Determining the optimal geostatistical method for drought monitoring based on different drought indices and its application in Neishabour Plain

Almost all regions of the world suffer from drought in one way or another. Various indices have been developed to better evaluate and understand the drought. Here, we investigate the spread of drought in the Neishabour plain using the data collected during the hydrological years of 1990 to 2018. Our dataset includes rainfall and minimum/maximum temperature collected by 13 climatological stations and 3 synoptic stations in the region.

The drought was analyzed based on 6 indices: SPI, RDI, SPEI, aSPI, eRDI, and SPEIEP. For the first 3 indices, the total amount of precipitation was used, while for the rest, the amount of effective precipitation based on the USBR method was used. To better understand how drought spreads in the plain we needed to convert point data to regional data, to this end we investigated simple, ordinary, and general kriging of geostatistical methods. Our results showed that the two-by-two correlation coefficient between indices varied between 0.65 and 0.97. In all stations, the lowest R2 value belonged to the comparison between SPI and SPEIEP. In general, there was a high correlation in comparing the indices that used the amount of effective precipitation with corresponding indices that used the amount of total precipitation. In specific, the comparison between eRDI-vs-RDI and aSPI-vs-SPI showed R2 values higher than 0.98 and SPEI-vs-SPEIEP showed R2 values higher than 0.96. In addition, SPI, aSPI, RDI, and eRDI indices were highly correlated with each other.

This similarity of behavior exists in determining the most appropriate method of drought zoning in the studied area. To better dissociate between different methods and determine the optimum one, we looked at root mean square error (RMSE) values. Based on RMSE values the optimum method of zoning severe and extreme drought events was the ordinary kriging method. However, in the case of SPEI and SPEIEP, the most suitable zoning method was simple kriging. Furthermore, indices that use evaporation and transpiration better showed the extent of the drought area in recent years compared to the SPI index that only considers rainfall. This could be due to the effect of global warming on increase in evaporation and transpiration. We may therefore conclude that effective precipitation, precipitation, and evapotranspiration had a positive spatial correlation with each other in the study area because the findings of Global Moran's spatial autocorrelation index indicated cluster patterns for all 6 drought indices in the area. In most of the stations, a prominent cluster pattern was observed in the years with severe and very severe droughts, according to the local Moran's index results.

Overall, our results suggest the usage of the SPEIEP index to monitor the drought in the Neishabour plain, considering the increasing temperature and also the smaller error of the SPEIEP index in determining the expansion of drought in the plain. In addition, considering that the SPEIEP index is relatively new and requires more studies in this basin and other basins, it is suggested that evaporation and transpiration be estimated with other methods such as Blaney-Kreider, Torrent-White, and Penman-Monteith and also effective precipitation with the other methods. Influence of the effective precipitation and evapotranspiration estimation method on SPEIEP index values should be investigated. Additionally, it is recommended that more research be done on the spatial autocorrelation of drought in various geographic and climatic conditions across the country (Iran). If at all possible, it should also be looked into how the severity of the drought or the geographical characteristics of the regions affect the degree of spatial autocorrelation of drought.