Sie sind auf Seite 1von 13

Meteorol Atmos Phys (2009) 105:167–179

DOI 10.1007/s00703-009-0045-4

ORIGINAL PAPER

Role of cumulus parameterization schemes in simulating heavy


rainfall episodes off the coast of Maharashtra state during
28 June–4 July 2007
Odury S. R. U. Bhanu Kumar Æ Sai Ramalingeswara Rao Æ
Kailasam Muni Krishna

Received: 29 July 2008 / Accepted: 3 September 2009 / Published online: 18 September 2009
Ó Springer-Verlag 2009

Abstract Indian summer monsoon gives on an average out that the KF-1 experiment has the best skill in predicting
250 cm of rainfall due to mesoscale/synoptic scale systems heavy rainfall episodes.
over west coast of India; now-a-days, MM5 model plays a
very crucial role in simulating such heavy rainfall episodes
like Mumbai (India) on 26 July 2005, which caused 1 Introduction
devastation through flash floods. The main aim of this
study is to simulate such heavy rainfall episodes using Heavy rainfall events (C15 cm) are very frequent and
three different cumulus parameterization schemes (CPS) wide spread during June through September over west
namely Kain–Fritsch-1, Anthes–Kuo and Grell and to coast of India, in general, and Maharashtra state, in
compare their relative merits in identifying the character- particular, due to onset vortex, active phase of intra
istics of mesoscale systems over 14 stations in coastal seasonal oscillations (ISO), mesoscale systems and
Maharashtra state during 28 June–4 July 2007. MM5 mid-tropospheric cyclonic circulations. An active phase of
control experiment results are analysed for the fields of ISO caused heavy rainfall episodes during the last week
mean sea level pressure, wind, geopotential height at of June and first week of July 2007. To identify such
850 hPa and rainfall with the above schemes. It is inter- episodes, we need high-resolution observations and
esting to note that Kain–Fritsch-1 scheme simulates heavy numerical modelling technique for better predicting heavy
rainfall amount of 48 cm for an observed rainfall of 51 cm rainfall events and understand the evolution and devel-
in 24 h. The Grell scheme underestimates heavy rainfall opment mechanisms of mesoscale convection responsible
episodes, while the Anthes–Kuo scheme is found to over for heavy rainfall incidents (Routray et al. 2005; Mohanty
predict rainfall on both temporal and spatial scales. The 2006). In recent decade, there have been many numerical
reason for better performance of KF-1 scheme may be due studies related to convective weather and improvement of
to inclusion of updrafts and downdrafts. Later the simu- rainfall forecasts that have provided a better understand-
lated rainfall quantities at 14 stations over study region are ing of physical and dynamical processes (Bhanu Kumar
validated with both 3B42RT and observed rain gauge data et al. 2008; Mohanty 2006; Ratnam and Krishna Kumar
of India Meteorological Department (IMD) and the results 2005; Kerkhoven et al. 2006; Wang and Seaman 1997)
are promising. Finally, for the heavy rainfall prediction over India and aboard. Previous studies on meso-b scale
cases, the best threat score is at 0.25 mm threshold for systems, research has concentrated on the development of
three CPSs. Thus, this study is a breakthrough in pointing conceptual cloud models for modelling the entrainment–
detrainment rate at cloud lateral boundaries and properties
of updrafts and downdrafts (Kuo 1965, 1974; Arakawa
and Schubert 1974; Anthes 1977; Anthes and Warner
O. S. R. U. Bhanu Kumar (&)  S. Ramalingeswara Rao  1978; Fritsch and Chappell 1980; Frank and Cohen 1985;
K. Muni Krishna
Betts and Miller 1986; Kain and Fritsch 1990, 1993; Grell
Department of Meteorology and Oceanography,
Andhra University, Visakhapatnam 530 003, India and Kuo 1991; Grell 1993; Arakawa 1993; Mapes 2000;
e-mail: osrubhanukumar@yahoo.com Xu and Randall 2001).

123
168 O. S. R. U Bhanu Kumar et al.

The difference between the resolution of mesoscale Heavy rainfall episodes were recorded at three stations
models (20–50 km) and the scale of a cumulus cell only out of 14 stations (Table 2). On 3 July evening it
(1–10 km), mesoscale models require the use of cumulus persisted over the same area; however, the associated upper
parameterization schemes (CPS). These schemes must air cyclonic circulation was up to 4.5 km above sea level. It
define the trigger of convection, how convection modifies moved away on 4 July. The synoptic sequence of above
moisture and temperature in a column, and how convection mesoscale system during 28 June to 2 July 2007 is shown
interacts with grid-scale dynamics using the grid-scale in terms of Satellite IR imageries over the study region
information of the main model. The purpose of this study is (Fig. 3). It is interesting to note that on 29 June convective
to evaluate the influence of the choice of CPSs (Table 1) in cloud amount of was very high and the OLR value was
MM5 model to predict heavy rainfall events, with an 130 W/m2 over the study region, where the heavy rainfall
emphasis on the accuracy of predicted rainfall amounts in episode occurred.
terms of spatial and temporal distributions.

3 Data and model description


2 Synoptic situation
In this study, National Center for Environmental Prediction
An active phase of ISO plays an important role in causing (NCEP) Final Analysis (FNL) model initialization fields
heavy rainfall episodes through a mesoscale system over (1°91°) at 6 h intervals, daily NCEP-GDAS (1°91°) and
the west coast of India in the present study. On 28 June, a Tropical Rainfall Measuring Mission (TRMM)-3B42RT-
feeble north–south trough formed off Maharashtra coast merged rainfall data sets are downloaded, while daily rain
and Chiplun station received 13 cm of rainfall. Next day gauge data set is obtained from the IMD for the period 28
(Figs. 1, 2) it has extended from North Maharashtra coast June–4 July 2007. The state-of-the-art to simulate the heavy
to Kerala coast on surface chart and the associated upper rainfall events in the present study is version 3 of the fifth
air cyclonic circulation extended over it up to 4.5 km generation of the Pennsylvania State University–National
above sea level, which resulted in 51 cm of rainfall in Center for Atmospheric Research MM5 model (Grell et al.
Mahad, while the neighbouring 11 stations also observed 1994; Dudhia et al. 2004) with three different CPSs, while
heavy rainfall amounts in the range 20–36 cm. On 30 June the following components of the scheme are common. The
morning, trough was further extended from North horizontal resolution of 45, 15 and 5 km is initially used on
Maharashtra coast to Lakshadweep area and corresponding a triple nested grid (Fig. 4). In the vertical, 23 r levels were
associated upper air cyclonic circulation also extended up used with increased resolution in the lowest kilometer in
to 7.6 km above sea level and on this day, ten stations order to adequately resolve the boundary layer. Topography
over the study region recorded 20–42 cm of rainfall. The is taken from the USGS 5-min resolution data set and
upper air cyclonic circulation over Saurashtra and interpolated onto the model grid using an overlapping
adjoining areas emerged with the above system on 30 June parabolic interpolation method. Initial and boundary
evening. conditions (atmospheric variables, soil moisture and
Above north–south, trough has confined south Gujarat to temperature) were interpolated from the NCEP FNL data at
Kerala coast on 1 July with a decreased vertical extent and 21 standard pressure levels under 100 hPa onto the MM5
intensity and six stations received greater than 20 cm of grid. Boundary layer processes were parameterized using
rainfall. Later it further weakened into a low pressure area the MRF scheme (Hong and Pan 1996), while surface, land
on 2 July evening and centred over West Rajasthan and temperature and moisture were calculated using the five-
adjoining areas and it persisted on 3 July morning tilting layer soil model (Dudhia 1996), which considers soil
southwards with height. properties up to 2 m below the surface. Cloud radiation

Table 1 Summary of cumulus parameterization schemes used in this study


CP schemes Reference Cloud model Closure assumption

Anthes–Kuo Anthes (1977) Simple 1-D, no downdraft Precipitation is proportional to large-scale moisture
convergence
Grell Grell (1993) 1-D updraft-downdraft couplet, The rate of production of instability at the large
no lateral entrainment scale equals the rate of removal of instability at the
small scale
Kain–Fritsch-1 Kain and Fritsch (1990) 1-D entraining–detraining plume, Convection is proportional to the available buoyant
with downdraft energy

123
Role of cumulus parameterization schemes 169

Fig. 1 Streamlines at 850 hPa from NCEP-GDAS valid at 0000 UTC during 28 June 2007–4 Jul 2007

123
170 O. S. R. U Bhanu Kumar et al.

Fig. 2 Streamlines at 200 hPa from NCEP-GDAS valid at 0000 UTC during 28 June 2007–4 Jul 2007

123
Role of cumulus parameterization schemes 171

Table 2 IMD rain gauge and model rainfall data


S. no. Station IMD rain gauge (Model-KF) (cm)
28-06-07 29-06-07 30-06-07 01-07-07 02-07-07 03-07-07 04-07-07

1 Chiplun 13 (11, 8, 29) 22 (19, 16, 49) 39 (35, 28, 52) 19 (17, 11, 34) – – –
2 Mahad – 51 (48, 38, 79) 42 (40, 26, 62) – – – –
3 Mhasala – 36 (31, 14, 69) 30 (28, 11, 53) – – – –
4 Poladpur – 34 (30, 12, 73) 35 (31, 13, 58) – – – –
5 Mahabaleshwar – 31 (27, 10, 59) 32 (29, 11, 62) – 33 (31, 11, 65) 20 (17, 5, 44) 20 (17, 5, 44)
6 Murud – 25 (22, 8, 50) 33 (30, 11, 64) – – – –
7 Mangaon – 24 (22, 10, 45) 27 (25, 10, 55) – – – –
8 Bhira – 23 (20, 9, 42) 21 (17, 7, 46) 24 (21, 7, 51) 21 (18, 7, 55) 29 (25, 11, 61) –
9 Rajapur – 22 (18, 8, 50) 30 (28, 9, 62) – – – –
10 Dapoli – 21 (19, 7, 41) 27 (23, 8, 58) – 18 (15, 6, 44) – –
11 Alibag – 20 (17, 11, 44) – – – – –
12 Hamai – 17 (14, 8, 43) – 17 (15, 8, 44) – – –
13 Santacruz – – 17 (14, 5, 33) 26 (22, 9, 59) – – –
14 Ratnagiri – – 17 (14, 6, 39) – – –
15 Thane – – – 33 (30, 10, 61) – – –
16 Taloda – – – 22 (20, 8, 50) – – –
17 Akkalkuwa – – – 24 (21, 9, 53) 25 (22, 8, 49) – –
18 Navapur – – – – 17 (15, 5, 39) 18 (16, 7, 42) –
The values in bold, italic and bold italic fonts are KF-1, GR and AK, respectively

Fig. 3 Dundee satellite IR


images during 29 June–2 July
2007 at 0600 GMT

123
172 O. S. R. U Bhanu Kumar et al.

Fig. 4 Computational domains


for the 45-, 15- and 5-km grid
sizes used in the study (numbers
in the second figure represent
IMD rain stations)

Fig. 5 Simulated sea level pressure for a NCEP-GDAS, b Kain–Fritsch, c Grell and d Anthes–Kuo schemes valid for at 0000 IST 29 June 2007

interaction is allowed between explicit cloud and clear air. which uses predictive equations for cloud water, rainwater,
Moist convection was parameterized using AK, Grell (1993) snow and ice, has been performed using the simple ice
and KF-1. The explicit microphysics parameterization, scheme of Dudhia (1996).

123
Role of cumulus parameterization schemes 173

Fig. 6 Streamlines at 850 hPa for a NCEP-GDAS, b Kain–Fritsch, c Grell and d Anthes–Kuo Schemes valid at 0000 IST on 29 June 2007

CP schemes have significant impacts on model predictions, precipitating systems. Then, the categorical statistics scores
just as cumulus convection affects the atmospheric (Hamill 1999; McBride and Ebert 2000) such as threat score
circulation by redistributing heat, moisture and momentum. (TS), equitable threat score (ETS), and bias score (BS) are
Adjustments to the atmosphere made by different CP evaluated for several threshold values (0.25, 2.5, 10, 15, 20
scheme can vary significantly. For the above reasons, we and 25 mm), based on the 6-h rainfall forecast by each CPS
choose three schemes namely AK, Grell and KF. AK experiment on the 5-km grid points. Note that total (both
scheme is a modification by Anthes (1977) of the Kuo parameterized and resolved) rainfall is evaluated. Because
(1965). Convection is determined from the vertically there are only a few 45- and 15-km grid points in Maha-
integrated moisture convergence. Grell scheme is a simple rashtra, categorical statistics scores with such a limited
single cloud version of the Arakawa–Schubert scheme sampling of points probably do not have much statistical
designed to avoid first-order sources of error (Grell 1993; meaning; thus the 45- and 15-km verification results over
Grell et al. 1994). The KF scheme (Kain and Fritsch 1990) the Maharashtra area are not discussed. Definitions and
is an extension of the Fritsch and Chappell scheme. It uses descriptions of the categorical statistical scores, and rain-
the same assumptions about the removal of CAPE, fall-amount parameters are given in ‘‘Appendix’’.
convective time scale, triggering of convective and
precipitation efficiency.
The model prognostic rainfall distribution is first 4 Results
validated against satellite imagery and objective analyses,
to make sure that the model has adequate skill to reproduce MM5 control experiments with three different CPS were
the mesoscale-scale features with embedded mesoscale performed for mean sea level pressure (MSLP), wind,

123
174 O. S. R. U Bhanu Kumar et al.

Fig. 7 Geopotential height (m) at 850 hPa for a NCEP-GDAS, b Kain–Fritsch, c Grell and d Anthes–Kuo schemes valid at 0000 IST on 29 June
2007

geopotential height and rainfall over off the coast of simulations show the trough formation along the west coast
Maharashtra during 28 June–4 July 2007 and the results are of India that extended from north Maharashtra coast to
analysed and discussed as follows. Kerala coast. This feature is also clearly supported by the
NCEP-GDAS trough shown in Fig. 5a. The manifestation
4.1 Mean sea level pressure of trough is most prominent in KF-1 (Fig. 5b). Although all
the figures show similar pressure pattern, there are a defi-
Figure 5 represents NCEP-GDAS and simulated MSLP nite differences in the pressure gradients along the west
valid at 0000 IST of 29 June 2007. Figure 5a shows the coast from Kerala coast to north Maharashtra coast. Grell
subjectively analysed MSLP by NCEP-GDAS, while and AK simulations establish a pressure change of 4 hPa
Fig. 5b–d depicts the model-simulated MSLP with KF-1, from southern part coastal Karnataka to north Maharashtra
Grell and Anthes–Kuo simulations, respectively. Comparing coast where as the northward decrease of pressure over same
the simulations from all the numerical experiments with the region is about 6 hPa in KF-1 simulations. Correspondingly,
NCEP-GDAS, it is noted that the general characteristics in KF-1 simulations generate strong pressure gradients
MSLP pattern over the study region is fairly captured in the along the west coast of India compared to Grell and AK
KF-1 simulations. It is noticed that over the Arabian Sea, simulations. Henceforth, the location of isobars and their
isobars are nearly parallel to each other and oriented along corresponding orientations in KF-1 fairly match with
the monsoonal flow during this period. The isobars in all NCEP-GDAS observations.

123
Role of cumulus parameterization schemes 175

Fig. 8 Convective instability (the difference in equivalent potential temperature (he) between 1,000 and 500 hPa in °K) for a Kain–Fritsch, b
Grell and c Anthes–Kuo schemes valid at 0000 IST on 29 June 2007

4.2 Upper air circulation turning of streamlines at 850 hPa is not very clear. The rate
of decrease in the geopotential height on 29 June 2007
Having studied surface distribution of pressure over the using KF-1 experiment resembles very well with NCEP-
study region, it is interesting to study vertical structure of GDAS. The simulated geopotential heights due to Grell
the mesoscale system using upper air data. Figure 6 and AK schemes are higher when compared to the NCEP
describes streamlines at 850 hPa valid at 0000 IST of 29 reanalysis.
June 2007. Figure 6a represents the streamlines from
NCEP-GDAS, where as KF-1, Grell and AK simulates are 4.3 Rainfall episodes
depicted in Fig. 6b–d, respectively. Figure 7 is same as
Fig. 6, but for geopotential height. A comprehensive Results from MM5 model simulations with KF-1 scheme
analysis of upper air circulation changes is presented in highlight that localized distribution of rainfall of 48 cm at
Fig. 6, which depicts a zonal flow pattern over the Arabian Mahad station in west coast for the observed IMD rainfall
Sea, which after crossing the west coast line acquires a of 51 cm on 29 June 2007 in 5-km resolution which is
meridional component towards the south. relatively better simulated than 45- and 15-km resolutions
The streamlines mostly follow geopotential height over to obtain the heavy rainfall episodes. Similarly, 13 other
the sea region in KF-1 simulations. Close observation of stations also simulated heavy rainfall which is very close to
near surface MSLP pattern and geopotential height at rain gauge data on the same day. Hence the specifications
850 hPa indicate the presence of trough. However, the of the experiments are satisfactory and repeated for the

123
176 O. S. R. U Bhanu Kumar et al.

Fig. 9 Comparison between


24 h accumulated rainfall (cm)
of model simulation (KF-1) and
3B42RT

123
Role of cumulus parameterization schemes 177

other days and the results are presented in Table 2. This experiments, the timing and downdrafts might cause such
study is similar to Routray et al. (2005), who assimilated bad rainfall forecasts than KF-1 experiment. Because
heavy rainfall events with the coarse resolution along the summer monsoon rainfall was mainly generated by tropical
west coast of India. An area of convectively unstable disturbances, active phases of ISOs and onset vortex, which
stratification [the difference in equivalent potential the 5-km MM5 could not adequately represent with such a
temperature (he) between 1,000 and 500 hPa] existed over coarse horizontal resolution, the model had poor TS
the heavy rainfall area near to west coast of India. This performance.
instability arose from warm moist air extending down the Table 3 reveals ETS and BS averaged over six 6-h
west coast towards Maharashtra with a tongue of high he periods at different thresholds for heavy rainfall episodes.
air present between 17°N and 19°N (Fig. 8). Thus, an Table 3a indicates that for heavy rainfall, the ETSs of all
extreme rainfall event was the result of warm moist air experiments were similar (around 0.2 at 0.25 mm threshold).
being advected onto coast of Maharashtra. For rainfall thresholds of 15 mm and above, all
Simulated rainfall using MM5 model with different experiments with near zero or negative ETSs showed little
CPSs is discussed as follows: Simulated rainfall is further predictive skill, consistent with the findings of Stensrud
compared with observational 24 h rainfall valid during et al. (2000) and Gallus and Segal (2001) that models with
heavy rainfall episodes separately. Although the relative grid sizes of 5–20 km have little skill for 6-h rainfall
paucity of rainfall stations prevents a detailed comparison, amounts of 12.5 mm or above. Table 3b further indicates
Fig. 9 gives an indication of the model performance for that MM5 model with AK overpredicted the rainfall
rainfall. The rainfall recorded over off coast of Maharashtra (with BSs more than one at 0.25 mm threshold), and
is 12–51 cm whereas the model simulated showed a underpredicted with Grell scheme (with BSs less than one
maximum of 10–48 cm. However, it is found that the for thresholds of 5 mm and more). For the BSs by three
timing of the most intense rainfall in the model compared CPSs for the heavy rainfall cases in Maharashtra, GR has a
well with that observed and the results of simulated rainfall clear under forecasting. It may be due to the various con-
are presented in Table 2. Present simulated study clearly vection nature, and different geographical location. For
demonstrates that the rainfall maximum episodes amount total rainfall amount prediction, KF has the best skill
up to 85% of observational value over the study region. among three CPSs.
Significance of model output performance of rainfall is also Finally, the model rainfall results with different CPS
discussed in terms of statistical skill scores. are validated (Table 3) with both the IMD rain gauge
observations and TRMM-merged rainfall product 3B42RT
4.4 Categorical statistics scores

Figure 10 shows TSs at the 0.25 mm threshold for 6-h Table 3 (a) Equitable threat scores and (b) bias scores averaged over
six 6-h periods for different precipitation thresholds of heavy rainfall
rainfall predictions. The 5-km MM5 using three CPSs
event
generally show good predictive skill for coverage of
measurable 6-h rainfall off coast of Maharashtra region, Threshold (mm) AK scheme GR scheme KF-1 scheme
with TSs greater than 0.4 at the rainfall threshold 0.25 mm. (a)
The model performed relatively poor for the AK and Grell 0.25 0.2041 0.2321 0.1978
2.5 0.1824 0.1975 0.1892*
5 0.1021 0.0924 0.1359*
10 0.0957 0.0125 0.0315*
15 0.0153 0.0071 0.0209*
20 0.0024 -0.0022 0.0156*
25 -0.0052 -0.0042 0.0021
(b)
0.25 1.2122 1.1622 1.1349*
2.5 0.7124 0.9965 1.0178*
5 0.5761 0.4954 0.6874
10 0.4622 0.3125 0.7216*
15 0.1825 0.1171 0.3697*
20 0.2522 0.2264 0.4511*
Fig. 10 Threat scores (TSs) at the 0.25-mm threshold for 6-h rainfall 25 0.2827 0.2679 0.5721*
predictions from 5-km MM5 runs. The times on the abscissa are
relative to the model’s initial time * The score indicates best forecast at a given precipitation threshold

123
178 O. S. R. U Bhanu Kumar et al.

(Fig. 7) for the same episodes off coast of Maharashtra. It Table 4 Rain contingency table used for verification
is noticed that the 3B42RT data set was underestimated as Forecasted Observed
compared to the observed rainfall. Similar type of work
was done by Yang and Tung (2003) over Taiwan region. Rain No rain

Rain A B
No rain C D
5 Discussions and conclusions Each element of the matrix (A, B, C and D) holds the number of
occurrence in which the observations and/or the model forecasts reach
Our study is based on a single localized event so that the a rainfall threshold amount for a given forecast period
findings are somewhat approximate in nature, and further
work is needed to fulfil our conclusions. The AK and Grell number of occurrences in the ‘‘A’’ element in Table 4 is
schemes are unable to produce sufficient rainfall than KF increased by one. Letters A, B, C and D are used as in
because of the inclusion of entrainment/detrainment Wilks (1995) and Colle et al. (1999).
processes. Statistical verification scores used in this study Based on this contingency table, one can calculate
do not always provide a fully ‘‘objective’’ assessment of a several evaluation parameters. The BS is defined as
rainfall forecast. F AþB
Development of better evaluation methods for high- BS ¼ ¼
O AþC
resolution nonhydrostatic models is greatly needed in the
future. The information obtained from these experiments where F is the number of grid points where rainfall was
should be useful for improvement of CPS intended for forecasted to exceed a given threshold, and O is the number
operational usage at a grid size of 5 km in a tropical of observed grid points where rainfall exceeded the
environment with substantial topography, or the Indian threshold. The ETS measures the skill in predicting a
region in particular. given threshold at a location and is defined as
In conclusion, quantitative rainfall forecasting in an HE AE
ETS ¼ ¼
operational model setting remains a very difficult task. FþOHE AþBþCE
There is no single cumulus scheme that consistently
where H is the number of grid points where rainfall was
outperforms the others in all evaluation parameters of the
correctly forecasted to exceed a given threshold (a ‘‘hit’’),
rainfall forecast and across all weather systems. Finally,
F and O are defined above, and the random-guess number
much work remains to be done to adequately understand
E is defined as
how a CPS interacts with other components in a numerical
weather prediction model. FO ðA þ BÞðA þ C Þ
E¼ ¼
N N
Acknowledgments The authors are thankful to D.R. Sikka, India,
where N is the total number of grid points verified
T.N. Krishnamurty, Florida State University, USA and D.B. Rao
NOAA, USA for their academic encouragement and financial support (A ? B ? C ? D). The ETS is similar to the TS, except
through research project (11/MRDF/1/41/P/08) from the Department that the ETS corrects for the expected number of hit by
of Science and Technology, Government of India, New Delhi. The chance (E).
authors are also thankful to Mesoscale group of UCAR and NCEP-
NCAR for providing the MM5 model and FNL. Two anonymous H A
TS ¼ ¼ :
reviewers provided helpful comments, which significantly improved FþOH AþBþC
the manuscript.

References
Appendix
Anthes RA (1977) A cumulus parameterization scheme utilizing a
Some evaluation parameters used in this study are derived one-dimensional cloud model. Mon Weather Rev 105:270–286
using a contingency table (Wilks 1995; Colle et al. 1999; Anthes RA, Warner TT (1978) Development of hydrodynamic
models suitable for air pollution and other meso meteorological
Hamill 1999). This table represents a 2 9 2 matrix
studies. Mon Weather Rev 106:1045–1078
(Table 4), where each element of the matrix holds the Arakawa A (1993) Closure assumptions in the cumulus parameter-
number of occurrences, in which the observations and the ization problem. In: Emanuel KA, Raymond DJ (eds) The
model did, or did not reach a certain threshold amount of representation of cumulus convection in numerical models.
American Meteorological Society, Boston, pp 1–15
rainfall for a given period of forecast (6 h in this study).
Arakawa A, Schubert WH (1974) Interaction of a cumulus cloud
For example, if both the observation and the model verified ensemble with the large-scale environment. Part 1. J Atmos Sci
at that point reach or exceed the threshold criteria, the 31:674–701

123
Role of cumulus parameterization schemes 179

Betts AK, Miller MJ (1986) A new convective adjustment scheme. Raymond DJ (eds) The representation of cumulus convection
Part II: single column tests using GATE wave, BOMEX, ATEX in numerical models. American Meteorological Society, Boston,
and arctic air-mass data sets. Q J Roy Meteorol Soc 112:693–709 pp 165–170
Bhanu Kumar OSRU, Muni Krishna K, Ramalingeswara Rao S Kerkhoven E, Gan TY, Shiiba M, Reuter G, Tanaka K (2006) A
(2008) Simulation of environmental heavy rainfall episodes comparison of cumulus parameterization schemes in a numerical
during June and July 2006—a case study. Canad J Pure Appl Sci weather prediction model for a monsoon rainfall event. Hydrol
2(1):211–220 Process 20:1961–1978
Colle BA, Westrick KJ, Mass CF (1999) Evaluation of MM5 and Kuo HL (1965) On formation and intensification of tropical cyclone
Eta-10 precipitation forecasts over the Pacific northwest during through latent heat release by cumulus convection. J Atmos Sci
the cool season. Weather Forecast 14:137–154 22:456–475
Dudhia J (1996) A multi-layer soil temperature model for MM5. Kuo HL (1974) Further studies of the parameterization of the
Preprints, the sixth PSU/NCAR Mesoscale Model User’s Work- influence of cumulus convection on large scale flow. J Atmos Sci
shop. National Center for Atmospheric Research, Boulder 31:1232–1240
Dudhia J, Gill D, Manning K, Wang W, Bruyere C (2004) PSU/ Mapes BE (2000) Convective inhibition, subgrid-scale triggering
NCAR mesoscale modelling system tutorial class notes and energy, and stratiform instability in a toy tropical wave model.
users’ guide (MM5 modelling system version 3). National J Atmos Sci 57:1515–1535
Center for Atmospheric Research, Boulder. http://www.mmm. McBride J, Ebert EE (2000) Verification quantitative forecasts from
ucar.edu/mm5/documents, 27 September 2004 operational numerical weather prediction models over Australia.
Frank WM, Cohen C (1985) Properties of tropical cloud ensembles Weather Forecast 15:103–121
estimated using a cloud model and an observed updraft Mohanty UC (2006) Challenges in simulation and forecasting
population. J Atmos Sci 42:1911–1928 extreme weather events. National workshop on dynamics and
Fritsch JM, Chappell CF (1980) Numerical prediction of convectively simulation of extreme rainfall C-MMACS, Bangalore, 16–17
driven mesoscale pressure system. Part 1. J Atmos Sci 37:1722– March 2006
1733 Ratnam JV, Krishna Kumar K (2005) Sensitivity of the simulated
Gallus WA, Segal M (2001) Impact of improved initialization of monsoons of 1987 and 1988 to convective parameterization
mesoscale features on convective system rainfall in 10-km Eta schemes in MM5. J Climate 18:2724–2743
simulations. Weather Forecast 16:69–680 Routray A, Mohanty UC, Das AK, Sam NV (2005) Study of heavy
Grell GA (1993) Prognostic evaluation of assumptions used by rainfall event over west coast of India using analysis nudging in
cumulus parameterizations. Mon Weather Rev 121:765–787 MM5 during ARMEX-I. Mausam 56:107–120
Grell GA, Kuo YH (1991) Semi prognostic tests of cumulus Stensrud DJ, Bao JW, Warner TT (2000) Using initial condition and
parameterization schemes in the middle latitudes. Mon Weather model physics perturbations in short-range ensembles. Mon
Rev 119:5–31 Weather Rev 128:2077–2107
Grell GA, Dudhia J, Stauffer DR (1994) A description of the fifth- Wang W, Seaman NL (1997) A comparison study of convective
generation Penn State/NCAR mesoscale model (MM5). NCAR parameterization schemes in a mesoscale model. Mon Weather
Technical Note 398. National Center for Atmospheric Research, Rev 125:252–278
Boulder Wilks DS (1995) Statistical methods in the atmospheric sciences.
Hamill TM (1999) Hypothesis tests for evaluating numerical precip- Academic Press, Dublin, p 467
itation forecasts. Weather Forecast 14:155–167 Xu KM, Randall DA (2001) Updraft and downdraft statistics of
Hong SY, Hl Pan (1996) Nonlocal boundary layer vertical diffusion simulated tropical and midlatitude cumulus convection. J Atmos
in a medium-range forcast model. Mon Weather Rev 124:2322– Sci 58:1630–1649
2339 Yang MJ, Tung QC (2003) Evaluation of rainfall forecasts over
Kain JS, Fritsch JM (1990) A one dimensional entraining/detraining Taiwan by four cumulus parameterization schemes. J Meteorol
plume model and its application in cumulus parameterization. Soc Jpn 81(5):1163–1183
J Atmos Sci 47:2784–2802
Kain JS, Fritsch JM (1993) Convective parameterization for meso-
scale models: the Kain–Fritsch scheme. In: Emanuel KA,

123

Das könnte Ihnen auch gefallen