پديد آورندگان :
بذرافشان، جواد نويسنده دانشگاه سيستان و بلوچستان Bazrafshan, J. , خليلي، علي نويسنده گروه مهندسي آبياري و آباداني- دانشكده مهندسي خاك و آب- پرديس كشاورزي و منابع طبيعي دانشگاه تهران Khalili, A. , هورفر، عبدالحسين نويسنده دانشكده مهندسي آب و خاك دانشگاه تهران Hoorfar, A. , ترابي، صديقه نويسنده وزارت نيرو , , حجام، سهراب نويسنده دانشگاه آزاد اسلامي واحد علوم تحقيقات دانشكده علوم پايه تهران Hajjam, S.
كليدواژه :
مقايسه مدل , ايران , مولدهاي وضع هوا , تنوع اقليمي , Climatic diversity , model comparison , Weather generators
چكيده لاتين :
Introduction
A stochastic weather generator is a numerical model
which produces synthetic daily time series of a suite of
climate variables with certain statistical properties such
as precipitation, temperature, and solar radiation.
Weather generators are widely used by researchers
from many different baekgrounds such as decision
support systems 10 agriculture and hydrology.
Stochastic weather generators were originally
developed for two main purposes: 1) To provide means
of simulating synthetic weather time series with
statistical characteristics corresponding to the observed
statistics at a site not long enough to be used in
hydrological or agricultural risk assessments and 2) To
provide means of extending the simulation of weather
time series to unobserved locations, through the
interpolation of the weather generator parameters
obtained from models at the neighboring sites.
Objectives
The aim of designing weather generators is to produce
the synthetic weather data which is statistically similar
to the observed data. The purpose of this study was to
test and compare two weather generators in simulating
the weather factors including daily total precipitation,
the minimum and maximum air temperatures, and the total solar radiation for diverse climates of Iran.
The ClimGen (Stockle et aI., 1999) developed in the
USA and the LARS-WG (Semenov et aI., 1998)
developed in Europe are compared in this study.
Metbodology
The two weather generators (ClimGen and LARS-WG)
work in a similar way. They analyze certain statistical
properties of the input observed daily weather data for
the chosen site and then by using these properties along
with pseudo-random number generation, produce
simulated weather data one day at a time. These
weather generators were run at 15 selected stations for
this study. The process of generating synthetic weather
data was divided into three distinct steps including
model calibration, model validation, and long-term
simulation of weather data. From 4S years of historical
data in each station, 40 years were used for calibrating
the two weather generators and the rest for validating
the models. For each of the 15 stations, 300 years of
daily weather data were generated using ClimGen and
LARS-WG. To evaluate the agreement between
observed and generated data, two indices were used;
the Root Mean Square Error (RMSE) and the
Coefficient of Determination (CD). The later differs
from R-ʹ . Indeed, a number of statistical tests including
t-student test. F test, and X.:.ʹ test were carried out to
compare a variety of characteristics of the data.
Results and Discussion
The results for generated and observed daily weather
data in calibration. validation, and long-term simulation
of two weather generators are compared in Tables 1 and 2. As shown in these tables in the calibration step,
LARS-WG has less error than ClimGen for modeling
the daily precipitation data (Table I) and wet and dry
spells, frequency distribution, the monthly mean and
variance of precipitation data, and the daily variance of
solar radiation data in stations of interest (Table 2).
ClimGen presented appropriate results for estimating
the daily and monthly mean of minimum and maximum
temperature, the monthly variance of minimum and
maximum temperature, heat and frost spells. the daily
and monthly mean of solar radiation, and the monthly
variance of solar radiation (Table 2). The same results
were obtained in models validation, but the level of
errors increased. In long-term simulation of weather
factors, It is revealed that LARS-WG has a better
performance for generating the synthetic precipitation
data. The ClimGen generated more acceptable
synthetic temperature data. Reciprocally, both models
failed to represent characteristics of the observed solar
radiation data. Conclusion
In this study two models of ClimGen and LARS-WG
were compared to simulate a suite of weather data
including daily total precipitation, minimum and
maximum air temperatures, and total solar radiation at
15 meteorologieal stations. Before long-term
simulation of the synthetic weather data, the calibration
and validation processes of both models were carried
out. To evaluate the agreement between observed and
synthetic data, a number of statistical tests and some
error indices were used. The results of the study
recommend LARS-WG for simulating the synthetic
precipitation data and the ClimGen for simulating
temperature data. Neither of the generators have
superiority in simulating solar radiation data.
Keywords: Weather generators, Model comparison,
Climatic diversity.
References
Stoekle, C.O., Campbell, G.S. and Nelson, R. (1999).
ClimGen manual. Biological Systems Engineering
Department, Washington State University, Pullman.
WA,28p.
Semenov, M.A., Brooks, RJ., Barrow, E.M. and
Richardson, C.W. (1998), Comparison of the
WGEN and LARS-WG stochastic weather
generators for diverse climates. Climate Research,
ʹ
