With challenge comes opportunity. The focus of the scientific community should be on merging the information provided from different satellites and sensors, to underpin their uncertainties, and to offer long‐term and consistent data sets for drought analysis.
1.1. INTRODUCTION
Drought is a recurring natural feature of climate and is defined as below‐normal precipitation, usually over an extended period of time (Wilhite & Buchanan‐Smith, 2005). Droughts cause billions of dollars of damage to multiple sectors globally, specifically to agriculture. Droughts may also cause, or co‐occur with, other hazards such as heatwaves, which collectively escalate the ramifications of this natural hazard (Raei et al. 2018). Indeed, the concurrence of climatic extremes, in particular droughts and heat waves, can result in forest fires (Goulden, 2018; Silva et al., 2018; Taufik et al., 2017), land degradation and desertification (Hutchinson & Herrmann, 2016; Olagunju, 2015; Vicente‐Serrano et al., 2015), water shortage for agriculture and urban water supply (AghaKouchak, Farahmand, et al., 2015; Gober et al., 2016; Khorshidi et al., 2019; Van Loon et al., 2016), and economic impacts, and may prompt water bankruptcy (Howitt et al., 2014; Madani et al., 2016). Therefore, the impacts of drought are complex and can propagate to regions outside the area of its occurrence. Drought is often categorized in four groups: meteorological, agricultural, hydrological, and socioeconomic (Dracup et al., 1980). Meteorological drought is defined as precipitation deficiency over a long period, and it best represents the onset of drought (Utah Division of Water Resources, 2007). An extended period of meteorological drought results in soil moisture deficit as evapotranspiration continues despite the lack of precipitation, which leads to agricultural drought (Cunha et al., 2015). Persistence of metrological drought ultimately reduces overall water supply and drought is manifested in a hydrological form (Modaresi Rad et al., 2016). Socioeconomic drought then occurs as supply and demand of some economic goods are impacted by meteorological, agricultural, and hydrological droughts (Shiferaw et al., 2014). The observed changes in temporal patterns of precipitation associated with unsustainable water withdrawal may escalate the drought severity around the globe (Mallakpour et al., 2018; U.S. Global Change Research Program, 2018); and large‐scale changes in weather patterns are likely to affect water storage around the globe and threaten water supply particularly in arid and semi‐arid regions (Ault et al., 2014).
Drought detection requires observation of a plethora of different climatic and biophysical variables. Observations in situ, however, do not provide a uniform spatial distribution and are limited to populated areas, hence satellite‐based observations provide a unique way to analyze and monitor drought at a global scale. Satellites offer observations for a wide range of climate variables such as precipitation, soil moisture, temperature, relative humidity, evapotranspiration, vegetation greenness, land‐cover condition, and water storage (Aghakouchak, Farahmand, et al., 2015; R. G. Allen et al., 2007; L. Wang & Qu, 2009; Whitcraft et al., 2015). Although remote sensing provides more opportunities for the scientific community to monitor Earth systems and offer better understanding of drought impact at regional to global scales, it is not without flaws or challenges. The main challenge is the insufficient length of the observed records provided for the variables of interest. Other challenges include data consistency, ease of access, quantifying uncertainty, and development of appropriate drought indices, which will be discussed throughout this chapter.
1.2. PROGRESS IN REMOTE SENSING OF DRIVERS OF DROUGHT
This section presents the recent remote sensing techniques used for identification and quantification of drought as characterized by different climatic and biophysical variables.
1.2.1. Precipitation
A meteorological drought can be described as precipitation deficiency over a period of time (WMO, 1975), often represented in terms of an index of deviation from normal. Drought indices not only serve the scientific communities but they are also great tools for facilitating the decision‐making and policy‐making processes for stakeholders and managers when compared with the raw data. One of the most widely used and informative meteorological drought indices is the standardized precipitation index (SPI) developed by Mckee et al. (1993). Several other meteorological drought indices have also been proposed, including, but not limited to, precipitation effectiveness (Thornthwaite, 1931), antecedent precipitation (API; McQuigg, 1954), rainfall anomaly (RAI; Van Rooy, 1965), drought area (Bhalme & Mooley, 1980), effective precipitation (Byun & Wilhite, 1999), and rainfall variability indices (Oguntunde et al., 2011). The SPI is currently being used in many national operational and research centers and was recognized as a global measure to characterize meteorological drought by the World Meteorological Organization (WMO, 2009). Computation of SPI requires measured rainfall data and a normalization process of monthly data, either by utilizing an appropriate probability distribution function (PDF) to transform the rainfall PDF (e.g., gamma or Pearson type III probability distribution) into a standard normal distribution (Khalili et al., 2011), or by utilizing a nonparametric approach (Hao & AghaKouchak, 2014). Precipitation deficit can be specified for different timescales (e.g., from 1 to 24 months) when using SPI, where precipitation abnormalities in shorter timescales reflect soil moisture wet/dry conditions and longer timescales portray the wet/dry conditions of subsequent processes such as streamflow, reservoir levels, and ultimately groundwater.
Since the root cause of droughts is deficit in precipitation, meteorological drought indices, and in particular SPI, are suitable indices for revealing the onset of drought (Hao & Aghakouchak, 2013). Indeed, precipitation is regarded as a key component in drought analysis. Clustering approaches have been used as a common practice to identify spatially homogeneous drought areas by utilizing meteorological drought indices such as SPI (Santos et al., 2010). Assessment of temporal variability of metrological drought utilizing SPI, however, has shown formation of noncoherent clusters in spatiotemporal clustering (Modaresi Rad & Khalili, 2015). This is due to precipitation’s large spatial variability, which creates diverse spatial patterns even at small scales. Considering spatial variability of precipitation is crucial, since a dense and evenly distributed network of gauging stations is required for describing spatiotemporal characteristics of drought. Similarly, ground‐based weather radars also suffer from spatial discontinuity and are error prone due to contamination by surface backscatter, uncertainty of approximation of relation between reflectivity and rain rate, and bright band effects, making them unfeasible for global applications (Kidd et al., 2012; Wolff & Fisher, 2008). As a result, a more robust approach would be to use satellite observations that would produce gridded data as an input not only for drought models, but also for meteorological and hydrological models such as weather research and forecasting (WRF) and variable infiltration capacity (VIC).
Figure 1.1 Rainfall map by NASA’s Tropical Rainfall Measuring Mission (TRMM) satellite. (a) Average rate of rainfall per day for the period of 1998‐2011. (b) A tropical storm in southeast Texas causing record‐breaking floods, produced using the IMERG precipitation product.
(Courtesy: NASA’s Earth observatory: https://earthobservatory.nasa.gov/images)
Visible (VIS) satellite images provide information about cloud thickness and infrared (IR) images provide information on cloud top temperature and cloud height that are used to estimate precipitation rate via different retrieval algorithms (Joyce & Arkin, 1997; Sapiano & Arkin, 2009; Turk et al., 1999). Geostationary (GEO) VIS/IR satellites offer approximately a 15–30 min frequency of observations, but their accuracies are disputed. On the other hand, passive microwave (MW) sensors capture data of hydrometeor signals and scattering signals of raindrops, snow, and ice contents in the lower atmosphere and sense the bulk emission from liquid water, and therefore provide a more accurate estimation of precipitation rate (Behrangi et al., 2014). The MW sensors, however, often face difficulties distinguishing between light rain and