najmeh khalili; Kamran Davary; Amin Alizadeh; Hossein Ansari; Hojat Rezaee Pazhand; Mohammad Kafi; Bijan Ghahraman
Abstract
Introduction:Many existing results on water and agriculture researches require long-term statistical climate data, while practically; the available collected data in synoptic stations are quite short. Therefore, the required daily climate data should be generated based on the limited available data. ...
Read More
Introduction:Many existing results on water and agriculture researches require long-term statistical climate data, while practically; the available collected data in synoptic stations are quite short. Therefore, the required daily climate data should be generated based on the limited available data. For this purpose, weather generators can be used to enlarge the data length. Among the common weather generators, two models are more common: LARS-WG and ClimGen. Different studies have shown that these two models have different results in different regions and climates. Therefore, the output results of these two methods should be validated based on the climate and weather conditions of the study region.
Materials and Methods:The Sisab station is 35 KM away from Bojnord city in Northern Khorasan. This station was established in 1366 and afterwards, the meteorological data including precipitation data are regularly collected. Geographical coordination of this station is 37º 25׳ N and 57º 38׳ E, and the elevation is 1359 meter. The climate in this region is dry and cold under Emberge and semi-dry under Demarton Methods. In this research, LARG-WG model, version 5.5, and ClimGen model, version 4.4, were used to generate 500 data sample for precipitation and temperature time series. The performance of these two models, were evaluated using RMSE, MAE, and CD over the 30 years collected data and their corresponding generated data. Also, to compare the statistical similarity of the generated data with the collected data, t-student, F, and X2 tests were used. With these tests, the similarity of 16 statistical characteristics of the generated data and the collected data has been investigated in the level of confidence 95%.
Results and Discussion:This study showed that LARS-WG model can better generate precipitation data in terms of statistical error criteria. RMSE and MAE for the generated data by LAR-WG were less than ClimGen model while the CD value of LARS-WG was close to one. For the minimum and maximum temperature data there was no significant difference between the RMSE and CD values for the generated and collected data by these two methods, but the ClimGen was slightly more successful in generating temperature data. The X2 test results over seasonal distributions for length of dry and wet series showed that LARS-WG was more accurate than ClimGen.The comparison of LARS-WG and ClimGen models showed that LARS-WG model has a better performance in generating daily rainfall data in terms of frequency distribution. For monthly precipitation, generated data with ClimGen model were acceptable in level of confidence 95%, but even for monthly precipitation data, the LARS-WG model was more accurate. In terms of variance of daily and monthly precipitation data, both models had a poor performance.In terms of generating minimum and maximum daily and monthly temperature data, ClimGen model showed a better performance compared to the LARS-WG model. Again, both models showed a poor performance in terms of variance of daily and monthly temperature data, though LAR-WG was slightly better than ClimGen. For lengths of hot and frost spells, ClimGen was a better choice compared to LARS-WG.
Conclusion:In this research, the performances of LARS-WG and ClimGen models were compared in terms of their capability of generating daily and monthly precipitation and temperature data for Sisab Station in Northern Khorasan. The results showed that for this station, LARS-WG model can better simulate precipitation data while ClimGen is a better choice for simulating temperature data. This research also showed that both models were not very successful in the sense of variances of the generated data compared to the other statistical characteristics such as the mean values, though the variance for monthly data was more acceptable than daily data.
N. Khalili; K. Davary; A. Alizadeh; M. Kafi; H. Ansari
Abstract
Modeling of crop growth plays an important role in evaluation of drought impacts on rainfed yield, choosing an optimum sowing date, and managerial decision-makings. Aquacrop model is a new crop model that developed by Food and Agriculture Organization (FAO), that is a model for simulation of crop yield ...
Read More
Modeling of crop growth plays an important role in evaluation of drought impacts on rainfed yield, choosing an optimum sowing date, and managerial decision-makings. Aquacrop model is a new crop model that developed by Food and Agriculture Organization (FAO), that is a model for simulation of crop yield based on “yield response to water“ with meteorological, crop, soli and management practices data as inputs. This model has to be calibrated and validated for each crop species and each location. In this paper, the Aquacrop has been calibrated and evaluated for rainfed wheat in Sisab station (Northern Khorasan). For this purpose, daily meteorological data and historical yield data from two cropping season (2007-2008 and 2008-2009) in the Sisab station have been used to calibrate this model. Next, meteorological data and historical yield data of five cropping season (2002-2003 to 2006-2007) are used to validate the model. The result shows that the Aqucrop can accurately predict crop yield as R2, RMSE, NRMSE, ME, and D-Index are achieved 0.86, 0.062, 5.235, 0.917 and 0.877, respectively.
