Sie sind auf Seite 1von 12

Journal of Hydrology 404 (2011) 112

Contents lists available at ScienceDirect

Journal of Hydrology
journal homepage: www.elsevier.com/locate/jhydrol

Spatiotemporal dynamics of landscape pattern and hydrologic process in watershed systems


Timothy O. Randhir , Olga Tsvetkova
Department of Environmental Conservation, University of Massachusetts, Holdsworth Natural Resources Center, Rm. 320, Amherst, MA 01003, United States

a r t i c l e

i n f o

s u m m a r y
Land use change is inuenced by spatial and temporal factors that interact with watershed resources. Modeling these changes is critical to evaluate emerging land use patterns and to predict variation in water quantity and quality. The objective of this study is to model the nature and emergence of spatial patterns in land use and water resource impacts using a spatially explicit and dynamic landscape simulation. Temporal changes are predicted using a probabilistic Markovian process and spatial interaction through cellular automation. The MCMC (Monte Carlo Markov Chain) analysis with cellular automation is linked to hydrologic equations to simulate landscape patterns and processes. The spatiotemporal watershed dynamics (SWD) model is applied to a subwatershed in the Blackstone River watershed of Massachusetts to predict potential land use changes and expected runoff and sediment loading. Changes in watershed land use and water resources are evaluated over 100 years at a yearly time step. Results show high potential for rapid urbanization that could result in lowering of groundwater recharge and increased storm water peaks. The watershed faces potential decreases in agricultural and forest area that affect open space and pervious cover of the watershed system. Water quality deteriorated due to increased runoff which can also impact stream morphology. While overland erosion decreased, instream erosion increased from increased runoff from urban areas. Use of urban best management practices (BMPs) in sensitive locations, preventive strategies, and long-term conservation planning will be useful in sustaining the watershed system. 2011 Elsevier B.V. All rights reserved.

Article history: Received 29 November 2010 Received in revised form 7 March 2011 Accepted 15 March 2011 Available online 21 March 2011 This manuscript was handled by G. Syme, Editor-in-Chief Keywords: Nonpoint source pollution Spatio-temporal modeling Watershed management Runoff GIS Dynamic modeling

1. Introduction Watershed systems provide multiple goods and services that sustain human population and ecosystems (Randhir and Shriver, 2009). Rapid resource depletion and increasing demands from human populations impact the structure and function of watersheds, thereby reducing the ability to sustain these services. Management and policies to protect these resources require information on the dynamics of the system, particularly an evaluation of the spatial and temporal interactions among watershed components. An assessment of the changes in the state of the system (pattern) and dynamics of ows (processes) is critical to evaluate and manage changes in landscape characteristics and environmental processes. Such integrated assessment can be used to address long term issues surrounding depletion of water resources and the cumulative impairment of the watersheds capability to sustain ecosystem services. Land use change and environmental outcomes are a result of the combined inuence of biophysical and socioeconomic drivers. Processes in watershed systems in particular, are essentially
Corresponding author. Tel.: +1 413 545 3969.
E-mail addresses: Randhir@eco.umass.edu, trandhir@gmail.com (T.O. Randhir). 0022-1694/$ - see front matter 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.jhydrol.2011.03.019

dependent on the spatial distribution of components, interactions, and temporal changes. For example, water purication and wildlife habitat protection are watershed ecosystem services that are dependent on land use pattern. Spatial patterns of land use, landcover change, and resource management change the nature and spatial distribution of pollutant loading (Randhir et al., 2000). A dynamic assessment of the landscape is useful in the development of spatial (targeting) and temporal (contracts) policies to protect water resources. The dynamic information will also allow insights into system trajectory, and adaptive management. Land use and land-cover (LULC) changes are inuenced by natural processes and are both direct and indirect effects of human activities (Turner and Meyer, 1991). Land use change inuences water-quality and habitat degradation (Randhir and Hawes, 2009; USGS, 2005), and loss of soil quality (Randhir, 2003). It also inuences water runoff, sedimentation rates (Marshall and Randhir, 2008), earth-atmosphere interactions, biodiversity, water budget, and biogeochemical cycling of carbon, nitrogen and other elements at regional to global scales (Vitousek, 1994). Evaluating the impacts of LULC change is important in land management decisions and in protecting natural resources of watershed systems (Marshall and Randhir, 2008). Documenting the rates of change, driving forces, and impacts of LULC on watershed systems is a

T.O. Randhir, O. Tsvetkova / Journal of Hydrology 404 (2011) 112

focus of government agencies such as the USEPA (Jones et al., 1999) to bring integrated solutions to resource management. Changes in LULC can be indicative of regional environmental problems that result from impairment of both abiotic (surface runoff dynamics, lowering of groundwater tables, impacts on rates and types of land degradation), and biotic (habitat loss, reduction in biodiversity, and species extinction) processes (Shriver and Randhir, 2006). An understanding of changes in land use and water use over the next 3050 years is central to attaining environmental sustainability (IIASA, 1999). The watershed landscape is dynamic in spatial, structural and functional patterns (Hobbs, 1997) that are essential characteristics of landscape ecology (McGarigal and Marks, 1995). Spatially explicit land use change models are needed. A system dynamics paradigm (Forrester, 1971) that is based on conceptualizing a system in terms of compartments (stocks) and ows provides an intuitive way of modeling differential or difference equations associated with system processes. The application of this framework to model landscape pattern and processes has an excellent scope toward integrated approaches to decision making (Randhir and Hawes, 2009). Modeling the dynamic emergence of landscape patterns and their dynamic implications are important to watershed research (Marshall and Randhir, 2008) and are important for adaptive management of watershed resources. The dynamic assessment (Randhir and Hawes, 2009) that links land cover to hydrologic dynamics is useful in predicting long term impacts of land use change on processes. A watershed manager can assess the system dynamics and gain insights into long term dynamics in complexity, structure, and functions of watershed systems. In addition, the development of site specic policies and implementation of management practices essentially depends on this spatial information. 1.1. Background Spatial decision support systems (Lovejoy et al., 1997; Djodjic et al., 2002) are becoming important to local and regional environmental impact assessment, planning, and implementation of regulatory policies (Munier et al., 2002) and in decision-making (Randhir and Shriver, 2009). Models of human decision-making with spatial and temporal dimensions cover a range of methods with different simulation techniques: statistical/econometric models, dynamic systems models, logistic function models, regression models, spatial simulation models, linear planning models, nonlinear mathematical planning models, mechanistic GIS models, and cellular automata models (Agarwal et al., 2000). Interest in the application of agent-based models to land-cover change (Marshall and Randhir, 2008) has also been growing rapidly during recent years. Agent-based models (Laine and Busemeyer, 2003) combine a cellular model representing the landscape with an agent-based model to represent decision-making entities (Marshall and Randhir, 2008). The agent-based model may include a variety of spatial processes and inuences relevant for land-cover change. Cellular automation models are important tools for the prediction of landscape changes such as the spread of urbanization and future land cover patterns at different spatial scales (Marshall and Randhir, 2008; Heiken et al., 2000). A Markovian cellular automation process can be used to predict land-cover changes (Marshall and Randhir, 2008). Differential equations in a visual environment are used to model system dynamics and implemented using Stella software (Randhir and Hawes, 2009; Costanza and Voinov, 2001; Costanza et al., 1998). Systems modeling can be linked to Geographic Information Systems (GIS) to model landscape changes (Maxwell and Costanza, 1994; Wilkie and Finn, 1988)). A dynamic landscape simulation of socio-economic effects on landscape change can be

applied to the built environment (Wang and Zhang, 2001) and to the peri-urban system (Anwar and Borne, 2005). Land use change can be analyzed using multivariate logistic regression (Allen et al., 1999) and remote sensing methods (Jane, 2003). A multiple criteria, dynamic spatial optimization can also be used to model water quality on a watershed scale (Randhir et al., 2000). A dynamic system approach can be used for managing and understanding complex ecological-economic systems (Costanza and Ruth, 1998). A multiattribute optimization can be used to model restoration options in watersheds (Randhir and Shriver, 2009). While several approaches have been used to study land use impacts on watersheds, there is a further need to develop a spatiotemporal, systems framework for modeling land use change and hydrological impacts for policy. Studies in spatio-temporal modeling of land use dynamics coupled with assessment of hydrologic process are limited in watershed literature (Marshall and Randhir, 2008). This study will address a portion of this gap with development of a system analysis of land use change and watershed response at a spatially and temporally explicit scale. The ability to use land use and hydrological assessment together to evaluate dynamic interactions and long term trends can support management decisions toward sustainability. This paper is unique in the development of such long term spatial simulation that combines Markovian process of land use change, cellular automation, nonlinear simulation, and spatially explicit assessment to model watershed processes. Specically, the watershed is rasterized (grid structure to represent spatial information) and subject to LULC change that is triggered by a Markov chain probabilistic transition matrix and spatial interactions. Each raster is modeled as a cellular agent with dynamic processes observed and modeled at a temporal scale of several decades to identify long term watershed impacts. The paper develops a unique approach to model spatial and temporal changes in watershed land use and hydrologic changes to evaluate opportunities to mitigate land use impacts on water resources. While land use modeling has been developed independently and later used in water resource and NPS quality assessment, there are no coupled assessments that are linked within a dynamic model framework. The spatiotemporal watershed dynamics (SWD) model is applied to a typical watershed to evaluate spatial distribution, temporal trends, and policy options. Such integrated assessment is critical as we move toward improving the sustainability of watershed systems. Land use change is inuenced by spatial and temporal factors that interact with watershed resources. Modeling these changes is critical to evaluate emerging land use patterns and to predict variation in water quantity and quality. This study models the nature and emergence of spatial patterns in land use and water resource impacts using a spatially explicit and dynamic landscape simulation. Use of urban BMPs in sensitive locations, preventive strategies, and long-term conservation planning will be useful in sustaining water and watershed systems. The general objective of this study is to assess the nature and emergence of spatial patterns in land use and the response of water resources in a small watershed system at long term temporal scales. Specically: (i) develop a spatially explicit, dynamic simulation model that links land use change with response of a watershed ecosystem; (ii) study the dynamic inuence of land use on runoff and sediment loading local and watershed scales; and (iii) to identify runoff and sediment reduction policies through spatial and temporal targeting. Hypotheses that will be tested include: (i) Spatial and temporal changes in land use have signicant effects on runoff and sedimentation in a watershed ecosystem (ii) the nature of spatial patterns in water quality in a watershed are inuenced by temporal transitions and spatial inuences; and (iii) dynamic modeling can result in more accurate assessment of

T.O. Randhir, O. Tsvetkova / Journal of Hydrology 404 (2011) 112

policy effectiveness. The hypotheses will be tested using comparison of dynamic and static (baseline) in trends. 2. Methodology We develop a decision model for watershed management and policy. The watershed is specied as gridded spatial units at location i and j within a matrix that includes the watershed. The state of land cover in spatial unit at i row and j column in time t is specied as Xijt. The state variable Xijt, dened in Eq. (1), is dependent on state of the spatial unit in time t 1 (Xij(t1)), dijt is the Markovian driver operating on spatial unit i, j at time t for the temporal processes, and X(i1)(j1)t is the states of spatial units neighboring the current spatial unit at time t.

X ijt f X ijt1 ; dijt ; X i1;j1;t

Let Yijt represent the outcome of hydrologic process g() in the spatial unit at i, j at time t as represented in Eq. (2). The process outcome of spatial unit is dependent on the state of the system Xijt and a policy variable Pijt representing conservation practices to change process outcomes in spatial unit i, j at t.

Y ijt g X ijt ; Pijt

The primary objective of the decision problems is in Eq. (3), which minimizes watershed outcomes that create negative externalities (such as storm water runoff, and pollution).

MinY ijt g X ijt ; Pijt

Other constraints to the decision problem include initial conditions and budget constraints. We focus on identifying Pijt at each spatial unit that minimizes negative externalities, especially soil loss and runoff generated by the watershed. The conceptual representation of the theoretical model is presented in Fig. 1. The spatiotemporal watershed dynamic model

Fig. 1. Spatiotemporal watershed dynamic model (SWD).

(SWD) consists of temporal and spatial drivers that operate on the baseline land use as the initial condition. Each spatial unit uses the two drivers to predict land use change in the next time step. The land use change is input into submodels to calculate runoff and sediment loading. This process is repeated on all spatial units in the modeling matrix of the watershed. The SWD model is applied to a small watershed in the Blackstone River Basin of central Massachusetts. This watershed is selected based on the diversity of land use types and the deterioration in water resources as a result of urbanization (Randhir, 2003). The watershed is divided into spatial units (grid objects) of 1 ha area, with four possible land use states: forest, agriculture, urban, and other. The initial states for each spatial unit are assigned to each grid object using GIS mapping of land use. Each grid object is modeled as automated agents that interact with neighboring grid agent. Transition coefcients from one land use state to another are based on historic land use probabilities. The SWD model is used for predicting the land use change and impacts on water resources at spatial and temporal dimensions. Predicted land use change is assumed to be based on current land use, knowledge of the past land use change, and the nature of spatial inuence. Land use transition (Clark and Mangel, 2000) is represented through a probabilistic progression from one state of land use to another through time. Land use change is dynamic with a cause-response behavior involving multiple variables: a cell may shift from one land use to another, driven by Markovian process and states of neighboring cells through cellular automation. The implementation of SWD as an object-based model (objects and submodels in a dynamic system) is modeled in a declarative modeling environment (Muetzelfeldt, 2004). Land use (state) change in a cell is inuenced by the interaction between temporal and spatial factors in the watershed. Modeling these complex changes is critical in evaluating emerging land use and potential problems in water quality in the watershed. In predicting land use change, two components that are integrated include: temporal change from one state to other and spatial dimension involving interaction between adjacent cells. The temporal change in land use is evaluated using transition probabilities evaluated from historic GIS land use datasets. The MCMC analysis and spatial analysis are used to model land use dynamics in geographic space and over time. The transition probabilities from the transition matrix are used in the MCMC analysis. The spatial dynamics are assessed using a cellular interaction of contiguous neighborhood cells (agents) through cellular automation. The modeling is performed using the SWD model for 3025 cells (agents representing a geographic unit) with geographic coordinates extracted from a land use map using GIS. Each cell is modeled in a state space to study the dynamics of shift in land use. The land use types are reclassied into four major categories; the proportional areas are calculated for each category, and the probabilities for each possible change are derived. The transition probabilities for the period 19711999 are used in the model. Watershed and matrix (square matrix that includes watershed cells) boundaries are converted to a grid format. Each grid represents a cell that is 100 m 100 m in size. The elevation layer is used to derive the aspect (ow direction) layer. The resulting grids are exported as ASCII text for use in the SWD model. The SWD model consists of cellular agents and processes as submodels (MCMC, land transition, neighborhood, runoff, and soil loss). The cell object is a xed-membership, multiple-instance submodel, and represents each land use location. Each cell type represents a land use state, which is arbitrarily labeled 1. for agriculture, 2. for forest, 3. for urban, 4. for others. Land use type cells are represented in a state equation that switch state based on incoming and outgoing land type, conditional on the outcome of MCMC

T.O. Randhir, O. Tsvetkova / Journal of Hydrology 404 (2011) 112

Fig. 2. Blackstone River watershed and subwatershed locus map.

and spatial process. An additional constraint is that any cell can only be in one state at any time. The land use, soil loss, runoff, and neighborhood submodels are specied as process equations and spatial relationships between objects. The runoff is assessed using the curve number method (USDA/SCS, 1972), while soil loss in each cell is estimated using the RUSLE method (USDA/NRCS, 2004). 2.1. Monte Carlo Markov Chain process The transition probability of each state denes the change in a land use patch. The transition probabilities derived from a time series of land use change in the study area (MassGIS) are incorporated into a transition matrix for MCMC analysis. The MCMC technique involves Markovian transitions of the land use at time (t) as dependent on land use at a previous time (t1) and a Monte Carlo process of simulation (Metropolis and Ulam, 1949). The likelihood of a change to a particular state is a combined outcome of MCMC and spatial inuence. Thus the specic state switches from 1 and 0 if the current land use switches to another type. Spatial modeling is in the form of disaggregating, with each spatial unit modeled as agents. Each spatial agent is given spatial attributes (area, location), and interactions between other spatial agents are represented (Muetzelfeldt and Duckham, 2005) using neighborhood association between the current cell and surrounding cells. A conditional evaluation is used to identify the neighboring status of each cell. The spatial dynamics are modeled through cellular interaction between contiguous neighboring cells.

2.1.1. Soil loss To model soil loss in each cell, the revised universal soil loss equation is used (USDA/NRCS, 2004), where soil loss is calculated as At = Rt Kt LSt Ct PRt , where At is the soil loss in tons per acre at time t, Rt is the rainfall factor at time t, Kt is the soil erodibility factor at time t, LSt is the land ow length and slope factor at time t, Ct is the cropping factor at time t, and the PRt is the support practice factor at time t. The Rt and Ct factors are derived from RUSLE2 software (USDA/NRCS, 2004). LSt factor is calculated as LSt = [0.065 + 0.0456(St) + 0.006541(St)2] (SLtK)NN, where, St = slope steepness (%), SLt = length of slope (ft.), and K = 72.5. The NN value varies from 0.2 to 0.5 with changes in value of St. The NN value equals 0.2 if St < 1. If 1 6 St < 3 then NN = 0.3. If 3 < St < 5, then NN = 0.4. NN = 0.5 for St P 5. The elevation layer is used to derive land slope and slope aspect using the BASINS software (USEPA, 2001). The data are transformed into ASCII les to calculate the LSt factor from the equation listed above. The results are specied as a table of LSt factors for each cell in the SWD model. The K-factor is derived from STATSGO data through BASINS (USEPA, 2001) from the soils attribute table and then clipped to the area of interest. The C values are assigned using conditional equations for each land use in the SWD model, which are estimated for each land use from the RUSLE2 software. The PRt value is set to no conservation for the baseline run, and the Rt value is calculated using the RUSLE2. 2.1.2. Runoff The runoff curve number method (USDA/NRCS, 1986) is used to estimate runoff from each patch. The curve number (CNt) method

T.O. Randhir, O. Tsvetkova / Journal of Hydrology 404 (2011) 112

(A) Agriculture

(B) Forest

(C) Urban

(D) Other

Fig. 3. Spatial distribution of baseline land uses (watershed shown in greyscale).

estimates direct surface runoff from a given amount of rainfall. This method incorporates soils permeability, land use, and antecedent soil moisture condition. The runoff is calculated as 2 t 0:2St Q t PP , where St 1000 10. Here, Qt is runoff in time t, CN t PPt 0:8St PPt is precipitation at time t, St is potential maximum retention (storage) after runoff begins, and CNt is the runoff curve number at time t. The SWD model (Fig. 1) is implemented in dynamic simulation software, SIMILE (Muetzelfeldt and Massheder, 2003; Muetzelfeldt and Taylor, 2001). SIMILE is a declarative modeling language which is appropriate to this study as it represents specications of the conceptual and mathematical structure of the model rather than a set of instructions (Muetzelfeldt, 2004) in a dynamic and spatial environment. It is also visual and declarative in approach (Muetzelfeldt and Taylor, 2001) that is useful in ecological and

environmental research and is used in a number of international research projects (Muetzelfeldt and Massheder, 2003). 2.1.3. Data Land use data from MassGIS has 37-categories for 1999, which are reclassied into four major land use types. The Ct values for soil loss estimation is estimated using RUSLE2 as 0.45, 1, 0.01, and 0 for agriculture, forest, urban, and other land uses, respectively. The Pt value is set at value 1 for the baseline run of the model, and the Rt value is calculated as 135 for the region using RUSLE2. The CNt values for the runoff estimation ranged from 30 to 100; lower numbers indicate low runoff potential while larger numbers are for increasing runoff potential. The model is run over 100 years at a yearly time step. The slope steepness can change in model time, with current policies and strict enforcement of BMPs in the study

T.O. Randhir, O. Tsvetkova / Journal of Hydrology 404 (2011) 112

(A) Land use distribution

cover for the same period is compared to existing land cover derived from GIS. The outputs from regional estimates (baseline states for the regional estimate for the watershed) and predicted (model run over 14 years) were analyzed in ArcMap using the geostatistical analysis tool. The Cohens Kappa Index (K) (Cohen, 1960) is calculated as K = (Po Pe)/(1 Pe), where Po is the observed match in prediction and Pe is hypothetical probability of agreement. Soil loss is validated by comparison with a sub-section time span (7 years from 1999 to 2005) prediction of subbasin outcome of sediment from a regional estimate (Randhir and Tsvetkova, 2009). Runoff is calibrated using 5 years data from 1999 to 2003 from Randhir and Tsvetkova (2009). Soil loss estimated by SWD model agreed with outcome for the watershed from regional estimates with an R2 of 0.59. Runoff estimates from SWD model was similar to regional estimates at R2 of 0.58.

3. Study area The study watershed is in the Blackstone River basin (Randhir, 2003) that is 48 miles long, owing from south-central Massachusetts into northeastern Rhode Island. The river basin (Fig. 2) has a historical role in the industrialization of the northeast, and is important to the environmental health of Narragansett Bay that receives the river waters. The Blackstone River Valley was formed by glacial action about fteen thousand years ago that slowly shaped the course of the Blackstone River. The Blackstone River drainage system is one of the seven major river systems of the northeast and its tributaries, banks and oodplains provide a rich habitat for ora, sh and wildlife (Wright et al., 2001). The current pattern of the landscape is a consequence of past patterns and land use activity in the region. During the 18th and 19th centuries, most forests in New England were partly cleared for agriculture, and partly harvested for wood products. By the late 19th century, agricultural lands were abandoned as a result of industrialization and forests grew over again. The modern vegetation in New England is a result of complex human-impacted disturbance histories, with a composition that developed as a result of prior land use (USEPA, 1999). Consequently, to forecast future patterns of land use and land-cover change, this research uses historical and contemporary drivers that operate at different spatial and temporal scales. The study watershed is located at the intersection of the four towns (Uxbridge, Northbridge, Sutton, and Douglas) located in Worcester County, Massachusetts. According to the 2000 Census, Uxbridge had a population of 11,156 with a total area of 78.7 km2. Northbridge had a population of 13,182 with a total area of 46.8 km2. The town of Sutton had a total population of 8250 with a total area of 87.9 km2. The town of Douglas had a population of 7045 with a total area of 97.7 km2. The total study area occupies 3172 ha in the modeling matrix and 1485 ha in the watershed itself. The study area is composed of a mixture of forest, agricultural, suburban and urbanized areas (Fig. 3). Most of the agricultural land is cropland and pasture and is focused in the northeast part of the area. Urban land use in the subbasin is generally concentrated in the southeast part of the area and represents 34% of the total area. Forested land is evenly distributed throughout the basin and tends to be along stream channels, especially in the southern and northern parts of the basin. Urbanization is a major problem in the watershed and impacts water resources through alterations in hydrology (Randhir, 2003), morphology, water quality (Randhir and Tsvetkova, 2009), and habitat in the watershed. Changes to the stream attributes are caused by an increase of impervious surface cover associated with the process of urbanization (Randhir, 2003; Schueler, 1992). Rainfall records from the National Weather Service (NWS) station in

(B) Soil loss

(C) Runoff
Fig. 4. Baseline spatial distribution of land use, soil loss, and runoff.

area we assumed no major slope changes over time. The system of differential equations is solved using RungeKutta solution method (Cartwright and Piro, 1992). The results are displayed within the Model Run Environment (MRE) of SIMILE platform as spatial grids and time series graphs. We use data from a regional study to calibrate and validate runoff and soil erosion estimates. The soil loss and runoff estimates from 1999 to 2005 of a regional study (Randhir and Tsvetkova, 2009) are used in calibrating and validating the SWD model. Results are calibrated through statistical evaluation of runoff and sediment estimates. Land use predictions are validated using the Kappa Index (Chust et al., 1999). The Kappa Index is calculated using MassGIS land use grids from 1985 and 1999. Simulated land

T.O. Randhir, O. Tsvetkova / Journal of Hydrology 404 (2011) 112

Cropland

Forest Urban

25 years

50 years

75 years

100 years
Fig. 5. Spatial land use distribution at varying time steps.

Fig. 6. Temporal land use change from baseline.

T.O. Randhir, O. Tsvetkova / Journal of Hydrology 404 (2011) 112

C
After 25 years of simulation After 50 years of simulation

After 75 years of simulation

After 100 years of simulation

Fig. 7. Soil loss at varying time steps.

Woonsocket, Rhode Island measure an average annual rainfall of 47.9 in/year or 121.67 cm/year. 4. Results and discussion 4.1. Baseline The land within the study watershed is predominately forested (46%) and distributed throughout the basin. The next largest land use is urban (34%) which is located mostly in the north and northeast parts, followed by agricultural land (12%) that is mostly in the northwest portion, and the remaining 8%is under other land use categories located mostly in the northern and western parts of the study area. The baseline land use distribution is presented in Fig 4a. The agricultural land use is distributed mostly on the northwest side of the watershed. It also has large contiguous and connected parcels going from the west side to the south end of the watershed. There is a small cluster of the cells in the north of the watershed

and several separated patches in the central part of the watershed. Forested land is mostly in south, west and central portions of the watershed with less density in the southern and northern end of the watershed. The baseline distribution of forest land is characterized by large connected and contiguous parcels. In the modeling matrix, outside the boundary of the watershed, agricultural land use is also dened by large connected cells in the north side and a number of contiguous parcels in the south. Urban land use is evenly distributed throughout the watershed with a concentration on the east portion of the watershed. Urban land use distribution consists of a large, contiguous region of connected cells with several disconnected cells in the central part of the watershed. Within the study matrix, urban land use is focused mostly in the southwest of the watershed. Land use categorized as Other is mostly located in the north and northeastern side of the watershed. Most of the parcels are connected and contiguous. The land use distribution of all types as simulated is presented in Fig 4a. There are three major soil classes in the study area with MUIDs MA007, MA014, and MA015. The soils texture in the study area is

T.O. Randhir, O. Tsvetkova / Journal of Hydrology 404 (2011) 112

After 25 years of simulation

After 50 years of simulation

After 75 years of simulation

After 100 years of simulation

Fig. 8. Runoff simulation at varying time steps.

composed of: ne sandy loam (FSL) at 28%; gravelly loamy sand (GR-LS) at 33%; and very stony ne sandy loam (STV-FSL) at 39% of the study watershed. The northern part of the watershed has a soil texture of the GR-LS type, while the southern, the southeast, and the northeast parts of the watershed have mostly a texture of STV-FSL type. The western, northwest, and southwest regions of the watershed are generally composed of FSL textured soils. All three types of soils texture are spatially distributed over the central part of the watershed. The baseline spatial distribution of soil loss in the watershed is presented in Fig. 4b. High soil loss is observed in areas located in the northwest and eastern parts of the study watershed, and there is also a long band of cells with high soil loss traversing from west to the southern part of the watershed. These areas have agricultural and urban land uses. A lower soil loss is observed in the central area of the watershed. The rate of soil loss is also variable throughout the watershed. This distribution can be explained with land use types where highest soil loss is associated with agriculture and urban areas, while forested areas had a lower rate of soil loss.

The spatial distribution of baseline runoff rates is presented in Fig. 4c. High runoff areas are located in the north and northeast of the watershed, and these areas are large, contiguous, and connected. These areas are associated with agriculture and urban land uses. The less intense runoff areas are mostly in the central and southern parts of the watershed with mostly forest cover. Runoff is correlated with land use type, with highest runoff areas associated with urban and agriculture and the lowest runoff areas in forests (Randhir and Tsvetkova, 2009). 4.2. Spatio-temporal simulation Results of spatial and temporal changes in land cover over 25year intervals are presented in Fig. 5. The results show a steady decrease in agricultural parcels in the watershed during the initial 25 years of simulation. The agricultural land is subject to fragmentation during the initial 25 years of the simulation. A comparison of the forest land use distribution shows that large, contiguous parcels that are observed in the central and

10

T.O. Randhir, O. Tsvetkova / Journal of Hydrology 404 (2011) 112

0.16 0.14 0.12

Soil Loss (t/Ha)

0.1 0.08 0.06 0.04 0.02 0 0 20 40 60 80 100

Time

(A) Soil Loss


107 106 105

Runoff (cm/Ha)

104 103 102 101 100 99 98 97 0 20 40 60 80 100

Time

(B) Runoff
Fig. 9. Temporal trend soil loss and runoff in the watershed.

southern parts of the watershed after 25 years of simulation are disappearing and dispersing over the years. It is observed that the forest land is fragmenting and declining in all areas of the watershed. The relatively low levels of parcels are accumulating in the northwest and eastern areas, which may indicate that forests are being replaced by urban areas. It is expected that high urbanization will dominate in eastern portions of the watershed. The pattern of rapid urbanization shows a trend toward suburbanization and later development of more dense urban areas. The urban pattern increased from the northeast portion to the central and other parts of the watershed. Urbanization trends show that after 100 years urban land use will be the major land use and lead to areas with the highest impervious cover in the watershed. The simulated land use over 100 years of the Other category shows a decrease in area over time. In general, land use changed over years and is concentrated in the different portions of the watershed over different simulation periods. In Fig. 6, the overall simulated land use change over 100 years is presented in the form of graph. Urban land increases rapidly during the rst 25 years to 25% of baseline and reaches a maximum increase of 35% of baseline. Forest land increases slightly during the rst 5 years and decreases gradually thereafter to 15% of the baseline. Agricultural land use decreases rapidly in the rst 10 years and reaches a maximum reduction of 12% of the baseline. The graph indicates that in comparison with the beginning period of time the results tend to increase

during the modeling period for urban land use and show a decrease in agricultural land use. There is a decline in the forest land use type over time. The spatial distribution of soil loss in each cell over 100 years is presented in Fig. 7. It can be observed that the soil loss expands from the middle section of the study area to the north and south sections over the simulation period. This is because of increased soil loss in cells changing from forest to urban land use (Randhir, 2003). These are also the most heavily urbanized areas of the watershed, especially in the central portion of the study area near the southeast boundary. Spatial distribution of runoff over the simulation period is shown in Fig. 8, which shows runoff during 25, 50 and 100 years compared to the baseline scenario. Runoff is highly correlated with land use types (Randhir and Tsvetkova, 2009) and is characterized by the distribution of large contiguous parcels throughout the watershed area with high levels in cells at the central section as well as in the east, northwest and southwest parts of the watershed. A high runoff is associated with the urbanization process in the watershed. The trend in sediment loss is depicted in Fig. 9a. The overall rate of overland sediment loss decreased in the rst 20 years to 0.06 t/ ha. The decrease remained at 0.05 t/ha from baseline over the following eight decades. The reduction in overland soil loss can be attributed to reduction in agricultural land and increased urban land that could have an armoring effect on land surface.

T.O. Randhir, O. Tsvetkova / Journal of Hydrology 404 (2011) 112

11

The total runoff is presented in the Fig. 9b. The results show an increase in runoff volume by 7 cm/ha over time and is associated with increase in urban land. It is observed that the increase in urban land use in the watershed is associated with decline in agricultural and forest areas. 4.3. Policy implications Results show a potential for rapid urbanization in the watershed that can result in water quality and storm water issues. Based on expected spatial land use change, site specic policies can be implemented in areas that have the potential to impact water resources. Nonpoint source reductions in the watershed can be achieved through the implementation of management measures that can reduce loads through land use policies and best management practices (BMPs) applied to sensitive sites. Three areas for targeting (AC in Fig. 7) are identied from sensitivity to sediment loading and runoff. It was observed that runoff and sediment impacts occur in early years of the land-cover change and need policies to protect from impacts. BMPs can reduce runoff and mitigate transfer of pollutants, reduce degradation of soil and water resources, and maintain inltration levels in the watershed. BMPs could include an array of practices to increase water inltration and reduce runoff that could be applied to urbanizing areas. Practices can also mitigate runoff, reduce the erosion, and attenuate transport of sediment from agricultural elds (USEPA, 2006). BMPs to reduce urban impacts on water resources include vegetative management practices such as shoreline revegetation, shoreline stabilization, urban forestry and urban practices such as porous pavement and water quality inlets. They may also consist of different structural practices such as water and sediment control basins, roof water collection devices, tree lter strips, terraces, diversions, grassed waterways, woodland fencing and wetland development. Another policy is to restrict land use and regulate land practices through zoning laws. Current zoning laws promote residential uses in most of the watershed. Zoning should be updated to reduce loss of open space under build-out conditions. Education in low impact development can also be useful in sensitive areas of the watershed. In summary, it is observed that spatial and temporal changes in land use have signicant effects on runoff in a watershed ecosystem. While the overall sediment loading is decreased, the spatial and temporal pattern of sediment loading is altered by land use change in the watershed. A spatially targeted approach to protect sensitive areas (both baseline and future potential) can be effective in enhancing water quality in the watershed. 5. Conclusion Urbanization and human population tend to grow and place demands on natural resources. Human activities can have a dramatic impact on water resources in the Blackstone River watershed. Dynamic simulation modeling is used to simulate changes that may occur to a subwatershed in Blackstone River watershed. Management of these resources and land areas is important for protecting water resources. With continued urban growth, policies to manage impervious areas are necessary to improve watershed hydrologic condition. The study models dynamic and spatial patterns of hydrologic processes of the watershed system. Spatio-temporal dynamic modeling is used to study the emergence of the landscape pattern over time. The model is run over a 100 year period and is used to assess potential land use change as a result of MCMC and spatial processes. The model is object-based and modular to represent change in land use states and hydrologic impacts on a yearly time-step.

The overall results show that land use modeling requires knowledge of multiple factors to determine land use change. Spatial and temporal changes in land use have signicant effects on a watershed ecosystem. The nature of spatial patterns is inuenced by temporal transitions and spatial inuences. Understanding the past and future impacts of changes in land cover is central to the study of land use change impacts on watershed system components. The modeling results show that the increase in urban land use in the watershed is associated with the decline in agriculture and forest land, indicating that urbanization could become a serious problem in the future. The results emphasize the need to protect agricultural area in rapidly changing watersheds. The highest soil loss is associated with agriculture areas, while forested areas had lower soil loss. A high runoff is associated with urban and agriculture types and a lower runoff in forested areas. There is the evidence that highest runoff and soil loss areas are associated with agriculture and early urban land uses. In summary, it is observed that spatial and temporal changes in land use are signicant in the watershed. There is a high potential for urbanization in most parts of the watershed. This urbanization could have signicant effects on runoff and inltration in the watershed ecosystem. Loss of recharge to groundwater can result in lowering of stream stages and groundwater levels. A higher volume of storm water can be expected from increased urban cover in the watershed due to a lesser lag time. While the overall sediment loading decreases at the watershed scale, there is potential for changes in spatial and temporal pattern of sediment loading. This potential has implications for stream morphology and erosion in certain sensitive areas of the watershed. A potential exists to anticipate and implement policies that mitigate urban inuences. Increasing inltration in recharge areas (Sekar and Randhir, 2006) can be an important strategy in urbanizing watersheds that can maintain groundwater and stream ows in the watershed. Runoff mitigation is critical to reduce stream bank erosion and loading of pollutants to the main stem of the river basin. A spatially targeted approach to protect sensitive areas (both baseline and potential) can be effective in enhancing water quality in the watershed. This study adds new knowledge to watershed literature through a dynamic assessment of impacts of land use changes on watershed hydrology. The spatio-temporal assessment allows for prediction of land use change and evaluation of potential water resource problems. Further research is needed in dynamic interaction involving multiple pollutants and under stochastic conditions. Such comprehensive assessments that result from spatio-temporal modeling are vital to achieving long-term sustainability of watershed systems. Acknowledgements We would like to thank the anonymous reviewers of the manuscript. This material is based upon work partially supported by the Cooperative State Research Extension, Education Service, US Department of Agriculture, Massachusetts Agricultural Experiment-Station, under Projects MA500864, MAS000943, NE-1024, NE-1044, MS 11, and MAS 00924. References
Agarwal, C. Green, G.M., Grove, J.M., Evans, T.P., Schweik, C.M.. 2000. A Review and Assessment of Land-Use Change Models: Dynamics of Space, Time and Human Choice. General Technical Report NE-297, Newtown Square, PA: US Department of Agriculture, Forest Service, Northeastern Research Station, pp. 61. Allen, J.S., Lu, K.S., Potts, T.D. 1999. A GIS-based analysis and prediction of parcel land use change in a coastal tourism destination area. Paper presented at the 1999 World Congress on Coastal and Marine Tourism Vancouver, British Columbia, Canada.

12

T.O. Randhir, O. Tsvetkova / Journal of Hydrology 404 (2011) 112 Muetzelfeldt, R., Massheder, J., 2003. The Simile visual modeling environment. European Journal of Agronomy 18, 345358. Muetzelfeldt, R., Taylor, J., 2001. Getting to Know Simile: the Visual Modeling Environment for Ecological, Biological and Environmental Research. Univ. of Edinburgh, Institute of Ecology and Resource Management. <http://www.ierm. ed.ac.uk/simile>. Munier, B., Schou, J.S., Birr-Pedersen, K., 2002. Ecological and economic modeling in agricultural land use scenarios. Ecological Modelling 174 (12), 518. Randhir, T.O., 2003. Watershed-scale effects of urbanization on sediment export: assessment and policy. Water Resources Research 39 (6), 113. doi:10.1029/ 2002WR001913. Randhir, T.O., Hawes, A.G., 2009. Watershed land use and aquatic ecosystem response: ecohydrologic approach to conservation policy. Journal of Hydrology 364, 182199. doi:10.1016/j.jhydrol.2008.10.017. Randhir, T.O., Shriver, D.M., 2009. Multiattribute optimization of restoration options: designing incentives for watershed management. Water Resources Research 45, W03405. doi:10.1029/2008WR007169. Randhir, T.O., Tsvetkova, O., 2009. Watershed-scale tradeoffs in water quantity and quality attributes for conservation policy. Water, Soil and Air Pollution 201 (14), 347363. doi:10.1007/s11270-008-9949-8. Randhir, T.O., Lee, J.G., Engel, B., 2000. Multiple criteria dynamic spatial optimization to manage water quality on a watershed scale. Transactions of the American Society of Agricultural Engineers 43 (2), 291299. Schueler, T.R., 1992. Mitigating the Adverse Impacts of Urbanization on Streams: a Comprehensive Strategy for Local Governments, Watershed Restoration Sourcebook. Metropolitan Council of Governments, College Park, Maryland, USA. pp. 2131. Sekar, I., Randhir, T.O., 2006. Spatial assessment of conjunctive water harvesting potential in watershed systems. Journal of Hydrology 334 (12), 3952. doi:10.1016/j.jhydrol.2006.09.024:1-14. Shriver, D., Randhir, T.O., 2006. Integrating stakeholder values with multiple attributes to quantify watershed performance. Water Resources Research 42 (8), 115. doi:10.1029/ 2005WR004413. Turner II, B.L., Meyer, W.B., 1991. Land use and land cover in global environmental change. Climatic Change 28, 4564. USDA/NRCS, 1986. Urban Hydrology for Small Watersheds. Technical Release 55 (TR-55), second ed. Natural Resources Conservation Service, Conservation Engineering Division. USDA/NRCS, 2004. RUSLE2 Instructions & User Guide, USDA-Natural Resources Conservation Service, Columbus, OH. USDA/SCS, 1972. National Engineering Handbook, United States Department of Agriculture, Soil Conservation Service, Section 4 Hydrology (Chapters 410). USEPA, 1999. Land Cover Trends: Rates, Causes, and Consequences of Late Twentieth Century US Land Cover Change. United States Environmental Protection Agency Ofce of Research and Development Washington, DC, 20460 EPA/600/R-99/105 June 1999. USEPA, 2001. Better Assessment Science Integrating Point and Nonpoint Sources BASINS, Version 3.0. Users Manual. Ofce of Water. EPA 823-B-01-001. USEPA, 2006. Management Measures for Agricultural Sources. USGS, 2005. Final Agenda for the USGS Chesapeake Bay Science Workshop, May 12 13. Vitousek, P.M., 1994. Beyond global warming: ecology and global change. Ecology 75, 18611876. doi:10.2307/1941591. Wang, Y., Zhang, X., 2001. A dynamic modeling approach to simulating socioeconomic effects on landscape changes. Ecological Modelling 140 (1), 141162. Wilkie, D., Finn, J.T., 1988. A spatial model of land use and forest regeneration in the Ituri forest of northeastern zaire. Ecological Modeling, 41. Wright, R.M., Nolan, P.M., Pincumbe, D., Hartman, E., Viator, O.J., 2001. Blackstone River Initiative: Water Quality Analysis of the Blackstone River Under Wet and Dry Weather Conditions. USEPA New England, Boston, MA.

Anwar, S.M., Borne, F. 2005. Companion modeling and multi-agent systems for integrated natural resource management in Asia Metro Manila: IRRI, pp. 237 253. Cartwright, J.H.E., Piro, O., 1992. The dynamics of RungeKutta methods. International Journal of Bifurcation and Chaos 2, 427449. Chust, G., DuCrot, D., Riera, J.L.L., Pretus, J.L.L., 1999. Characterizing humanmodelled landscapes at a stationary state: a case study of Minorca, Spain. Environmental Conservation 26, 322331. Clark, C.W., Mangel, M., 2000. Dynamic State Variables in Ecology: Methods and Applications. Oxford University Press, New York, Oxford, 289 pp. Cohen, J., 1960. A coefcient of agreement for nominal scales. Educational and Psychological Measurement 20, 3746. Costanza, R., Ruth, M., 1998. Using dynamic modeling to scope environmental problems and build consensus. Environmental Management 22 (2), 183195. Costanza, R., Voinov, A., 2001. Modeling ecological and economic systems with Stella: part III. Ecological Modelling 143, 1143. Costanza, R., Duplisea, D., Kautsky, U. (Eds.), 1998. Modelling Ecological and Economic Systems with Stella. Ecol. Model. 110, 1_/103 (Special Issue). Djodjic, F., Montas, H., Shirmohammadi, A., Bergstrm, L., Uln, B., 2002. A decision support system for phosphorus management at a watershed scale. Journal of Environmental Quality 31, 937945. Forrester, J.W., 1971. World Dynamics. Wright-Allen Press, Cambridge, MA. Heiken, G., Valentine, G., Brown, M., Rasmussen, S., George, D., Green, R., Jones, E., Olesen, K., Andersson, C., 2000. Modeling cities: the Los Alamos urban security initiative. Journal of Public Works Management and Policy 4, 198. Hobbs, R., 1997. Future landscape and the future of landscape ecology. Landscape and Urban Planning 37 (1), 19. IIASA, 1999. Chinas Changing Land, Options of International Institute for Applied Systems Analysis (IIASA), Laxenburg. Austria. Jane, M., 2003. Spatial analyses of logging impacts in Amazonia using remotely sensed data. Photogrammetric Engineering and Remote Sensing 69 (3), 275 282. Jones, P.D., New, M., Parker, D.E., Martin, S., Rigor, I.G., 1999. Surface air temperature and its changes over the past 150 years. Reviews of Geophysics 37 (2), 173 199. Laine, T., Busemeyer, J., 2003. Comparing agent-based learning models of land-use decision making. In: NAAC[A]SOS (North American Association for Computational Analysis of Social and Organizational Systems) Conference 2004 Proceedings. <http://www.casos.cs.cmu.edu/events/conferences/2004/2004_ proceedings/Laine_Tei.pdf>. Lovejoy, S.B., Lee, J.G., Randhir, T.O., Engel, B.A., 1997. Research needs for water quality management in the 21st century: a spatial decision support system. Journal of Soil and Water Conservation (JanuaryFebruary), 1821. Marshall, E., Randhir, T.O., 2008. Spatial modeling of land cover change and watershed response using markovian cellular automata and simulation. Water Resources Research 44, W04423. doi:10.1029/2006WR005514. Maxwell, T., Costanza, R., 1994. Distributed object oriented spatial ecosystem modeling. In: Proceedings of the Society for Computer Simulation Western Multiconference: Object Oriented Simulation Conference, January. 2426 1994. Tempe Arizona. McGarigal, Kevin, Marks, B.J., 1995. FRAGSTATS: Spatial pattern Analysis Program for Quantifying Landscape Structure, PNW-GTR_351, United States Department of Agriculture, Pacic Northwest Research Station, Oregon, USA, 1995. Metropolis, N., Ulam, S., 1949. The Monte Carlo method. Journal of the American Statistical Association 44 (247), 335341. doi:10.2307/2280232. Muetzelfeldt, R., 2004. Declarative Modeling in Ecological and Environmental Research. Position Paper EUR 20918. Luxembourg, Germany: DirectorateGeneral for Research, European Commission. Muetzelfeldt, R., Duckham, M., 2005. Dynamic spatial modelling in the simile visual modelling environment. In: Fisher, P.F., Unwin, D.J. (Eds.), Re-Presenting GIS. Wiley John Wiley & Sons, Chichester, United Kingdom, pp. 243256.

Das könnte Ihnen auch gefallen