Comparison of Artificial Neural Network and SDSM Methods in the Downscaling of Annual Rainfall in the HadCM3 Modelling (Case study: Kerman, Ravar and Rabor)

Abstract

Nowadays, it is believed that anthropogenic activities, such as changes in land use and deforestation, have resulted in atmospheric concentrations of greenhouse gases. One consequence of this ruinous activity, is an alteration of the energy balance that tends to warm the atmosphere that has resulted in climate change. Precipitation forecast is one of the most important element in water resources management and planning. In this study, precipitation depth of Kerman, Ravar and Rabor Stations have been predicted using the HadCM3 model outputs under the A2 scenario, SDSM downscaling models and artificial neural network, for three periods: 2010-2039, 2040-2069 and 2070-2099. Precipitation data for the 1971- 2001 period were selected as the base one. The results obtained by using the two models were evaluated and compared according to the statistical criteria. The artificial neural network model showed superior performance for the Kerman and Ravar stations. Annual precipitation of Kerman, Ravar and Rabor stations by 2099, using the AMM model decreases by 12.86, 11.68, and 11.39 percentage points, respectively. These are for 0.89, 18.48, and 1.55 percentage points, respectively, for the same year.

Keywords

• precipitation
• climate change
• downscaling hadcm3 model