Assessment and forecasting spatial pattern changes of dust and wind speed using ARIMA and ANNs model in Helmand Basin, Iran

The aim of this study is to assess and forecast spatial pattern changes of dust and wind speed in Hamun-e Helmand basin, Sistan region. This region has the most dust events and the strongest winds, including 120-day winds. Wind speed and dust data of seven synoptic stations were therefore extracted from the Iran Meteorological Organization during 1990-2018. Wind and dust speed were predicted for 2020-2030 period using Autoregressive Integrated Moving Average (ARIMA) model and neural network, then analyzed by spatial pattern distribution using Hotspot G Index. The results showed that both models have a good efficiency in predicting wind and dust speed. However, due to error assessment in ARIMA model, neural network was more accurate in prediction. The results of spatial autocorrelation showed that cluster pattern of dust formed a pattern based on ARIMA model with an area of 21.01 and 20.64, neural network with an area of 19.67 and 19.47 for the two statistical periods 2018-2025 and 2026-2035, respectively, in the eastern half of the basin, namely Zabol and Zahak. The condition of autocorrelation of wind speed pattern was similar to that of the dust, except that the wind speed not only extended to the south of the basin, but also had spatial autocorrelation positive pattern in the northern half of the basin as small spots based on ARIMA model with an area of 18.71 and 15.16, neural network with an area of 22.12 and 15.68 for the two statistical periods 2018-2025 and 2026-2035, respectively.