Australian Climate Initiative downscaling

WaterCAST

Water Contaminant Analysis and Simulation Tool

SIMHYD

Simplified HYDROLOG model

MSM-BIGMOD

Monthly Simulation Model BIGMOD

AWBM

Australian Water Balance Model

Macaque

PERFECT (in CAT)

Productivity, Erosion and Runoff Functions to Evaluate Conservation Techniques (in the Catchment Analysis Tool)

Model Acronym / Model Name Simulates terrestrial ecosystem production and soil microbial respiration. Includes global soil emissions of nitrous oxide and carbon dioxide. Recent versions are coupled to a Dynamic Global Vegetation Model (DGVM). NASA; (Potter et al. 2001; Potter and Klooster 1999; Potter et al. 1993)

Purpose

Developer / Owner

CASA (to be coupled with CABLE)

Carnegie Ames Stanford Approach Biosphere model

CENTURY (v4) Simulate plant-soil carbon and nutrient dynamics for different types of ecosystems including grasslands, agricultural lands, forests and savannas, capable of simulating detailed daily soil water and temperature dynamics and trace gas fluxes (CH4, N2O, NOx and N2).

DAYCENT

CENTURY Soil Organic Matter Model Environment modified for a daily time step

Colorado State University; USDA-ARS; (Parton et al. 1988)

CMSS Predicts impacts of nutrient management on water quality.

Catchment Management Support System Supports constituent generation for sediment, nitrogen, phosphorus and litter. Predicts runoff and total suspended sediment on a daily timestep Modelling framework that integrates SedNet, enabling calculation of average annual sediment and nutrient loads. Regional sediment and nutrient budgets for river networks. Spatially accounts for sediment and nutrient stores, sources and fluxes. Hydrological salt, sediment and nutrients transport model.

eWater, (Davis et al. 1991), www.toolkit.net.au/cmss

WaterCAST

Water and Contaminant Analysis Simulation Tool

eWater CRC,(Argent et al. 2009); Replacement for E2

EMSS

Environmental Management Support System

(Vertessy et al. 2001)

BIOGEOCHEMISTRY

CatchMODS

Catchment Scale Management Of Diffuse Sources Model

iCAM, icam.anu.edu.au/products/catchmods.html, (Newham et al. 2002) eWater CRC, CSIRO Land and Water, (Prosser et al. 2001a), www.toolkit.net.au/Tools/SedNet CWR, University of Western Australia; (Sivapalan et al. 1996a; Sivapalan et al. 1996b; Sivapalan et al. 1996c) (Coleman and Jenkinson 1999)

SedNet

LASCAM

Large Scale Catchment Model

RothC

Rothamsted Carbon Model

Simulates turnover of organic carbon in soils. Calculates total organic carbon, microbial biomass carbon and 14C over timescales up to centuries. Simulates carbon and nitrogen biogeochemistry in agricultural systems. Primarily a nitrogen cycling and distribution model, also models hydrology and channel routing, sedimentation, crop growth.

DNDC

DeNitrification-DeComposition Model

(Li et al. 1992a; b)

SWAT (in CAT)

Soil & Water Assessment Tool (the nutrient submodel in Catchment Analysis Tool)

DPI Victoria; (Weeks et al. 2008) (originally USDA)

Page 7

Model Acronym / Model Name Generalised stand model, for relatively even-aged homogenous forest or plantations. 3-PG calculates the radiant energy absorbed by forest canopies and converts it into biomass production, models water and biomass/carbon (modified for modelling carbon balance specifically). (Battaglia et al. 2004; Landsberg and Waring 1997)

VEGETATION

Page 8
Purpose Developer / Owner Calculates carbon, water and heat exchanges between the land surface and atmosphere and is suitable for use in climate models and in the form of a one-dimensional stand-alone mode. (Kowalczyk et al. 2006) Competition-based growth model (growth dependent on size and distance of competitor trees), includes a random mortality ftn, recruitment/new trees need to be specified/told to occur, although can be linked to a Leaf Area Index (LAI) threshold (Crookston and Dixon 2005) Spatial forest landscape disturbance and succession model. (Mladenoff 2004) Cell-based individual tree forest gap model. (Botkin et al. 1972) Individual tree-based forest gap model. (Pacala et al. 1996; Pacala et al. 1993) A Dynamic Global Vegetation Model (DVGM) Potsdam PIK (should have source for versions also) A Dynamic Global Vegetation Model (DVGM) (Krinner et al. 2005)

3-PG (and CABALA)

Physiological Principles Predicting Growth (and the CArbon BALAnce model based on this)

CABLE

CSIRO Atmosphere Biosphere Land Exchange

FVS

Forest Vegetation Simulator

LANDIS

JABOWA

Janak-Botkin-Wallis

SORTIE

LPJ

Lund-Potsdam-Jena

ORCHIDEE

Organizing Carbon and Hydrology in Dynamical Ecosystems

Model Acronym / Model Name Purpose-built statistical distributional models based on bioclimatic envelopes, topography, and habitat characteristics (e.g. tree species, habitat structure [tree spacing, size]) Many authors in Australia, and world-wide

Purpose

Developer / Owner

Species-specific and guildbased distribution models

TERRESTRIAL BIODIVERSITY

VORTEX, MARXAN, RAMAS, CIRCUITSCAPE Application of birth-death-emigration-immigration simulations to species population dynamics, including spatially explicit variation in population parameters

Spatially explicit dynamic demographic models

Many authors in Australia, and world-wide

AUSRIVAS Rapid prediction system used to assess the biological health of Australian rivers. Focused on physical assessment and bioassessment of streams. Includes predictive modelling software for macroinvertebrates.

Australian River Assessment Scheme

eWater CRC; DEWHA; State government departments

Ecological Modeller Statistical modelling linking eWater hydrological tool outputs to ecological processes (e.g. fish spawning). In development.

eWater CRC

AQUATIC BIODIVERSITY

FILTERS

An aquatic bio-assessment predictive model for deriving reference conditions without the use of reference sites. More suitable for use in disturbed sites than AUSRIVAS.

MDBA/SRA; (Chessman and Royal 2004)

CAEDYM

Computational Aquatic Ecosystem Dynamics Model

Generic aquatic ecological model designed to be linked to hydrodynamic models (e.g. ELCOM, the Estuarine and Lake COmputer Model), includes biogeochemical cycling and phytoplankton. Options for limited other biological options.

CWR, University of Western Australia

Page 9

2.1. Climate and weather models


The climate data and predictions most commonly utilised by the natural resources management community in Australia are model outputs published in the 2007 Fourth Assessment Report on Climate Change by the Intergovernmental Panel on Climate Change (IPCC). Since the release of the Fourth Assessment Report, much work has been done on downscaling model outputs to produce regional climate and weather data.

2.1.1. SEACI
In its first operational phase (2006 2009), the South Eastern Australian Climate Initiative (SEACI) addressed a range of policy-relevant climate and weather research questions, pertaining to climate variability and drivers of climate change, in order to develop improved regional climate information for south-eastern Australia. SEACI is managed by the Murray-Darling Basin Authority and involves the Victorian Department of Sustainability and Environment, Department of Climate Change, Managing Climate Variability Program, CSIRO and the Bureau of Meteorology. SEACI utilised six models from the IPCC Fourth Assessment report as the basis of its major climate forecasting projects, selected for their ability to accurately model historical records (Howden et al. 2008; Timbal and Jones 2008). Among the many outputs of this program, the initiative produced daily 110-year long time series data using statistical downscaling, based on four emissions scenarios. These were used to generate high resolution rainfall, temperature, evaporation and water balance projections for the whole Murray-Darling Basin, as well as specific water balance projections for catchments in the southern Murray-Darling Basin. SEACI has also produced improved techniques for seasonal forecasting in the Murray-Darling Basin. These techniques are used to produce probabilistic forecasts of seasonal rainfall and temperature, with some work done to integrate these forecasts with management. The independent mid-term review of SEACI identified that the individual project components did not effectively interact, such that several projects researched global climate model downscaling using different methods.

2.1.2. ACCESS
Currently under development, the Australian Community Climate and Earth-System Simulator (ACCESS) will couple both climate and earth-system simulation to enable improved meteorological forecasting and prediction of climate scenarios over a 50+ year timeframe. The first phase of ACCESS is to enable meteorological forecasting for Australia. The second phase is to complete a physical global climate model (GCM) that includes earth-system simulation; with the aim of including model outputs in the IPCC Fifth Assessment Report. Once developed, ACCESS (and the IPCC Fifth Assessment database) with be the model of choice for obtaining climate forecasts for south-eastern Australia. There is potential to use ACCESS global climate outputs using downscaling models such as the Weather Research and Forecasting (WRF) model (Lo et al. 2008), however in the first iteration of this project, given the IPCC Fifth Assessment modelling is still being developed, we will use the output from the IPCC Fourth review, which has already been downscaled as part of SEACI.

2.1.3. Summary
The statistically downscaled regional climate data and seasonal forecasts produced by SEACI are ideal inputs to a whole-of-ecosystem model, as described in this report. The large research effort put into SEACI, as well as the peer-review of project outputs gives strong incentive for utilising this climate data in this project.

Page 10

The seasonal forecasts produced by the Bureau of Meteorology can provide up to nine months of weather projections (rainfall and temperature) across south-east Australia with a higher degree of certainty than downscaled global climate models. With regard to producing a whole ecosystem simulation model, using SEACI data has the disadvantage that there is no opportunity to input into the climate and weather forecasts. For example, changes in water vapour flux and albedo from disturbance of forested areas, as would occur from fire, have the potential to impact significantly on regional climate.

Page 11

2.2. Hydrology models


A large number of hydrological models have been developed for simulating the flow of water through a catchment over the past 40+ years. A number of these have been specifically developed for Australian conditions, and are in wide use here, with these models having more of a focus on water yield, as opposed to the flood prediction focus of much of the rest of the world. The subset of Australian models alone remains unwieldy, with at least 100 models developed (Boughton 2005; Ranatunga et al. 2008). However a number of reviews (Boughton 2005; Boughton 1988; Croke and Jakeman 2001; Marston et al. 2002) have begun to indicate that hydrological modelling is mature, and that there is a consensus that input data (rainfall distribution) is a more significant factor than the specific hydrological model which is used, producing similar output from a range of models (Boughton 2005), however, although a number of hydrological models have been applied in the Murray-Darling region, a specific comparison of the results has not been made. Three main approaches to hydrological modelling have been identified: 1) simple rainfall-runoff modelling; 2) physically-based hydrological process modelling; and 3) modular systems. An example of each type is discussed in more detail. The models discussed in this section are currently under continuous use and development by the Australian hydrological modelling community.

2.2.1. Simple rainfall-runoff models


The simplest hydrological models are rainfall-runoff models that typically have only a few key input parameters, such as precipitation, evaporation and some form of catchment parameter (e.g. permeability), and can simulate surface runoff and baseflow (Boughton 2005). More advanced models can simulate storages for soil moisture and groundwater, as well as interflow between stores. WaterCAST (previously known as E2, applications and validation of E2 are thus considered here) was developed by the eWater CRC for predicting and managing the quantity and quality of water resources in a catchment (Argent et al. 2009). WaterCAST can use SIMHYD or AWBM for the hydrological component (Chiew et al. 2002). SIMHYD utilises seven input parameters to simulate surface runoff, baseflow and interflow and has previously been used to investigate climate change impacts on runoff (Boughton 2005; Jones et al. 2006). However, these are not physically-based parameters and therefore need extensive calibration data. The Australian Water Balance Model (AWBM) is a simple three parameter catchment water balance model that can calculate runoff from rainfall, evaporation and a baseflow index (Boughton 2004). Coupling WaterCAST to models outside the eWater Toolkit (described at www.toolkit.net.au) requires rescaling of upstream weather from the spatial grid used to the subcatchment units (the basic functional spatial unit used in WaterCAST). WaterCAST already includes a pre-processing plug-in to perform a similar rescaling for SILO (the Australian historical meteorological database) data, and this could be used for other weather projections (e.g. SEACI) once appropriately formatted. The WaterCAST soil moisture component has not been reliably validated and output for the vegetation module would need to be carefully considered. To fully couple these modules, additional code would be needed to insert the vegetation model evapotranspiration output to replace the WaterCAST evapotranspiration component (which relies on a simple calculation based on potential maximum evapotranspiration and the current soil moisture). Again, rescaling would be required. Sediment (constituent) generation from the sub-catchments is simply and empirically defined, requiring data on sediment erosion rates and nutrient concentration in eroded soil and calibration. Unlike most other hydrology models available, WaterCAST includes a riparian buffer filtering (of particulates and nitrogen) component, which would be potentially useful. Additionally, catchment vegetation models are also not concerned with this aspect of a catchment, and significant work would have be required to adapt any vegetation models to the higher resolutions and additional processes of interest in riparian zones. Note however, this component is largely untested. WaterCAST (E2) has previously been used to examine the impact of fire on water quality (Feikema et al. 2005). The 2003 pre- and post-bushfire water quality data was used to calibrate constituent generation rates for each subcatchment unit, and these were then used to predict the long term average changes Page 12

(increases) in loads (TN, TP and TSS) at specific points of interest. Subcatchments of 20-50km were used, and detailed subcatchment rainfall data was used, but note this model application was not process-based, and changes in runoff generation are not accounted for, nor are gradual changes such as recovery of the catchment. Extending this application and addressing some of these limitations (in particular the lack of changes in runoff, static catchment vegetation, etc.) would be of significant benefit to a more comprehensive assessment. MSM-BIGMOD (a Monthly Simulation Model with a daily timestep modification) is another key example of a simple flow and salinity routing model that has been applied to the Murray-Darling Basin. However, it does not model catchment runoff processes, which would be required for full interaction with the biogeochemistry and vegetation components (Ravalico et al. 2007). Given its development and application in the region of interest, some of the previous modelling is likely to be of use in parameterizing and validating any integrated process-based models. Similarly, other Murray Flow Assessment Tools for floodplain and wetland vegetation, fish, birds and algal risk are also relatively simple flow preference curves (Young et al. 2003), but some of the information used may be of use in parameterizing the process-based models proposed here.

2.2.2. Physically based hydrological process models


More complex hydrological models include detailed representations of biophysical processes, but require greater validation of model parameters. Such physically-based hydrological models are suited for predicting the hydrological consequences of disturbances such as deforestation and bushfire (Lane et al. 2007). Macaque is an example of a physically-based model, which was developed specifically to look at water yield impacts in forested Victoria catchments, where parameterization is based on direct measurements of catchment parameters (Lane et al. 2007; Watson 1999). Macaque is particularly suited to the upland catchments of the southern Murray-Darling Basin where mountain ash forests dominate, as the model was developed based on the observed changes water yield seen in the Victorian central highlands after the 1939 bushfires (Lane et al. 2007). Macaque has been used to examine the long-term (250 year) impact of the 2002-2003 bushfires on catchment water yield, however, this application did not look at sediment or nutrients (Watson et al. 1999). One of the core parameters represented in Macaque is leaf area index (LAI), which is a measure of leaf surface area for a given area of ground surface and is used to determine retention of rainfall and potential evapotranspiration by forests of different age structure. LAI can be physically measured using plant canopy analysers, hemispherical photography or remotely using satellites, in order to validate this model parameter (Watson et al. 1999). LAI would be a primary intersection when coupling hydrology with the vegetation module.

2.2.3. Modular systems


The Catchment Analysis Tool (CAT) links individual component models into a single hydrological model (Weeks et al. 2008). CAT has subsequently been utilised as the basis of the Catchment Modelling Framework (CMF), which is described later in section 3.1.4. The processes represented in CAT predominantly include land use (crop growth, forest growth, grazing) and hydrology (water balance, groundwater). The hydrology component of CAT is based on the Productivity Erosion Runoff Functions to Evaluate Conservation Techniques (PERFECT) (Littleboy et al. 1992), with additional water balance improvements adapted from the Soil Water and Assessment Tool (SWAT). CAT also includes an option to link surface and subsurface flow modelling with a spatially distributed groundwater model, MODFLOW (McDonald and Harbaugh 1988; Weeks et al. 2008). The authors of CAT note that the performance of the model (and this type of landscape model) is constrained by the availability of suitable validation data (Weeks et al. 2008). This is a serious concern for the approach described in this report, and is discussed in more detail in section 4.2. Page 13

The Catchment Modelling Toolkit is another collection of individual models with some potential for linkage, with a strong hydrological component, including WaterCAST as described in section 2.2.1 (Argent et al. 2009).

2.2.4. Summary
The hydrological models described use core hydrological parameters such as rainfall and evaporation in order to model runoff for a given catchment area. Beyond these basic functions, for which many of the models are likely to produce similar outputs, some models represent more complex physically-based water balance concepts for different forest types or for agricultural applications. Of the three models described, WaterCAST (E2) has the most published applications in Australia. Additionally, the previous application of WaterCAST in examining fire impacts provides an analogue for large scale modelling of bushfire impacts on hydrological parameters. Although Macaque has been applied for longer term water yield impacts, overall it has less testing and general validation. Additionally, integration of WaterCAST with vegetation models will allow similar modelling to the Macaque study to be performed. CAT may be the most difficult of the models to evaluate, although many of its components are based on well validated models, the overall system does not appear to have been peer reviewed. The fact that it already integrates a number of modules of interest (hydrology, biogeochemistry, vegetation) could be a significant benefit and is discussed further in the section on component-based modelling systems.

Page 14

2.3. Biogeochemistry models


Most models do not solely simulate biogeochemistry (Ciais et al. 2001; Moldan 1994). Models including biogeochemistry are usually 1) part of hydrological models, and focused on nutrient transport and export to the waterways (e.g. E2/WaterCAST), or 2) vegetation models focused on yield/growth of vegetation/agriculture/forests (e.g. CENTURY). Most vegetation models include at least one potentially limiting nutrient, and many of the hydrology models discussed in Section 2.2 include transport of nutrients through the catchment. Some of these models are discussed in more detail below, with additional details in Table 6 and Table 7. Although there is a general consensus that carbon and nitrogen are the minimum requirements for examining ecosystem nutrient impacts, phosphorus is also increasingly regarded as necessary, particularly when the impact on aquatic systems is considered (Townsend et al. 2008). The level of detail needed to represent these nutrients at an appropriate level also needs to be considered, and depends on the aims of the modelling exercise. In terms of a whole-of-ecosystem model, it is appropriate to model total N and P only (as some models do), given the importance of inorganic forms to primary production in the catchment and in the waterways. There remain many models which examine the breakdown of organic matter and cycling of nitrate/ammonium/phosphate. Although not intended to immediately couple fully with the climate model, nitrous oxide and methane generation may also be of interest in this module. Our understanding of biogeochemical cycling in the soil is quite well developed (Townsend et al. 2008), but different models place their emphasis on different parts of this cycling, representing processes at different levels of detail. Weathering of bedrock and soil sediments is a primary source of phosphorous, while atmospheric carbon dioxide and nitrogen fixed by plants and the primary source of carbon and nitrogen in terrestrial ecosystems. Further cycling within the system occurs as organic matter (mainly plant material) on and in the soil decomposes. This organic matter is generally taken as being composed of 1) easily decayed matter (typically labelled as labile, fast, active, etc.), and 2) matter which is less easily broken down (typically labelled as resistant, refractory, slow or passive). These decompose at different rates according to first order kinetics, and turn into CO2, inorganic nutrients (primarily NH4, NO3 and PO4), microbial biomass, humus, and very resistant organic residue (inert/passive). Temperature, moisture and soil texture affect soil nutrient processes. However, quantification of these processes is difficult (not least because of heterogeneity). Nutrients are rarely modelled conservatively, and uptake and partitioning of nutrients by vegetation is usually quite simply and empirically represented. Quantification and modelling of the link from nutrients in the catchment to waterways is especially poor, particularly of dissolved nutrient (and sediment) removal in riparian zones (Drewry et al. 2006).

2.3.1. Export models


There are a number of models which look at export of nutrients from the catchment to the waterways. These include: CMSS (the Catchment Management Support System), which is essentially a database of nutrient generation rates for different land uses and calculates nutrient loads from the entire catchment (Marston et al. 1995). It is an empirical model and does not model hydrology or nutrient cycling processes. CatchMODS (Catchment scale Management Of Diffuse Sources) has a finer spatial resolution, capable of resolving subcatchments, but it simulates TN and TP only, and has some significant limitations/assumptions (e.g. TP is only transported adsorbed to sediment), and a coarse temporal scale (annual only) (Newham et al. 2002). EMSS (the Environmental Management Support System), models daily loads at a subcatchment scale, however, there are apparently problems at this temporal resolution, and aggregating the daily output into monthly loads is recommended (Drewry et al. 2006). The E2/WaterCAST constituent generation component uses empirical relationships to predict nutrient loads based on past data and generation relationships. It should also be noted that WaterCAST does include a component for nutrient/sediment removal by riparian buffers, but this is untested. Drewry et al. 2006 discuss some of the complications associated with modelling riparian processes from a biogeochemical perspective (e.g. limited decreasing effectiveness with time/lifespans, dissolved vs. suspended forms, N vs. P). Page 15

SedNet is another model in the Catchment Modelling Toolkit. It was developed by CSIRO Land and Water and supported by the CRC for Catchment Hydrology. It creates nutrient (nitrogen and phosphorus) and sediment budgets from catchment to waterway (Prosser et al. 2001b), explicitly representing catchment erosion processes (e.g. hillslope erosion vs. gullying) and also accounting for sediment deposition on floodplains and within the stream network. However, although daily loads can be disaggregated from the mean annual timestep (and annual budgets produced), the developers actually recommend budgets are averaged over long time periods (20 years) to identify spatial patterns and long term trends and impacts, so daily outputs would need careful validation and interpretation before passing to another model.

2.3.2. Biogeochemical cycling models


RothC, the Rothamsted Carbon model, was originally developed for arable soils, but it has now been applied to grasslands and forests, however this focuses on carbon cycling only and uses an annual timestep, which is not at an appropriate temporal resolution for simulation of whole ecosystem processes (Coleman and Jenkinson 1999). DNDC is a denitrification-decomposition model, which models in detail the carbon and nitrogen cycles, however it does not include phosphorus (Miehle et al. 2009). The Soil and Water Assessment Tool (SWAT) is in use by the DSE, and is the basis for the biogeochemistry module in CAT. Carbon fixing via primary production is modelled for different crops, pasture, forest and native vegetation. Carbon allocation occurs to live, senescing, dead, litter, above and below ground pools. Four primary nitrogen processes are modelled, mineralisation, nitrification, volatilization and denitrification. Two inorganic pools (nitrate and ammonium) and three organic pools (fresh, active humic, stable humic) are used. Phosphorus is not currently modelled. Additionally, as with most vegetation models, growth does not appear to be linked to explicit uptake of nut rient content in the soil and instead uses a soil fertility parameter which needs to be calibrated for different sites and vegetation types (e.g. in the 3PG submodel). CENTURY is a general plant-soil-nutrient dynamics model with components for soil organic matter, the water budget, grassland/crop production and forest production, and was developed by the Natural Resources Ecology Laboratory at Colorado State University (Parton et al. 1988). It is freely available (downloadable online), and has been used widely (including sites in Victoria, South Australia and Queensland) for different soil and vegetation types. It simulates C, N and P (as well as S), and a newer version, DAYCENT, has been adapted to use a daily rather than original monthly timestep. It has been applied at a range of resolutions, down to 1kmx1km and daily timesteps). However, the quite detailed process representation requires a large number of parameters to be defined.

2.3.3. Summary
Conceptually, biogeochemical models take inputs (from atmosphere and bedrock), cycling (via weathering, leaching, uptake and return by plants/organisms in the soil) and provide outputs (partitions of nutrients within the system and export back to the atmosphere and to the waterways) of nutrients from catchment soils, as affected by parameters such as moisture and temperature. The models differ with regard to which cycles are regarded as important and are thus represented, how empirically they are represented and the amount of detail used. A particular difference is the detail with which vegetation is modelled, having significant impact on nutrient inputs (fixing) from the atmosphere, uptake, and cycling. It should be noted however, that reviews by (Drewry et al. 2006) and (Letcher et al. 2002) indicate that modelling of the physics of nutrient export from catchments may be inappropriate for Australian catchments where data is more sparse. Some of the more complex models require significant parameterization and calibration (e.g. LASCAM, requiring 18 parameters for N alone, 11 for P, and 6 for sediment). Simpler empirical models of export may produce more accurate results, however this may be of less use when examining future scenarios (outside the observations of historical data), and in research into the processes of interest for management.

Page 16

Additionally, many of the more sophisticated developments in modelling of nutrient processing in catchments are not yet incorporated into the general models, despite showing significant improvements e.g. for catchment DIN uptake (Wang and Linker 2006).

Page 17

2.4. Vegetation models


A large range of vegetation models have been developed, with the primary objectives of: 1) modelling projections of vegetation distribution spatially; and 2) modelling growth behaviour within vegetation stands, aspatially (Port and Bartelink 2002). Many of these models have been developed for plantation industries to predict vegetation growth/yield for productive forestry. The models under consideration are generally hybrid models, being primarily process -based, but combined with empirical representations (Miehle et al. 2009). This is generally necessary due to measured data being very limited for many of the detailed processes of interest (e.g. with allocation of biomass to different parts of each plant). Recent modelling has also combined a process-based stand model with a mechanistic model of fire spread, but this requires a large amount of parameterization (Perry and Enright 2006) and significant computational power (He and Mladenoff 1999) which would only be exacerbated with further model coupling. Modelling fire impacts on vegetation as a series of disturbances impacting forest composition and age structure, as has been done for forests in China and the USA, may be more practical (He et al. 2008). Areas that are not yet well addressed by modelling include the impacts of fire on understorey vegetation, and also where low- and moderate-intensity fires inflict varying degrees of damage to general vegetation processes involving water and carbon (Beringer et al. 2002).

2.4.1. Vegetation distribution models


Vegetation distribution can be modelled at either a broad landscape scale or at a finer plot scale (Perry and Enright 2006). We discuss one landscape-scale model, LANDIS, and two plot-scale models, JABOWA and SORTIE. The LANDIS model represents vegetation disturbance and succession on a landscape scale, using a cell-based grid with each cell containing information on tree species age classes, as well as spatial process such as seed dispersal and fire (Mladenoff 2004). JABOWA (acronym derived from authors names) is a spatial forest simulation model, originally designed for use in the Hubbard Brook Ecosystem Study, now with many derivatives for local forest simulation applications (Botkin et al. 1972; Bugmann 2001). JABOWA contains three main sub-routines for modelling tree growth, death and establishment, which use cell-based plots as the spatial unit. While the model represents stands of trees spatially across a landscape, there is no interaction between the patches (Bugmann 2001). SORTIE is a spatial model (derived from JABOWA) that represents long-term dynamics of forest communities using mechanistic sub-models of individual tree growth and competition (Pacala et al. 1996; Pacala et al. 1993). Sub-models include growth, mortality, recruitment and resources, which together make up a population dynamic model. SORTIE differs from JABOWA in that individual trees each have a unique spatial location, rather than using spatial cells. This is important, as one of the main driving factors of SORTIE is competition for light; however, this comes at a high computational cost (Bugmann 2001).

2.4.2. Stand growth models


Stand growth models represent forest growth dynamics aspatially within stands. Heavy calibration may be required for different forest types; for example, lots of work has been previously done on Eucalyptus globulus (Miehle et al. 2009), while there has been relatively little done on remnant forest. A review of forest succession models provides useful background information on the models described in this section (Taylor et al. 2009). The Physiological Principles Predicting Growth (3PG) model is a generalised stand model, which can be used to estimate carbon production from photosynthetically active radiation (PAR) received at the forest canopy, stand age, soil moisture and atmospheric vapour pressure (Landsberg and Waring 1997). Page 18

3PG needs to be parameterised for specific species and is intended for relatively even-aged homogenous forests or plantations (Landsberg and Waring 1997; Nightingale et al. 2008). It is included as part of the Catchment Assessment Tool (other components of CAT are described in other sections). The Carbon Balance model (CABALA) is a modification of 3PG developed by CSIRO, used to estimate the tree growth and carbon sequestration in plantations and managed forests. Model inputs include rainfall, temperature, salinity, water table depth and tree species data, which CABALA can use to estimate biomass production, carbon sequestration, nitrogen content and canopy height of trees in plantations and forests (Battaglia et al. 2004). The Lund-Potsdam-Jena (LPJ) model is a dynamic global vegetation model that represents largescale vegetation dynamics and land-atmosphere exchange of carbon and water (Sitch et al. 2003). LPJ uses a modular framework to link vegetation dynamics with land-atmosphere interactions. LPJ defines plant functional types to generically represent the different vegetation types found globally. The CSIRO Atmosphere Biosphere Land Exchange (CABLE) model is a land surface simulation model that can be used independently offline or in conjunction with a GCM online (Kowalczyk et al. 2006). CABLE represents soil, vegetation and atmosphere interactions and can be used to calculate carbon, water and heat exchange and will eventually be implemented in ACCESS. The Forest Vegetation Simulator (FVS) is a stand-based vegetation model that can be used to predict individual tree growth and mortality after harvesting (Crookston and Dixon 2005). The Organizing Carbon and Hydrology in Dynamical Ecosystems (ORCHIDEE) model is a dynamic global vegetation model used to simulate surface-vegetation-atmosphere interactions (Gerten et al. 2008; Krinner et al. 2005).

2.4.3. Summary
The available vegetation models cover a wide variety of spatial scales and processes, many of which could potentially be used in an integrated modelling framework. Spatially explicit models are of most interest for landscape-scale modelling and models exist that can do this (e.g. LANDIS). However, many of the stand growth models can be interfaced with geographic information systems (GIS) to produce spatial vegetation models that can, at least, represent vegetation processes within cell-based landscape models. Individual tree-based models, such as JABOWA and SORTIE, while designed for small plot sizes, could potentially be used on larger scales if high performance computing resources are applied to the task. At the other end of the scale, vegetation models designed to model carbon dynamics and biomass production on global scales, such as CABLE, can be applied to landscape-scale simulation to produce coarse predictions of vegetation distribution and composition. These have the advantage that they are tightly coupled to global climate models, thus ideal for climate change research. However, the coarse-scale outputs of such models would not be of much use in modelling terrestrial biodiversity (as habitat input data) or for catchment-scale hydrological modelling using models like Macaque.

Page 19

2.5. Terrestrial biodiversity models


As with the other modules, there are many models available for examining terrestrial ecology (and specifically biodiversity), but as with the vegetation modelling (or even more so), there is a lack of convergence of accepted techniques and process-representation for simulating wildlife dynamics (Shifley et al. 2009). There are two major forms of terrestrial (animal) biodiversity modelling. The first deals with modelling of species distributions based on bioclimatic envelopes, topograp hy, soil types (as a surrogate for site productivity) and habitat characteristics (Ferrier et al. 2002). There are literally thousands of such models, but little overall consensus about which statistical methods should be used and how validation is to be conducted (Elith et al. 2006; Fleishman and Mac Nally 2007). A systematic approach to this is needed (Thomson et al. 2009). Some work uses guilds species that use similar resources (food, nesting sites, etc.) as the basis for models because often there are few data for many species of conservation concern (Mac Nally et al. 2008). The alternative approach is one that focuses on building spatially explicit demographic models. These typically aim to identify whether a particular species, usually represented as a metapopulation, is likely to persist given the spatial pattern of habitats of various value to the species, birth and death rates, and rates of movement across the landscape. Additional variables include simulated threats (e.g. habitat degradation, hunting) and species support (e.g. augmented food). Models include VORTEX, MARXAN and RAMAS. A combination of the two approaches is one that uses raster-based GIS data to infer metapopulation persistence (Drielsma and Ferrier 2009), although whether such models can be validated is at this time unclear. Being more specific about the goals of this module is essential for the suggested model development. We envisage attempting to make general predictions about change in biodiversity over decadal timescales on the catchment scale and gain insight into the theoretical basis and representation of system processes, to more solidly define what we understand about the system and identify the gaps and uncertainties in that knowledge, and finally to gain a better idea of the key requirements for long term monitoring to improve our knowledge in this area. Incorporation of fire impacts is likely to require the more process-based explicit demographic models to incorporate repeated disturbances and resetting of populations, superimposed on habitat suitability.

2.5.1. Summary
Terrestrial biodiversity models take inputs (from the climate module, temperature and rainfall), from the biogeochemistry module (for soil nutrients as a proxy for vegetation quality), from the vegetation module for vegetation characteristics (e.g. stand area/extent, age structure, size of trees, connectivity, etc.) to give as outputs, abundance of indicator species or a measure of biodiversity. There is currently little consensus in this area, and some development of the goals of this module and the modelling required to meet these will be necessary. A thorough review of current terrestrial biodiversity modelling is recommended as a first step.

Page 20

2.6. Aquatic biodiversity models


Certain parts of the state of modelling in aquatic ecology are relatively advanced. Previous discipline-based models in the fields of hydrodynamics, nutrient cycling and eutrophication (Reckhow and Chapra 1999), and biology (in particular, primary production and algal processes), have given way to relatively sophisticated models integrating these processes within aquatic ecosystems (Robson and Hamilton 2004). However, modelling of higher ecosystem trophic levels remains limited. In this section we look at two approaches: 1) the process-based integrated biophysical approach; 2) a statistical approach looking at measures of biodiversity; and 3) population models. As mentioned in 2.3, although sources of nutrients from the catchment to the waterways are known (e.g. P input to waterways is often primarily from bank and gully erosion (Drewry et al. 2006)), quantification is difficult and modelling of nutrient inputs can be poor. Both inorganic N and inorganic P can be the limiting nutrient in freshwater/estuarine systems. Land use associated with nutrient loss to waterways is summarised in the Nutrient Data Book (Marston et al. 1995).

2.6.1. Process-based ecosystem models


The process-based integrated biophysical approach is used by many models such as PROTECH (Reynolds et al. 2001), CE-QUAL (Wlosinski and Collins 1985), SALMO (Recknagel et al. 1995), and ELCOMCAEDYM (Robson and Hamilton 2004). We use ELCOM-CAEDYM (the Estuarine and Lake Computer Model with the Computational Aquatic Ecosystem Dynamics Model) as an example of this approach, having been developed in Australia, and incorporating a high level of ecosystem detail. Some additional details on ELCOM-CAEDYM are included in Table 11. Hydrodynamics, nutrient cycling and phytoplankton dynamics are simulated, with additional options for zooplankton, macroalgae, and sediment biogeochemical interactions, however higher trophic levels and populations of interest such as macroinvertebrates and fish are not simulated. As noted with detailed models in other modules, the level of detail involved requires a large amount of parameterization and calibration, and relatively large amounts of processing power. It should also be noted that the instream hydrodynamic part of the modelling (ELCOM) is modular and CAEDYM has previously been coupled with a range of 1-, 2-, and 3dimensional hydrodynamic applications up to very high resolution spatial scales (metre scales). It should be possible to couple simpler river routing models already specifically applied to the Murray-Darling Basin such as MSM-BIGMOD (Monthly Simulation Model with a daily timestep modification). Alternatively, such previous modelling may be of use for comparison and validation of flows or salinity transport.

2.6.2. Aquatic biodiversity models


A primary aquatic biodiversity approach used in Australia is AusRivAS, the Australian River Assessment Scheme (Reynoldson et al. 1997). This scheme compares the modelled distribution of invertebrates calibrated in so-called pristine conditions with the distributions in equivalent disturbed environments. This approach contains some difficulties in lowland areas of interest where pristine reference sites are often not available. There are also problems in using this as a predictive tool, either under a climate change scenario where reference sites will also be affected, or for fire impact scenarios, where the areas affected may be skewed toward the pristine reference sites. More recent invertebrate distribution models, Filters, have been developed for south-east Australia to address some of these difficulties (Chessman et al. 2008; Chessman and Royal 2004). These models use the tolerances or preferences for specific environmental factors (e.g. climatic, geomorphological and hydrological factors) as filters for the potential macroinvertebrate taxa inhabiting a certain site, identifying the range of taxa which might be expected at the site under natural conditions.

Page 21

2.6.3. Population models


Stochastic population models for species of interest (or indicator species) such as the Trout Cod (Maccullochella macquariensis) and Murray Cod (Maccullochella peelii peelii) have also been developed, using limited ecological data and understanding of temperature preferences, fecundity, spawning behaviour, age-specific survival and density dependence given variations over time (Todd et al. 2004; Todd et al. 2005). However in models such as this, there is usually limited validation of the faunal abundances due to limited data, although (as for the cod models) there can be a rigorous analysis of model behaviour/sensitivity including plausibility of outcomes.

2.6.4. Summary
The aquatic ecology models described examine inputs from the climate module (temperature and rainfall), from the combined hydrology, biogeochemistry and vegetation modules (for runoff, evaporation, sediment, nutrient and organic matter inputs) to give as outputs: aquatic variables (flow, nutrients, sediment), risk of eutrophication/algal blooms, abundance of indicator species or a measure of biodiversity. However, there is a significant gap between the process-based simulation of aquatic conditions and ecological/biodiversity measures of interest that is currently spanned by statistical/empirical relationships. Two models may be required in this module to cover the range of impacts in the aquatic system.

Page 22

3. Linking models together


The ability to link separate models together into a single virtual model is crucial to the development of a whole ecosystem model. There is increasing development of land surface models covering aspects of hydrology, biogeochemistry and vegetation, as these are strongly linked components of the system (Abramowitz et al. 2008). In this section, we outline a number of systems for model integration, as well as data management systems.

3.1. Modelling systems


3.1.1. The Nimrod toolkit
The Nimrod toolkit is a parametric modelling system, which can automate the running of software models by collecting and staging data, running and monitoring experiments and collecting outputs. The scheduling features of Nimrod allow models to be run on local computer networks and distributed across global computer grids (Abramson et al. 2000). The Nimrod toolkit was initially developed by the Distributed Systems Technology Centre (funded through the Australian Research Councils Cooperative Research Cent re program) and its continued development is led through Monash Universitys Faculty of Information Technology. The Nimrod toolkit can perform parameter sweeps (Nimrod/G), search the parameter domain using non-linear optimization algorithms (Nimrod/O) or can enact fractional factorial design (Nimrod/E). The most relevant tool for this assessment is Nimrod/K, based on the Kepler system (described in the next section), which can link different models together, optimize model parameters and spread computational load across global computing grids.

3.1.1.1. Nimrod/G Nimrod/G is a version of the Nimrod parametric modelling system that can take advantage of grid computing resources. Essentially, this version can distribute modelling tasks across multiple high performance computer resources across the global computing grid.

3.1.1.2. Nimrod/O Nimrod/O is a part of the Nimrod toolkit that uses a range of parameter optimization algorithms, which can be used to find parameter sets that give rise to a series of observed results (Abramson et al. 2001; Abramson et al. 2006). Parameter optimization can be computationally intensive, as a large combination of parameters can result from just a few model variables. As Nimrod/O is grid-enabled, different sets of parameter sweeps can be sent to clusters of processors on the grid, meaning that parameter optimization can be performed much more rapidly. The parameter domain (i.e. the full range of available parameters) can be specified and constraints can be set to define soft or hard limits to the range of parameters. A range of optimization algorithms are used to cover the parameter domain in different ways. Some will sample throughout the whole domain at a given resolution, while others will sample an iteratively finer grid around the best point from previous sweeps (Abramson et al. 2006). Other search methods include non-linear techniques and genetic algorithms, with the potential to add custom search algorithms. The user can also prioritize the model output of interest, so that the parameter set can best describe certain model outputs.

3.1.1.3. Nimrod/E When running a set of models using a large number of input variables, covering the full set of possible combinations (i.e. full factorial design) becomes impractical. Nimrod/E uses a fractional factorial design, Page 23

which reveals the most important interactions between parameters by using a subset of the full set of parameters. This means that a good approximation of the information that would be derived from a full factorial design can be achieved using a more practical number of model runs.

3.1.1.4. Nimrod/K Nimrod/K represents the grid workflow component of the Nimrod toolkit. The workflow engine on which Nimrod/K is built (Kepler) is described in the section below. Essentially, this grid workflow system enables different software models to be linked together into a single virtual model, with the ability to run locally on a desktop computer, or on a computing grid ranging from a grid of locally networked computers all the way up to a global computing grid. The Nimrod/K prototype has the ability to run models on the grid in a way that is parallel, distributed and dynamic. In other words, different software models can be run at the same time, from different locations and using data stored either locally or at another location on the grid, while still dynamically feeding model results back into new model runs. Nimrod/K can wrap around models to feed in inputs and take off outputs and stream them around to other models or statistical analysis tools. This can be conc eptualised as model plumbing, but in reality the system is more sophisticated, for example, enabling archiving of model connections for future reference.

3.1.2. Kepler
Kepler is a software application for retrieving input data from locally or remotely stored files and executing component models and performing statistical analyses on the retrieved data. Users can capture workflows, so they can be easily exchanged, archived, versioned, and executed. The workflow system is ideal for scientists with little background in computer science. The graphical interface allows components (i.e. models or data sources) to be dragged and dropped onto a Workflow canvas for connection and execution. Kepler supports foreign language interfaces via the Java Native Interface (JNI), so that component models written in different programming languages can be integrated. For example, Kepler includes the ability to execute Matlab scripts and R code. Kepler is thus used to tie together diverse computational systems into a unified framework. The flow of data from one analytical step to another is captured in a formal workflow language. Component models can either be loosely coupled, where each model runs on a single batch of data and the outputs are transferred to another model, or more tightly coupled, where a continuous stream of data outputs are fed into another model, allowing feedback mechanisms to occur at a high temporal resolution.

3.1.3. Interactive Component Modelling System


Developed by CSIRO Land and Water, Australian National University and Land & Water Australia, the Interactive Component Modelling System (ICMS) is a software modelling system targeted for use by catchment managers. It includes a number of simple models for rainfall-runoff, flow routing, crop selection and management, salinity and nutrients, which can be linked and executed with a graphical interface called the ICMSBuilder. ICMS has been designed for simple catchment representations and is not suitable for complex examinations and detailed spatially explicit modelling (Newham et al. 2004).

3.1.4. Catchment Management Framework


The Catchment Management Framework (CMF) is a modelling framework for connecting modelling tasks across different scientific disciplines. The primary model involved is the Catchment Analysis Tool (CAT), developed by the Victorian Department of Primary Industries, originally for farming systems, and incorporating and adapting a number of submodels such as PERFECT (Productivity, Erosion, Runoff Page 24

Functions to Evaluate Conservation Techniques, a crop production model), SWAT (the Soil and Water Assessment Tool), CERES-Wheat and CERES-Maize (Crop Environment Resource Synthesis), GRASP (a dynamic pasture model), 3PG (Physiological Principles Predicting Growth) amongst others. CMF enables coupling of CAT with models such as MODFLOW (MODular three-dimensional finite-difference ground-water FLOW model) and 2CSalt. CMF also includes a number of other modelling tools for data analysis and visualisation.

3.1.5. Ecological Modeller


Currently being developed by the eWater CRC, the Catchment Planning Tool is intended to be the interface between WaterCAST (discussed in section 2.2.1) and Ecological Modeller (previously called the Ecological Response Modelling tool). Ecological Modeller is a database system for running a range of ecological models using time series data of habitat (e.g. streamflow). Ecological Modeller allows the user to run different biodiversity models using common time series data for habitat conditions. For example, Ecological Modeller could model the range of a fish species in a lowland river given a set of hydrological conditions over time. Ecological Modeller is not a model in itself, simply a tool for writing and collating models or relationships, allowing easy comparison of output.

3.2. Data management systems


3.2.1. National Data Grid Demonstrator Project (formerly PEMS)
The National Data Grid (NDG) Demonstrator Project refers to the Cooperative Research Centre for Spatial information (CRC SI) project that builds on the earlier Platform for Environmental Modelling Support (PEMS) Demonstrator Project. The NDG is a system for managing grid-based spatial data, which can integrate different data sources into a shared, online database, using standard grids and projection systems. The NDG Demonstrator is proposed to be operational by the end of 2009.

Page 25

4. A conceptual framework for linking models


In this section, we present a generic conceptual framework for linking component models together that would be suitably flexible to address a range of natural resource management questions. The conceptual model presented details how different model variables connect to each other across discipline boundaries (Figure 2). Note that this is a generic diagram, and many of the models reviewed may cover only part of what is shown here (e.g. biogeochemistry models may not include phosphorus) or may simulate additional components not detailed (e.g. vegetation fragmentation effects on terrestrial ecology, more specific biogeochemistry compartments for cycling and species). More specific versions of this diagram should be constructed once specific requirements are detailed (e.g. we are interested in denitrification to nitrous oxide, or transport of phosphate to waterways), and again once specific models are selected for each module. The explicit process names for the linking arrows are not given in the diagram, but are described in the next section (4.1). However, this initial representation of the linkages between the six model components is the first major step in building a whole ecosystem model. To construct a conceptual model as a blueprint for building a whole ecosystem model would require a larger investment of resources.

HYDROLOGY CLIMATE

(DOWNSCALED WEATHER)

CLOUD COVER P.A.R.

RAINFALL

TEMPERATURE

AIR PRESSURE

HUMIDITY

WIND

SURFACE RUNOFF

SUBSURFACE RUNOFF
EROSION

EVAPORATION
CH4 CO2 GROUNDWATER

N2O

INTERCEPTION

LEAF AREA INDEX

SOIL MOISTURE EVAPOTRANSPIRATION SOIL TEMP SOIL TYPE

NITROGEN
NO2NO3NH4+

PRIMARY PRODUCTIVITY & GROWTH


ORGANIC INPUTS VEGETATION

Labile OM

Organic N Passive Organic Matter

Inorganic N

Resistant OM

Organic P

VEGETATION

FUNCTIONAL TYPE/EVC STAND HEALTH

Humic OM
Organic C Inorganic C Inorganic P

CARBON
NUTRIENTS

EXTENT STAND STRUCTURE

PHOSPHORUS

Inorganic C

FLOW
FOOD SPECIES X POPULATION HABITAT SUITABILITY IN-STREAM TEMP

Organic N Inorganic N Organic P

Inorganic P Adsorbed P

SPECIES Y POPULATION

SUSPENDED SEDIMENT

IN-STREAM P.A.R. ALGAE

NUTRIENTS

TERRESTRIAL ECOLOGY

METAPOPULATION DYNAMICS/PERSISTENCE

MACROINVERTEBRATES

Figure 2 A conceptual model of integrated modelling across ecological disciplines

Page 26

AQUATIC ECOLOGY

RIPARIAN VEGETATION

Organic C

4.1. Processes represented


The conceptual model aids in understanding the processes that can be represented within each component and how these processes interact. In this section we list the specific ecosystem processes we think should be represented in a whole-of-catchment integrated ecosystem model. These processes describe many of the links (arrows) in Figure 2, although some are not explicitly represented in the diagram. The more detailed models discussed in each module cover many of the processes listed below, however no single model in each module may cover all the processes mentioned. Additionally, the italicised entries are often not considered as important and are more rarely simulated. In combination with Figure 2, this list may be of use in model selection.

4.1.1. Climate and weather


Climate and weather parameters can be derived from data of solar radiation, albedo of land and sea surfaces, greenhouse gas concentrations and fluxes and land topography. The following parameters can be modelled: 1. 2. 3. 4. 5. Heat exchange/temperature Rainfall generation Air pressure (from heat exchange, Coriolis force, friction, topography wind) Humidity generation (from rainfall, evapotranspiration, water balance) Cloud formation (from humidity)

4.1.2. Hydrology
Hydrological parameters are mainly derived from rainfall and temperature (evaporation) data, with more complex models incorporating vegetation to calculate more accurate values for evapotranspiration, and land use to calculate permeability more accurately. The following parameters can be modelled: 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. Interception (from rainfall and leaf area) Infiltration (from interception and rainfall and soil chars) Saturation excess (from infiltration and rainfall) Surface flow (from saturation excess + runon from upslope) Soil moisture (from infiltration) Subsurface flow (from infiltration and soil moisture) Soil evaporation (from temperature and soil moisture) Plant uptake (from temperature and soil moisture and vegetation chars) Plant transpiration (from temperature and wind and vegetation chars) Drainage to groundwater (from soil moisture) Groundwater flow Erosion/transport (nutrients/sediment)

Page 27

4.1.3. Biogeochemistry
Biogeochemistry parameters, for both soil and hydrological components, can be defined by the element of interest; mainly carbon, nitrogen and phosphorus. In addition, temperature and soil moisture are important for modelling plant uptake and nutrient cycling processes. Transport of particles and dissolved salts is considered in the aquatic biodiversity section ( 4.1.6). The following parameters can be modelled: Carbon cycling 1. Fixing by plants 2. Breakdown of plant material/organic matter (potentially to labile and refractory, with further processing/decay for further breakdown of refractory into labile and refractory, etc., alternatively to structural and metabolic, active and passive) 3. Mineralisation 4. Leaching 5. Respiration 6. Fermentation to methane Nitrogen cycling 1. Fixation 2. Breakdown of organic matter 3. Mineralization (to ammonium) 4. Nitrification (to nitrite then nitrate) 5. Uptake by plants 6. Denitrification (to N2 and N2O) 7. Ammonification 8. Volatilization of ammonia 9. Adsorption of ammonia 10. Leaching Phosphorus cycling 1. Weathering of bedrock/sediment 2. Breakdown of organic matter 3. Mineralisation 4. Uptake by plants 5. Adsorption 6. Leaching

4.1.4. Vegetation
Stand growth models typically utilize light (photosynthetically active radiation or PAR) to drive tree growth, which can be constrained by parameters describing soil moisture, rainfall and stand age. Spatially distributed models take into account competition for light to differentiate tree growth for individual trees. Some of the parameters that can be represented by models are: 1. Photosynthesis 2. Growth leaf, stem, root allocation (giving outputs for stem size/basal area, crown height, stem density) 3. Competition for light 4. Limitation by nutrients/temperature/salinity/conditions 5. Litterfall 6. Mortality 7. Disturbance (including harvesting/logging, disease/pathogens and fire often impacts only modelled, not actual disturbance) 8. Seeding 9. Recruitment

Page 28

4.1.5. Terrestrial Biodiversity


Models of terrestrial wildlife dynamics based on bioclimatic envelopes can include a number of parameters related to habitat suitability. Demographic models can represent species birth and death rates. A summary of parameters that can potentially be modelled are as follows: 1. Habitat suitability and quality from climate (temp, rainfall, etc.), soil (nutrients, texture, etc.), topography (elevation, aspect, etc), vegetation (type, extent/patch sizes, tree size distribution, etc.) (statistical relationship for) extent/area suitable 2. Species (statistical relationship for) estimate of biodiversity 3. Species birth and death rates 4. Rates of species movement through landscape 5. Metapopulation persistence

4.1.6. Aquatic Biodiversity


Aquatic biodiversity (macroinvertebrates, fish, algae, etc) has strong conceptual links to water flow and quality, as well as productive factors such as light, oxygen, nutrients and carbon. Some of the parameters represented in aquatic biodiversity models include: 1. 2. 3. 4. 5. 6. 7. Physical flow/circulation (and effects on transport, nutrient distribution/processing) Suspended sediment and sedimentation Nutrient processing (largely as for biogeochemistry, including uptake for primary productivity) Oxygen dynamics (respiration/photosynthesis) Photosynthesis/growth (for phytoplankton from nutrients, light, temp) Photosynthesis (for macroalgae, macrophytes) (statistical relationship for) macroinvertebrates/other biodiversity.

4.2. Limitations
Obviously, there are major limitations on the degree to which some biophysical processes can be represented, either because of a lack of knowledge about the process, lack of data to parameterize the process, or purely because of limits to computing power and the time available. The ability to perform integrated modelling for a particular location depends on data availability.

1. Data limitations occur for initial conditions, for calibration of parameters, and for validation. These limitations may occur for a variety of reasons, including the long-term scale of interest, because were interested in scenarios such as climate change which we are not able to experiment with, or because some of the data which would be of use in validation is difficult/expensive to collect (e.g. physiological tree data, rare fauna surveys), 2. Natural variability can be immense. Spatial and temporal heterogeneity of physical factors such as soil characteristics across a catchment and sub-daily rainfall events would be easy to omit or misrepresent. There is possibly even greater variability in biological factors such as the growth or behaviour between different species and even of individuals within a single species, which may be represented by an average behaviour for a limited number of modelled types. 3. Traditional validation approaches may not be possible due to the large spatiotemporal extents under consideration (Oreskes et al. 1994). Lack of data is one of the primary motivations in building a large-scale, long-term integrated model (He et al. 2008). Techniques to analyse time series of spatially explicit data are also currently lacking (Perry and Enright 2006). It may be more appropriate to evaluate the integrated model in terms of how well (or plausibly) ecological processes are represented, and how useful the model is for hypothesis testing and learning about the modelled system (Perry and Enright 2006; Shifley et al. 2009).

Page 29

As with any simulation model, the proposed integrated model needs specific management questions to frame modelling tasks around. Given the specific management questions, the general approach outlined in this report is to make the integrated model as flexible as possible to address a large combination of NRM questions. Where models have been developed and applied only by a particular group, access to the model and the expertise needed to apply the model may be a limitation. Most of the models discussed in this report are not proprietary, or are available at minimal cost. An extended range of commercial simulation models are available (e.g. MIKE), but do not necessarily provide greater modelling capacity than those reported. It should be noted that the primary focus of this report is the long-term impact of fires, and does not look at modelling of fire or fire risk. Overall, fire modelling is a relatively recent field that is still under development and has some difficulties, particularly at the catchment scale (McKenzie et al. 1996), however there is the possibility of eventually coupling the integrated ecological model with some recent limited fire focused models which may contribute to our understanding of fire risk. For example, a recent process-based model (developed in Australia) could be linked to weather (precipitation, temperature) and outputs from the vegetation module (litter) to be used as inputs to simulate wetting and drying and fire risk to the fuel load (Matthews 2006).

4.3. Issues of scale


There are a number of issues of scale which will need to be considered in integrating models for each module. Specifically, the boundary conditions between models will need to be matched; a task that might be made easier with the development of the National Data Grid. The output of each module will need to be matched (and possibly up- or down-scaled) to match the input of the coupled models. Specifically, the anticipated rescaling required includes:

1. Use of climate data that has already been downscaled spatially and temporally to weather for the region; 2. Rescaling the weather data to the selected hydrological model spatial resolution (this may involve rescaling grid to sub-catchment units); 3. Rescaling of the hydrology output data (soil moisture where available) for the vegetation model in space (from sub-catchment or grid to patch or EVC) and from daily hydrological temporal scale to the vegetation model time-step (which may be monthly or yearly); 4. Up-scaling the vegetation output data spatially if trees are modelled individually to stand/patch or to larger units (possibly whole-of-catchment-scale) for terrestrial ecology; 5. Rescaling the hydrology, vegetation and biogeochemistry patches and/or grids for the boundaries of the aquatic ecology model to provide runoff, sediment, organic matter and nutrient inputs.

There is also a disjunct between the landscape vegetation scale (km scale) and riparian buffer zones (scale of metres), which may present problems for incorporating these important river-side zones into vegetation simulations. Rescaling and inter/extrapolation will be required for collected field data (e.g. weather stations, soil sampling) to match resolution required for model input data. There are also issues with the resolution of the temporal scale. Modelling of long-term forest growth often uses a time-step of more than a year (Crookston and Dixon 2005), while runoff is usually modelled on a daily time-step. However, many studies indicate that nutrient inputs into waterways can be significantly affected by storm events which may require a sub-daily resolution (Chessman 1986; Drewry et al. 2006).

Page 30

4.4. Issues of uncertainty


Any single simulation run of a model (an instantiation) is subject to many sources of variance. Uncertainties arise from these sources: Initial conditions and values of all variables, which can have profound effects (Lundberg et al. 2000; May 1976) Model functional relationships (what are the actual shapes of functional relationships between response and predictor variables?) Response and predictors may have different scaling in space and time, making their functional forms change at different scales Imperfect knowledge of functional relationships (variation explained in known relationships is usually much less than 100%) Model structure (are the important relationships included?) Non-stationarity relationships may change through time some relationships may become unimportant while others emerge.

These points have stimulated some ecologists to seek an alternative approach, which involves specifications of scenarios, imagined futures, for which key measures are assessed (Carpenter 2002). However, such approaches involve little incentive for on-going learning about knowledge gaps and refinement of important relationships. As we have seen with climate change modelling, it is crucial that numerical values be associated with forward projections not merely comparatively vague futures. Moreover, scenario methods cannot provide a pathway for informing management and policy about going from the present to the future because specific pathways need to be developed to do so. These comments indicate why we favour an approach in which we will build the complete system model, propagate uncertainties, and run the model many thousands of time to produce probability distributions for variables about which we are interested. This is linkable to risk-based assessment of various options that might be envisaged by any stakeholder group. Without a probability distribution, one cannot assess the likelihood that undesirable outcomes may emerge with higher-than-acceptable chances. We think that ensuring systems are managed so that they are bounded away from catastrophic results is a critical lesson to convey, but this needs the many-instantiations approach to make judgements.

Page 31

4.5. Issues of model integration


Linking models together is a challenging task, and although coupling of two models has been done relatively widely (Lynch et al. 2007; Robson and Hamilton 2004; Sherman et al. 2007), and there is some theory developed as to re-using, coupling and integrating models (Brandmeyer and Karimi 2000; Parker et al. 2002; Rizzoli et al. 1998) and conceptual work on whole-of-system integration (Gentile et al. 2001; Ogden et al. 2005), the larger scale of the project proposed here means the actual effort required, cannot be known until it is attempted. Some of the issues that may potentially arise in a model coupling exercise include: Data format spatial resolution, file formats, boundaries Model programming language Model licensing Increased processing and runtime requirements

The above points are certainly challenging, but are not in any way fatal to this exercise. The grid workflows technology, notably Kepler, includes actors that can reformat data streams into a large number of common file formats, with the capacity to also use custom formats. Kepler can also deal with programs written in different programming languages, as in many cases it launches the modelling software in its own environment. Programming languages only limit the degree to which any particular model can be altered in Kepler. Scripts written in Matlab or R can be natively implemented in Kepler. Licensing of models can be an issue, particularly where a licence is held on a USB memory stick. However, many of the models reviewed in this report do not have such strict licensing requirements. Distributed computing and increasing availability and access to supercomputer facilities as well as sophisticated data scheduling and management will help address the potential problems in passing large arrays of data back and forth between models.

Page 32

5. Program of work
The program of work required to build and run an integrated ecological model that can operate on a catchment scale is considered in this section. A number of important factors would firstly need to be addressed regarding the purpose and scope of the model. The framework proposed in this report, of using six model components to describe a whole ecosystem, provides flexibility to make specific model interconnections in order to address individual management and research questions and scenarios. An understanding of the feasibility of linking models and identification of modelling gaps would emerge from the production of a simplified draft model, which would be a useful first step towards a comprehensive whole system model.

5.1. Recommended approach


We recommend the following logical approach to scoping, designing and implementing the modelling framework outlined in this report, using the case study of upland bushfires and their impacts on catchmentscale ecological processes.

5.1.1. Define specific research questions and goals


In the first instance, specific queries related to more general management goals need to be formulated to input into the model. In the case of studying bushfire impacts, specific queries may be produced from more general management questions as shown in the flowchart in Figure 3. For example, many vegetation models do not directly simulate fire, but can simulate what will happen following a disturbance that removes mature vegetation, such as logging or fire.

Large scale f ires in uplands

Impacts of f orest disturbance

Erosion

Forest regrowth

Multiple f ires

Water quality

Habitat availability

C production

Impacts on water yield

Community composition

Aquatic biodiversity

Figure 3 Model query flowchart

Page 33

5.1.2. Select a case study catchment.


Applying a linked ecological model to specific catchments within the Murray-Darling Basin is the most practical way to implement this assessment. Operating an ecological model at a catchment scale provides a number of advantages, including: data availability, consistent boundaries for input data, alignment with waterways and alignment with existing management boundaries. For example, more than two decades of ecological research has taken place within the boundaries of the Goulburn-Broken Catchment in Victoria. Figure 4 shows the geographic distribution of published ecological research projects within the boundary of the Goulburn-Broken Catchment. As can be seen from the map, a large amount of aquatic ecology research has been conducted in the fire-affected Acheron River catchment.

Figure 4 Location of published ecological study sites in the Goulburn-Broken Catchment

Page 34

5.1.3. Identify data requirements


A review of ecological studies within a catchment is a useful first step for identifying potential sources of data. Information on ecosystem processes can be derived from these studies and used to parameterize an ecosystem model. Figure 5 shows the breakdown of ecological research themes within the Goulburn-Broken Catchment. In addition, input data requirements will need to be established for each individual model. Input data usually needs to be formatted, which the Kepler system can be setup to perform automatically through data filters. For some models, extensive parameterization and validation may be required.

Freshwater
Effect of river regulation (8)

Terrestrial

Restoration (4) Colonisation (14)


Morphology and sedimentation (9) Effect of restoration (3) In-stream (6) Macroinvertebrates (18)

Disturbance

Flooding (6) Habitat Fragmentation (6)

Carbon cycling

Forests (3)

Riparian zone (1)

Fish (6) Platypus (1)


Macroinvertebrates (4)

Habitat-fauna interactions Sampling methods and classification

Woodlands (13) Floodplain (8) Woodland fauna (3)

Fish (8) Stream condition (1)

Life cycle
Fish (7)

Figure 5 Themes of published ecological research undertaken in the Goulburn-Broken Catchment: (number of studies in each theme).

5.1.4. Obtain access to component models


In most instances, the component models discussed are freely available. However, in a few cases the publicly available model version is an executable, and access to the raw code would be of use in development, additionally some models are under continued development by small research groups, and in these instances further access may need to be negotiated.

Page 35

5.1.5. Implement grid workflows


In the first instance, wrapping workflows around a single model so that input data can be formatted and routed to the model should be tested. This would involve setting file locations for input data and model outputs and producing variable strings to be tested by the model. A workflow actor would implement the model script or program, and other actors would handle file fetching, data formatting, writing outputs to file, setting file locations and model variables. Following this, a second model could be coupled to the first by wrapping the model in the same way as described above, but adding workflow actors that can re-format data to be compatible with the first model. A recommended order for model coupling would be 1. climate/weather with hydrology, 2. adding vegetation, 3. then biogeochemistry, 4. aquatic ecology, 5. terrestrial ecology (parallel development of the model for this module is likely to occur during the previous stages). Consideration of rescaling input/output data will be required for each coupling. Additionally, consideration of runtimes should occur in this phase, with adjustment of model resolutions or investigation of additional processing power to ensure practical application in subsequent phases.

5.1.6. Calibration and parameter optimization


As each model is implemented for the case study catchment, empirical and catchment-specific parameters will need to be calibrated. As each additional module is coupled, an additional check will be required to see if cross-cutting calibrated parameters maintain expected behaviour. Note that as discussed in section 3.1.1, there are specific parameter optimization capabilities in the software integration systems proposed. An examination of which parameters may be tuned at the scale of the entire Murray-Darling Basin should be made (i.e. parameters which should not vary much between catchments/sub-catchments). This requires review of current measured data, which may not be available across the basin, although nearby studies should also be of use in identifying variability. There will still be some parameters which will be catchment/sub-catchment specific, however it may be possible to produce a set of best-guess or default parameter values for use at the basin scale, or in catchments where no specific data has been collected.

5.1.7. Validation, analysis and scenarios


As discussed in section 4.2, traditional validation approaches may not be possible for the fully integrated model. However, where long term monitoring of certain variables or parameters is feasible, it may be possible to verify and validate certain process representations within the model, particularly if the model is used to guide future monitoring efforts. This approach is widely recommended for addressing knowledge gaps and improving our understanding and modelling of the system (Shifley et al. 2009). Multiple instantiations of the integrated model will provide a probability distribution for variables of interest and estimates of uncertainty in the model outputs. Analysis of the ecological processes significant in different outcomes can then occur, allowing identification of management practices favourable to desired outcomes. Examination of specific scenarios within the distributions can also occur.

5.1.8. Iterative model development


There are several areas that existing models address poorly, such as process-based export of dissolved catchment nutrients to waterways and changes or movement of vegetation functional types. Development in these areas would add significantly to this research effort into a whole-of-catchment ecological model. Additionally, depending on the final model selection, a number of potential gaps in ecological representation may occur, given choice of model in each module will not solely be based on processes represented, but will also take into account model quality (e.g. how appropriately the included processes are represented) as well as practical implementation issues, access, and available expertise.

Page 36

It is also important to keep in mind that any model, no matter how complex, is a simplification of reality, and as model development and application proceeds and we learn more about the system which we are modelling, improvements can be made. In using the model to guide management decisions and monitoring of outcomes, the processes modelled can be better represented and the model improved for future management decisions. This iterative approach to model development can feed directly and usefully into the on-ground adaptive management cycle.

Page 37

6. Conclusions
Two broad approaches to modelling a whole ecosystem can be identified; 1) build a single, highly parameterized model from the ground-up and attempt to represent as many ecosystem processes as possible in one framework; or 2) assemble a group of existing component models, validated and known to perform well for their defined modelling task (with a mind to their limitations), linking inputs and outputs across disciplinary boundaries. Where necessary, purpose-built models for components not well developed yet can be linked into the second approach. The latter has significant advantages, in that it is built on proven modelling techniques and is a way of integrating models from different disciplines without the need to completely re-write a modelling framework. This represents an enormous saving of human and financial resources that would be required to build and operate a single does -it-all model, at the expense of intrinsically linked ground-up model integration. We found that modelling capacity differs substantially among the six components: climate and weather, hydrology, biogeochemistry, vegetation, terrestrial biodiversity and aquatic biodiversity. We have identified that climate, and in particular weather, data should be a driver of ecosystem processes and not a functional process component itself. High resolution, statistically downscaled climate data already exist for south-eastern Australia, produced by SEACI, thus eliminating the need to undertake additional downscaling. Furthermore, without the ability to tightly couple models to global climate models, such as occurs with some global carbon or vegetation models, climate data becomes a one-way driver of ecosystem processes. As the next iteration of global climate modelling (ACCESS) is developed, the next iteration of our integrated ecosystem model can incorporate appropriate feedback loops, in particular for carbon and water. Hydrology models are capable of modelling water balance and groundwater flows at landscape scales over long time periods, with the potential to incorporate physical processes, such as leaf area index. The major limitation of hydrology models in this context is the availability of high-quality data to appropriately drive and validate models. The data might be appropriate for use in water resource planning, but may not be of sufficient resolution, temporally or spatially, for modelling ecological processes within catchments of intermediate sizes, which are nevertheless ecologically significant. However, a certain amount of essential on-ground sampling would be undertaken in the modelling phase of this project to ensure that the models are validated. Biogeochemistry models differ with regard to which cycles are regarded as important and are thus represented. Some of the more complex models require significant parameterization and calibration. Simpler empirical models of export are available but are probably not appropriate. Vegetation models range in scale from individual trees to global vegetation dynamics coupled to global climate models. Landscape-scale models that can represent spatial processes, such as fire, would be most suitable for a whole ecosystem model. The transfer and feedback of water and nutrients between the vegetation, biogeochemistry and hydrology modules is a key component of integration. Models of both terrestrial and aquatic (animal) biodiversity deal with modelling of species distributions based on bioclimatic envelopes, topography, soil types and habitat characteristics. A more systematic approach to this is needed. An alternative approach is one that focuses on building spatially explicit demographic models. These typically aim to identify whether a particular species, usually represented as a metapopulation, is likely to persist given the spatial pattern of habitats of various value to the species, birth and death rates, and rates of movement across the landscape. The availability of sophisticated model-linking software and the increasing computational power offered by distributed systems makes it feasible to couple multiple models from the different modules. Advances in eResearch provide opportunities for collaborating with other ecological research groups internationally. One example is the Open Wildland Fire Modelling e-Community (www.openwfm.org), which provides a portal for sharing modelling software, data and expertise. The approach to constructing a whole ecosystem model outlined in this report is feasible and can be implemented immediately using a suite of some of the existing models described. The conceptual framework for linking models presented in this report, while not a blueprint, is a valuable device for formulating research questions that can be used to query a whole of system model. We envisage that in implementing the next Page 38

stage of this approach, by actually constructing a whole ecosystem model, the conceptual framework for linking models will grow in size and detail in parallel with our understanding of ecological processes and our ability to represent them with models. An investment in this modelling approach is critical for initiating the next phase of natural resources modelling in Australia. A linked, whole ecosystem model of this type will be highly compatible with other modelling initiatives and will enhance, rather than compete with, other modelling systems being developed in Australia and internationally. This approach is arguably the most promising method for modelling whole ecosystems at catchment scales and, eventually, on the scale of the whole Murray-Darling Basin.

Page 39

7. Glossary
3-PG ACCESS AUSRIVAS AWBM CABALA CABLE CAEDYM CASA CatchMODS CMSS DNDC ELCOM EMSS FVS Grid ICMS IPCC JABOWA LASCAM LPJ Model ORCHIDEE PAR PERFECT R RothC SEACI SWAT WaterCAST Workflow Physiological Principles Predicting Growth Australian Community Climate and Earth-System Simulator Australian River Assessment Scheme Australian Water Balance Model CArbon BALAnce CSIRO Atmosphere Biosphere Land Exchange Computational Aquatic Ecosystem Dynamics Model Carnegie Ames Stanford Approach Biosphere model Catchment Scale Management Of Diffuse Sources Model Catchment Management Support System DeNitrification-DeComposition Model Estuarine and Lake COmputer Model Environmental Management Support System Forest Vegetation Simulator A computational grid, consisting of a distributed network of computers Integrated Catchment Management System Intergovernmental Panel on Climate Change Janak-Botkin-Wallis Large Scale Catchment Model Lund-Potsdam-Jena A computational model, consisting of a computer program that simulates a natural system Organizing Carbon and Hydrology in Dynamical Ecosystems Photosynthetically active radiation Productivity, Erosion and Runoff Functions to Evaluate Conservation Techniques A programming language and software environment for statistical computing and graphics Rothamsted Carbon Model South Eastern Australian Climate Initiative Soil & Water Assessment Tool Water Contaminant Analysis and Simulation Tool A sequence of operations automated by a software application

Page 40

8. References
Abramowitz G., Leuning R., Clark M. & Pitman A. (2008) Evaluating the Performance of Land Surface Models. J. Clim. 21, 5468-81. Abramson D., Giddy J. & Kotler L. (2000) High Performance Parametric Modeling with Nimrod/G: Killer Application for the Global Grid? In: International Parallel and Distributed Processing Symposium (IPDPS) pp. 520-8, Cancun, Mexico. Abramson D., Lewis A., Peachey T. & Fletcher C. (2001) An Automatic Design Optimization Tool and its Application to Computational Fluid Dynamics. In: SuperComputing 2001, Denver. Abramson D., Peachey T. & Lewis A. (2006) Model Optimization and Parameter Estimation with Nimrod/O. In: The International Conference on Computational Science, University of Reading, UK. Argent R. M., Perraud J. M., Rahman J. M., Grayson R. B. & Podger G. M. (2009) A new approach to water quality modelling and environmental decision support systems. Environmental Modelling & Software 24, 80918. Battaglia M., Sands P., White D. & Mummery D. (2004) CABALA: a linked carbon, water and nitrogen model of forest growth for silvicultural decision support. Forest Ecology and Management 193, 251-82. Beringer J., Hutley L. B., Tapper N. J., Coutts A., Kerley A. & O'Grady A. P. (2002) Fire impacts on surface heat, moisture and carbon fluxes from a tropical savanna in northern Australia. pp. 333-40. C S I R O Publishing, Darwin, Australia. Botkin D. B., Janak J. F. & Wallis J. R. (1972) Some Ecological Consequences of a Computer Model of Forest Growth. Journal of Ecology 60, 849-72. Boughton W. (2004) The Australian water balance model. Environmental Modelling & Software 19, 943-56. Boughton W. (2005) Catchment water balance modelling in Australia 1960-2004. Agric. Water Manage. 71, 91-116. Boughton W. C. (1988) Partitioning streamflow by computer. Transactions of the Institution of Engineers, Australia. Civil engineering CE30, 285-91. Brandmeyer J. E. & Karimi H. A. (2000) Coupling methodologies for environmental models. Environmental Modelling and Software 15, 479-88. Bugmann H. (2001) A Review of Forest Gap Models. Climatic Change 51, 259-305. Carpenter S. R. (2002) Ecological futures: Building an ecology of the long now. Ecology 83, 2069-83. Chessman B. C. (1986) Impact of the 1983 wildfires on river water quality in East Gippsland, Victoria. Australian Journal of Marine and Freshwater Research 37, 399-420. Chessman B. C., Muschal M. & Royal M. J. (2008) Comparing apples with apples: Use of limiting environmental differences to match reference and stressor-exposure sites for bioassessment of streams. River Res. Appl. 24, 103-17. Chessman B. C. & Royal M. J. (2004) Bioassessment without reference sites: use of environmental filters to predict natural assemblages of river macroinvertebrates. Journal of the North American Benthological Society 23, 599-615. Chiew F. H., Peel M. C. & Western A. W. (2002) Application and testing of the simple rainfall-runoff model SIMHYD. In: Mathematical Models of Small Watershed Hydrology (ed V. P. Singh, Frevert, S.K.) pp. 33567. Water Resources Publications, Littleton, Colorado. Ciais P., Friedlingstein P., Friend A. & Schimel D. S. (2001) Integrating Global Models of Terrestrial Primary Productivity. pp. 449-79. Academic Press. Page 41

Coleman K. & Jenkinson D. S. (1999) A model for the turnover of carbon in soil. Model description and windows users guide. Rothamsted Research, Harpenden, Herts. Croke B. F. W. & Jakeman A. J. (2001) Predictions in catchment hydrology: an Australian perspective. Mar. Freshw. Res. 52, 65-79. Crookston N. L. & Dixon G. E. (2005) The forest vegetation simulator: A review of its structure, content, and applications. Comput. Electron. Agric. 49, 60-80. CSIRO. (2007) Climate change in Australia. p. 152. Commonwealth Scientific and Industrial Research Organisation, Australia. Davis J. R., Nanninga P. M., Biggins J. & Laut P. (1991) Prototype Decision Support System for Analyzing Impact of Catchment Policies. Journal of Water Resources Planning and Management 117, 399-414. Drewry J. J., Newham L. T. H., Greene R. S. B., Jakeman A. J. & Croke B. F. W. (2006) A review of nitrogen and phosphorus export to waterways: context for catchment modelling. Marine and Freshwater Research 57, 757-74. Drielsma M. & Ferrier S. (2009) Rapid evaluation of metapopulation persistence in highly variegated landscapes. Biol. Cons. 142, 529-40. Elith J., Graham C. H., Anderson R. P., Dudik M., Ferrier S., Guisan A., Hijmans R. J., Huettmann F., Leathwick J. R., Lehmann A., Li J., Lohmann L. G., Loiselle B. A., Manion G., Moritz C., Nakamura M., Nakazawa Y., Overton J. M., Peterson A. T., Phillips S. J., Richardson K., Scachetti-Pereira R., Schapire R. E., Soberon J., Williams S., Wisz M. S. & Zimmermann N. E. (2006) Novel methods improve prediction of species' distributions from occurrence data. Ecography 29, 129-51. Feikema P., Lane P. & Sherwin C. (2008) Hydrological studies into the impact of timber harvesting on water yield in state forests supplying water to Melbourne - Part 2 of Hydrological studies (Climate Change and Bushfire). School of Forest and Ecosystem Science, The University of Melbourne for the Victorian Government, Melbourne. Feikema P. M., Sheridan G. J., Argent R. M., Lane P. N. J. & Grayson R. B. (2005) Using E2 To Model The Impacts Of Bushfires On Water Quality In South-Eastern Australia. In: MODSIM 05 International Congress on Modelling and Simulation (eds A. Zerger and R. M. Argent) pp. 1126-32. Modelling and Simulation Society of Australia and New Zealand, Melbourne, Australia. Ferrier S., Drielsma M., Manion G. & Watson G. (2002) Extended statistical approaches to modelling spatial pattern in biodiversity in northeast New South Wales. II. Community-level modelling. Biod. & Cons. 11, 230938. Fleishman E. & Mac Nally R. (2007) Contemporary drivers of fragmentation and measurement of their effects on animal diversity. Can. J. Zool. 85, 1080-90. Gentile J. H., Harwell M. A., Cropper W., Harwell C. C., DeAngelis D., Davis S., Ogden J. C. & Lirman D. (2001) Ecological conceptual models: a framework and case study on ecosystem management for South Florida sustainability. Sci. Total Environ. 274, 231-53. Gerten D., Luo Y., Le Maire G., Parton W. J., Keough C., Weng E., Beier C., Ciais P., Cramer W., Dukes J. S., Hanson P. J., Knapp A. A. K., Linder S., Nepstad D., Rustad L. & Sowerby A. (2008) Modelled effects of precipitation on ecosystem carbon and water dynamics in different climatic zones. Global Change Biology 14, 2365-79. He H. S., Keane R. E. & Iverson L. R. (2008) Forest landscape models, a tool for understanding the effect of the large-scale and long-term landscape processes. Forest Ecology and Management 254, 371-4. He H. S. & Mladenoff D. J. (1999) Spatially explicit and stochastic simulation of forest-landscape fire disturbance and succession. Ecology 80, 81-99. Howden M., Carberry P. & Hayman P. (2008) SEACI technical report for Milestone 3.1.6. SEACI. Page 42

Howe C., Jones R. N., Maheepala S. & Rhodes B. (2005) Melbourne Water Climate Change Study: Implications of potential climate change for Melbourne's water resources. p. 26. CSIRO, Melbourne. Jakeman A. J., Letcher R. A. & Norton J. P. (2006) Ten iterative steps in development and evaluation of environmental models. Environmental Modelling & Software 21, 602-14. Jones R. N., Chiew F. H. S., Boughton W. C. & Zhang L. (2006) Estimating the sensitivity of mean annual runoff to climate change using selected hydrological models. Advances in Water Resources 29, 1419-29. Kandel D., Chiew F. & Grayson R. (2005) A Tool for Mapping and Forecasting Soil Moisture Deficit over Australia. Cooperative Research Centre for Catchment Hydrology. Kowalczyk E. A., Wang Y. P., Law R. M., Davies H. L., McGregor J. L. & Abramowitz G. (2006) The CSIRO Atmosphere Biosphere Land Exchange (CABLE) model for use in climate models and as an offline model. CSIRO Marine and Atmospheric Research, Aspendale, Victoria. Krinner G., Viovy N., de Noblet-Ducoudr N., Oge J. r. m., Polcher J., Friedlingstein P., Ciais P., Sitch S. & Prentice I. C. (2005) A dynamic global vegetation model for studies of the coupled atmospherebiosphere system. Global Biogeochem. Cycles 19. Landsberg J. J. & Waring R. H. (1997) A generalised model of forest productivity using simplified concepts of radiation-use efficiency, carbon balance and partitioning. Forest Ecology and Management 95, 209-28. Landsberg J. J., Waring R. H. & Coops N. C. (2003) Performance of the forest productivity model 3-PG applied to a wide range of forest types. Forest Ecology and Management 172, 199-214. Lane P., Feikema P. M., Sherwin C., Peel M. C. & Freebairn D. M. (2007) Physically-based prediction of water yield from disturbed forested water supply catchments. In: MODSIM 2007 International Congress on Modelling and Simulation. Modelling and Simulation Society of Australia and New Zealand (eds L. Oxley and D. Kulasiri) pp. 2953-9. MODSIM, Christchurch, New Zealand. Letcher R. A., Jakeman A. J., Calfas M., Linforth S., Baginska B. & Lawrence I. (2002) A comparison of catchment water quality models and direct estimation techniques. Environmental Modelling & Software 17, 77-85. Li C., Frolking S. & Frolking T. A. (1992a) A Model of Nitrous Oxide Evolution From Soil Driven by Rainfall Events: 1. Model Structure and Sensitivity. J. Geophys. Res. 97. Li C., Frolking S. & Frolking T. A. (1992b) A Model of Nitrous Oxide Evolution From Soil Driven by Rainfall Events: 2. Model Applications. J. Geophys. Res. 97. Littleboy M., Silburn D. M., Freebairn D. M., Woodruff D. R., Hammer G. L. & Leslie J. K. (1992) Impact of Soil-Erosion on Production in Cropping Systems .1. Development and Validation of a Simulation-Model. Australian Journal of Soil Research 30, 757-74. Lo J. C.-F., Yang Z.-L. & Pielke R. A., Sr. (2008) Assessment of three dynamical climate downscaling methods using the Weather Research and Forecasting (WRF) model. J. Geophys. Res. 113. Lundberg P., Ranta E., Ripa J. & Kaitala V. (2000) Population variability in space and time. Trends in Ecology & Evolution 15, 460-4. Lynch A. H., Abramson D., Gorgen K., Beringer J. & Uotila P. (2007) Influence of savanna fire on Australian monsoon season precipitation and circulation as simulated using a distributed computing environment. Geophys. Res. Lett. 34, 5. Mac Nally R., Fleishman E., Thomson J. & Dobkin D. S. (2008) Use of guilds for modeling avian responses to vegetation change in the Intermountain West (U.S.A.). Global Ecology and Biodiversity 17, 758-69. Marston F., Argent R., Vertessy R., Cuddy S. & Rahman J. (2002) The Status of Catchment Modelling in Australia. Technical Report 02/4. Cooperative Research Centre for Catchment Hydrology, Canberra.

Page 43

Marston F. M., Young W. & Davis R. (1995) Nutrient Data Book: Nutrient generation rates data book. CSIRO Division of Water Resources, LWRRDC, SWB, Canberra. Matthews S. (2006) A process-based model of fine fuel moisture. Int. J. Wildland Fire 15, 155-68. May R. M. (1976) Simple mathematical-models with very complicated dynamics. Nature 261, 459-67. McCarthy M. A., Malcolm Gill A. & Lindenmayer D. B. (1999) Fire regimes in mountain ash forest: evidence from forest age structure, extinction models and wildlife habitat. Forest Ecology and Management 124, 193203. McDonald M. G. & Harbaugh A. W. (1988) A modular three-dimensional finite-difference ground-water flow model. In: Techniques of Water-Resource Investigation p. 586. McKenzie D., Peterson D. L. & Alvarado E. (1996) Extrapolation problems in modeling fire effects at large spatial scales: A review. Int. J. Wildland Fire 6, 165-76. Miehle P., Battaglia M., Sands P. J., Forrester D. I., Feikema P. M., Livesley S. J., Morris J. D. & Arndt S. K. (2009) A comparison of four process-based models and a statistical regression model to predict growth of Eucalyptus globulus plantations. Ecological Modelling 220, 734-46. Mladenoff D. J. (2004) LANDIS and forest landscape models. Ecological Modelling 180, 7-19. Moldan B., ern, J. (1994) Small catchment research. In: Biogeochemistry of small catchments, a tool for environmental research. (ed B. Moldan, ern, J.) pp. 1-30. John Wiley & Sons, Chichester, UK. Murphy B. F. & Timbal B. (2008) A review of recent climate variability and climate change in southeastern Australia. International Journal of Climatology 28, 859-79. Newham L. T. H., Letcher R. A., Jakeman A. J. & Kobayashi T. (2002) A framework for integrated hydrologic, sediment and nutrient export modelling for catchment-scale management. pp. 1029-38. Elsevier Sci Ltd, Lugano, SWITZERLAND. Newham L. T. H., Letcher R. A., Jakeman A. J. & Kobayashi T. (2004) A framework for integrated hydrologic, sediment and nutrient export modelling for catchment-scale management. Environmental Modelling & Software 19, 1029-38. Nightingale J. M., Hill M. J., Phinn S. R., Davies I. D., Held A. A. & Erskine P. D. (2008) Use of 3-PG and 3PGS to simulate forest growth dynamics of Australian tropical rainforests: I. Parameterisation and calibration for old-growth, regenerating and plantation forests. Forest Ecology and Management 254, 107-21. Ogden J. C., Davis S. M., Barnes T. K., Jacobs K. J. & Gentile J. H. (2005) Total System Conceptual Ecological Model. Wetlands 25, 955-79. Oreskes N., Shraderfrechette K. & Belitz K. (1994) Verification, Validation, and Confirmation of NumericalModels in the Earth-Sciences. Science 263, 641-6. Pacala S. W., Canham C. D., Saponara J., Silander Jr. J. A., Kobe R. K. & Ribbens E. (1996) Forest Models Defined by Field Measurements: Estimation, Error Analysis and Dynamics. Ecological Monographs 66, 1-43. Pacala S. W., Canham C. D. & Silander J. A. J. (1993) Forest models defined by field measurements: I. The design of a northeastern forest simulator. Can. J. For. Res. 23, 1980-8. Parker P., Letcher R., Jakeman A., Beck M. B., Harris G., Argent R. M., Hare M., Pahl-Wostl C., Voinov A., Janssen M., Sullivan P., Scoccimarro M., Friend A., Sonnenshein M., Baker D., Matejicek L., Odulaja D., Deadman P., Lim K., Larocque G., Tarikhi P., Fletcher C., Put A., Maxwell T., Charles A., Breeze H., Nakatani N., Mudgal S., Naito W., Osidele O., Eriksson I., Kautsky U., Kautsky E., Naeslund B., Kumblad L., Park R., Maltagliati S., Girardin P., Rizzoli A., Mauriello D., Hoch R., Pelletier D., Reilly J., Olafsdottir R. & Bin S. (2002) Progress in integrated assessment and modelling. Environmental Modelling & Software 17, 209-17.

Page 44

Parton W. J., Stewart J. W. B. & Cole C. V. (1988) Dynamics of C, N, P and S in Grassland Soils: A Model. Biogeochemistry 5, 109-31. Peel M. C., Watson F. G. R. & Vertessy R. A. (2003) Modelling impacts of land use change on low flows in the North Esk river using the macaque model. pp. 849-54. Univ Western Australia, Townsville, AUSTRALIA. Perry G. L. W. & Enright N. J. (2006) Spatial modelling of vegetation change in dynamic landscapes: a review of methods and applications. Progress in Physical Geography 30, 47-72. Port A. & Bartelink H. H. (2002) Modelling mixed forest growth: a review of models for forest management. Ecological Modelling 150, 141-88. Potter C. S., Alexander S. E., Coughlan J. C. & Klooster S. A. (2001) Modeling biogenic emissions of isoprene: exploration of model drivers, climate control algorithms, and use of global satellite observations. Atmospheric Environment 35, 6151-65. Potter C. S. & Klooster S. A. (1999) Dynamic global vegetation modelling for prediction of plant functional types and biogenic trace gas fluxes. Global Ecology and Biogeography 8, 473-88. Potter C. S., Randerson J. T., Field C. B., Matson P. A., Vitousek P. M., Mooney H. A. & Klooster S. A. (1993) Terrestrial Ecosystem Production - a Process Model-Based on Global Satellite and Surface Data. Global Biogeochemical Cycles 7, 811-41. Prosser I., Rustomji P., Young B., Moran C. & Hughes A. (2001a) Constructing River Basin Sediment Budgets for the National Land and Water Resources Audit. CSIRO Land and Water, Canberra. Prosser I. P., Rutherfurd I. D., Olley J. M., Young W. J., Wallbrink P. J. & Moran C. J. (2001b) Large-scale patterns of erosion and sediment transport in river networks, with examples from Australia. Marine and Freshwater Research 52, 81-99. Ranatunga K., Nation E. R. & Barratt D. G. (2008) Review of soil water models and their applications in Australia. Environmental Modelling & Software 23, 1182-206. Ravalico J. K., Dandy G. C. & Maier H. R. (2007) More sensitivity analysis of the MSM-BIGMOD River Murray flow and salinity model. In: MODSIM 2007 International Congress on Modelling and Simulation (ed L. O. a. D. Kulasiri) pp. 2754-60. Modelling and Simulation Society of Australia and New Zealand, Christchurch, New Zealand. Reckhow K. H. & Chapra S. C. (1999) Modeling excessive nutrient loading in the environment. Environmental Pollution 100, 197-207. Recknagel F., Hosomi M., Fukushima T. & Kong D.-S. (1995) Short- and long-term control of external and internal phosphorus loads in lakes--A scenario analysis. Water Research 29, 1767-79. Reynolds C. S., Irish A. E. & Elliott J. A. (2001) The ecological basis for simulating phytoplankton responses to environmental change (PROTECH). Ecological Modelling 140, 271-91. Reynoldson T. B., Norris R. H., Resh V. H., Day K. E. & Rosenberg D. M. (1997) The reference condition: a comparison of multimetric and multivariate approaches to assess water-quality impairment using benthic macroinvertebrates. Journal of the North American Benthological Society 16, 833-52. Rizzoli A. E., Davis J. R. & Abel D. J. (1998) Model and data integration and re-use in environmental decision support systems. Decis. Support Syst. 24, 127-44. Robson B. J. & Hamilton D. P. (2004) Three-dimensional modelling of a Microcystis bloom event in the Swan River estuary, Western Australia. Ecological Modelling 174, 203-22. Sherman B., Todd C. R., Koehn J. D. & Ryan T. (2007) Modelling the impact and potential mitigation of cold water pollution on Murray cod populations downstream of Hume Dam, Australia. River Res. Appl. 23, 37789. Page 45

Shifley S. R., Rittenhouse C. D. & Millspaugh J. J. (2009) Validation of Landscape-Scale Decision Support Models That Predict Vegetation and Wildlife Dynamics. In: Models for Planning Wildlife Conservation in Large Landscapes pp. 415-48. Academic Press, San Diego. Sitch S., Smith B., Prentice I. C., Arneth A., Bondeau A., Cramer W., Kaplan J. O., Levis S., Lucht W., Sykes M. T., Thonicke K. & Venevsky S. (2003) Evaluation of ecosystem dynamics, plant geography and terrestrial carbon cycling in the LPJ dynamic global vegetation model. Global Change Biology 9, 161-85. Sivapalan M., Ruprecht J. K. & Viney N. R. (1996a) Water and salt balance modelling to predict the effects of land-use changes in forested catchments. 1. Small catchment water balance model. Hydrol. Process. 10, 393-411. Sivapalan M., Viney N. R. & Jeevaraj C. G. (1996b) Water and salt balance modelling to predict the effects of land-use changes in forested catchments. 3. The large catchment model. Hydrol. Process. 10, 429-46. Sivapalan M., Viney N. R. & Ruprecht J. K. (1996c) Water and salt balance modelling to predict the effects of land-use changes in forested catchments. 2. coupled model of water and salt balances. Hydrol. Process. 10, 413-28. Taylor A. R., Chen H. Y. H. & VanDamme L. (2009) A Review of Forest Succession Models and Their Suitability for Forest Management Planning. For. Sci. 55, 23-36. Thomson J. R., Moilanen A., Vesk P., Bennett A. F. & Mac Nally R. (2009) Where and when to revegetate: a quantitative method for scheduling landscape reconstruction. Ecol. Appl. 19, 817-28. Timbal B. & Jones D. (2008) Future projections of winter rainfall in southeast Australia using a statistical downscaling technique. Climatic Change 86, 165-87. Todd C. R., Nicol S. J. & Koehn J. D. (2004) Density-dependence uncertainty in population models for the conservation management of trout cod, Maccullochella macquariensis. Ecological Modelling 171, 359-80. Todd C. R., Ryan T., Nicol S. J. & Bearlin A. R. (2005) The impact of cold water releases on the critical period of post-spawning survival and its implications for Murray cod (Maccullochella peelii peelii): A case study of the Mitta Mitta River, southeastern Australia. River Res. Appl. 21, 1035-52. Townsend C. R., Begon M. & Harper J. L. (2008) Essentials of Ecology. Wiley-Blackwell, Oxford. Vertessy R. A., Watson F. G. R., Rahman J. M., Cuddy S. M., Seaton S. P., Chiew F. H., Scanlon P. J., Marston F. M., Lymburner L., Jeannelle S. & Verbunt M. (2001) New software to aid water quality management in the catchments and waterways of the south-east Queensland region. In: Third Stream Management Conference: The Value of Healthy Streams (eds I. Rutherfurd, F. Sheldon, G. Brierley and C. Kenyon) pp. 611-6. Cooperative Research Centre for Catchment Hydrology, Brisbane. Wang P. & Linker L. C. (2006) A correction of DIN uptake simulation by Michaelis-Menton saturation kinetics in HSPF watershed model to improve DIN export simulation. Environmental Modelling & Software 21, 45-60. Watson F. G. R. (1999) Large scale, long term modelling of the effects of land cover change on forest water yield. In: Department of Civil and Environmental Engineering p. 443. University of Melbourne, Melbourne. Watson F. G. R., Vertessy R. A. & Grayson R. B. (1999) Large-scale modelling of forest hydrological processes and their long-term effect on water yield. Hydrological Processes 13, 689-700. Weeks A., Christy B., Lowell K. & Beverly C. (2008) The Catchment Analysis Tool: Demonstrating the Benefits of Interconnected Biophysical Models. In: Landscape Analysis and Visualisation pp. 49-71. Wlosinski J. H. & Collins C. D. (1985) Evaluation of a water quality model (CE-QUAL-R1) using data from a small Wisconsin reservoir. Ecological Modelling 29, 303-13. Young W. J., Scott A. C., Cuddy S. M. & Rennie B. A. (2003) Murray Flow Assessment Tool. A Technical Description (based on MFAT version 1.4). p. 81. CSIRO Land and Water, Canberra.

Page 46

Model Acronym

Availability

Outputs (variables/features/capability)

Potential Resolution

Quality

Rainfall

Temperature

Humidity

Cloud cover

Potential Evapotranspiration Wind

Spatial Resolution

Temporal Resolution

Previous rescaling

Peer Reviewed

Used by more than developer

ACCESS
1 1

CLIMATE

Currently in development ? ? ? 1x1km Daily

Table 2 Climate (and Weather) Module Models Comparison Matrix

Appendix 1 Detailed model comparison tables

IPCC 4 SEACI downscaling for weather 1x1km

WRF

Daily

These may not have been archived.

Previously coupled to other models

Page 47

LAND SURFACE (COMBINED) MODELS

Table 3 Mixed/Overlapping Modules Models Comparison Matrix

EWater Toolkit subcatchment daily

CABLE

Model Acronym

MODELS

Availability

Hydrology

Rainfall-runoffevapotranspiration Soil/sub-surface water Sediment Salinity Vegetation

Forest stands

Carbon

Biogeochemistry

Nitrogen

Phosphorus

20x20m

subcatchment

Resolution

Spatial resolution

daily

annual

Temporal resolution

Quality

Peer Reviewed

Used by more than developer

Manual acceptable

Page 48
CAT ICMS ? ? ~ ~ ~ Vegetation

Table 4 Hydrology Module Models General Comparison Matrix

Page 49
HYDROLOGY

Model Acronym

Macaque

PERFECT and MODFLOW (in CAT)

SIMHYD (in E2/WaterCAS T) subcatch ments (>100s) 1 daily > 100 years

Slopes (>1000) 2 daily > 250 years ?

Availability Rainfallrunoff Surface runoff Subsurface runoff Soil Water

Outputs (variables/features/capability)

20x20m 3+ daily > decad al ~ ~ ?

Groundwater

Sediment transport Salinity transport

Resolution

Min spatial resolution

Soil layers Min temporal resolution Length of simulations

Quality

Peer Reviewed Used by more than developer Manual acceptable

HYDROLOGY

Topopgraphy (DEM) Soil type Vegetation/crop types Rainfall Temperature Evapotranspiration Land use Crop rotation Management strategy Gauged streamflow Bore data

Soil evaporation Evapotranspiration

PROCESSES

Rainfall-runoff Saturation Interception

Constituent generation Transport Runoff Soil erosion/loss Evaporation Soil Water Groundwater Drainage Crop growth Salinity transport Nutrient transport

Page 50

Macaque

Model Acronym

Table 5 Hydrology Module Models Comparison Matrix

SIMHYD (in E2/WaterC AST)

PERFECT and MODFLO W (in CAT)

INPUTS OUTPUTS

Outputs (variables/features/capability)

Resolution

Quality

Availability

Model Acronym
Nitrogen Phosphorus Species (N2O?) Nutrient Processing Sediment Soil layers Spatial resolution Temporal resolution Length of simulation Peer Reviewed

Carbon

SWAT and DAISY (in CAT) 3+ 20x20 m Daily

CASA

Table 6 Biogeochemistry Module Models Comparison Matrix

SedNet

Subca tchme nts 1 Subca tchme nts <1x1k m

~Daily

BIOGEOCHEMISTRY

E2/CMSS (eWater Toolkit)

Daily

CENTURY/

Daily

4+, 1cm resolu tion

DAYCENT

Used by more than developer Manual

Page 51

BIOGEOCHEMISTRY

Rainfall Temperature

Soil characteristics (e.g. Initial concentrations clay/sand Atmospheric inputs Plant lignin content Fertilizer inputs Gauged nutrients

Fixation Plant uptake Denitrification Leaching Volatilisation Erosion

Phosphotase production Mineralization

Sediment

OUTPUTS

SOC Organic N Organic P

DIN PO4

TN TP CH4 CO2

N2O NOx

Page 52
CASA SedNet Model Acronym E2/CMSS (eWater Toolkit)

Table 7 Biogeochemistry Module Models Comparison Matrix

DAYCENT

SWAT and DAISY (CAT)

CENTURY/

INPUTS PROCESSES

Table 8 Vegetation Module Models Comparison Matrix

Page 53
VEGETATION

LPJ
> 100s

FVS

CAT crop, pasture, forest models (include s 3PG)

1 yr-daily > 100s

CENTUR Y/DAYCENT
000s of stands 5 years > 100s

Model Acronym


20 x 20m Crops daily, Forests monthly Decadal

Availability Stand Representation Individual Tree Representation

Age structure Species succession EVCs Riparian Vegetation Stand Health Water Carbon Spatial resolution Temporal resolution Length of simulations

Peer Reviewed Used by more than developer Australian Applications Previous coupling

Stand Health

Riparian Vegetation Evapotranspiration Stand Relative Age


~

Rainfall

Temperature

Solar radiation

Vapour Pressure Deficit

CO2

Slope/aspect/DEM

Elevation

Soil chars (e.g. texture, water capacity)

Initial stem number

Initial mass fractions: stem, foliage, roots Soil Fertility Ratio

Litterfall Rate

Maximum Stomatal Conductance

Canopy Quantum Efficiency

Net Primary Production

Biomass allocation

Water Use

Soil Water Balance

Nutrient uptake

Stem Mortality

Litterfall

Root turnover

Establishment/recruitment

Species specific traits

Biomass pools (stems, foliage, roots) Tree density

Biomass fixed (growth)

Water Use (Soil Water left)

Stem Diameter Distn (Basal Area)

3PG (in CAT)

FVS

Leaf Area Index

VEGETATION

LPJ

DAYCEN T

Table 9 Vegetation Module Models Detailed Inputs/Outputs Comparison Matrix

UNCERTAINTIES: Root depth, soil heterogeneity, fertility measures not necessarily explicitly nutrient related (Landsberg et al. 2003),

PROBLEMS: Scale issues/multiple disturbances, validation/lack of data (long term slow phenomena), mixes of tree species not additive, effects on soil chem/nutrients (Landsberg et al. 2003), poor nutrient cycling (Miehle et al. 2009), poor establishment/recruitment modelling (Crookston and Dixon 2005; Port and Bartelink 2002). Most tree models growth/yield focused better growth predictions, but not very good for succession/dynamics. Gap models (patches of forest with list of species) appear best for succession modelling, especially where heterogeneous/mixed species forests are of interest (Port and Bartelink 2002). Large gap in Stand Health. Still a very active area of research and development no consensus or convergence of models (He et al. 2008).

EVC/Species Succession

Page 54
PROCESSES OUTPUTS

INPUTS

Model Acronym

Table 10. Terrestrial Ecology Module Models Comparison Matrix

Page 55
TERRESTRIAL ECOLOGY

Model

Species specific models (e.g. Population models)

Demographic spatially dynamic models Variable Variable

Availability

? Variable Variable

Processbased

Statistical

Indicator species

Biodiversity

Spatial resolution

Temporal resolution

Peer Reviewed

Used by more than developer

AQUATIC ECOLOGY

Process-based

Statistical

Reach

Reach

arbitrary < 1x1m

Spatial resolution (less than catchment)

arbitrary < 1hr

Temporal resolution (less than daily)

Sediment

Carbon

Nitrogen

Phosphorus

Algae

Higher ecology

Biodiversity

Peer-reviewed

Used by more than developer

Page56

Model Acronym

AUSRIVAS

Ecological Modeller

CAEDYM

Table 11 Aquatic Ecology Module Models Comparison Matrix

Models similar to CAEDYM CEQUAL, PROTECH, AQUAMOD ~

Das könnte Ihnen auch gefallen