Analysis of Trends and Patterns of Droughts Using Satellite Data and Climate Model Simulations

Abstract: Numerous studies indicate that the frequency of extreme events such as droughts have increased particularly in the 20th century. Droughts are considered among the most costly natural disasters due to their impacts on crop yield, infrastructure, industry, and tourism. This thesis investigates historical global trends and patterns of droughts and drying areas. A satellite-based (model independent) climate data record is used along with historical information from the latest CMIP5 multi-model simulations. For the CMIP5 simulations the objective include a comparison with validated ground observations. The focus is on meteorological droughts, defined as a lack of precipitation, and the Standardized Precipitation Index (SPI) is used to analyze fractions of area under drought (SPI $\leq$ -1) and area experiencing extreme droughts (SPI $\leq$ -2) at a quasi-global scale, for the length of the records available. From the 32-year long record, based on merged satellite product from GPCP and PERSIANN data, and from the CMIP5 simulations between 1901 -2005, the results show similar pattern for the Southern Hemisphere (SH) where variability is larger than in the Northern Hemisphere (NH). When investigating trends of land area in drought, only the SH show a significant trend using Mann-Kendall's test on the merged satellite based product. However, over oceans drought areas indicate an increase, for both hemispheres and the globe as a total. When studying the CMIP5 simulations and comparing them to ground observations 73\% of the models recapture the trends in the observations, for the global land area, as well as for the NH and the SH, separately. Both datasets indicate larger variability of area in drought for the SH than for the NH, for extreme as well as more moderate droughts. From the analysis of spatial patterns of wetting and drying trends maps were created from the pixel-based satellite data and several regions, such as the southwestern United States, Texas, parts of the Amazon, the Horn of Africa, and northern India exhibit a drying trend. The results from the CMIP5 models were generally consistent with observations at a global (semi-global) scale, but most models were not in agreement with observations with respect to regional drying and wetting trends. However, CMIP5 simulations of regional trends are collectively in better agreement with observations over high latitudes, as well as northeastern India and the western United States.