High-resolution gridded climate dataset for data-scarce region

Abstract

The knowledge of spatiotemporal distribution of climate variables is essential for most of hydro-climatic studies. However, scarcity or sparsity of long-term observations is one of the major obstacles for such studies. The main objective of this study is to develop a methodological framework for the generation of high-resolution gridded historical and future climate projection data for a data-scarce region. Egypt and its densely populated central north region (CNE) were considered as the study area. First, several existing gridded datasets were evaluated in reproducing the historical climate. The performances of five high-resolution satellite-based daily precipitation products were evaluated against gauges records using continuous and categorical metrics and selected intensity categories. In addition, two intelligent algorithms, symmetrical uncertainty (SU) and random forest (RF) are proposed for the evaluation of gridded monthly climate datasets. Second, a new framework is proposed to develop high-resolution daily maximum and minimum temperatures (Tmx and Tmn) datasets by using the robust kernel density distribution mapping method to correct the bias in interpolated observation estimates and WorldClim v.2 temperature climatology to adjust the spatial variability in temperature. Third, a new framework is proposed for the selection of Global Climate Models (GCMs) based on their ability to reproduce the spatial pattern for different climate variables. The Kling-Gupta efficiency (KGE) was used to assess GCMs in simulating the annual spatial patterns of Tmx, Tmn, and rainfall. The mean and standard deviation of KGEs were incorporated in a multi-criteria decision-making approach known as a global performance indicator for the ranking of GCMs. Fourth, several bias-correction methods were evaluated to identify the most suitable method for downscaling of the selected GCM simulations for the projection of high-resolution gridded climate data. The results revealed relatively better performance of GSMaP compared to other satellite-based rainfall products. The SU and RF were found as efficient methods for evaluating gridded monthly climate datasets and avoid the contradictory results often obtained by conventional statistics. Application of SU and RF revealed that GPCC rainfall and UDel temperature datasets as the best products for Egypt. The validation of the 0.05°×0.05° CNE datasets showed remarkable improvement in replicating the spatiotemporal variability in observed temperature. The new approached proposed for the selection of GCMs revealed that MRI-CGCM3 gives the best performance and followed by FGOALS-g2, GFDL-ESM2G, GFDL-CM3 and lastly MPI-ESM-MR over Egypt. The selected GCMs projected an increase in Tmx and Tmn in the range of 2.42 to 4.20°C and 2.34 to 4.43°C respectively for different scenarios by the end of the century. Winter temperature is projected to increase higher than summer temperature. For rainfall, a 62% reduction over the northern coastline is projected where rain is currently most abundant with an increase of rainfall over the dry southern zones. Linear and variance scaling methods were found suitable for developing bias-free high-resolution projections of rainfall and temperatures, respectively. As for the CNE, the high-resolution projections showed a rise in maximum (1.80 to 3.48°C) and minimum (1.88 to 3.49°C) temperature and change in rainfall depth (-96.04 to 36.51%) by the end of the century, which could have severe implications for this highly populated region.