1.5. REMAINING CHALLENGES AND OPPORTUNITIES
The number of different satellite sensors observing our planet is ever increasing and as a result, better resolutions and different physical variables can be obtained. Some of these satellite missions appropriate for monitoring drought related variables that recently have been launched include ECOSTRESS, GPM, GOES R series, SMAP, Ice Cloud and Land Elevation Satellite 2 (ICESat‐2), and GRACE Follow‐On, and some are planned for launch in the near future such as Surface Water and Ocean Topography (SWOT), Landsat 9, Biomass, and FLuorescence EXplorer (FLEX). Initial challenges associated with the launch of new satellites, however, include inconsistencies in observations due to sensor changes, continuity of data, unforeseen uncertainties, data maintenance, and community acceptability. Among them, data continuity represents a great challenge both in terms of cost and time. Since satellite missions are often expected to have a life span of near a decade and an equal amount of time is required for the design of a new satellite, preparations must be made to ensure continuity of data through follow‐up missions. This is particularly essential for drought monitoring purposes, as over 30 years of data are required for drought monitoring and this often surpasses the ideal operational lifetime (e.g., a decade) of most satellites. Some of these follow‐up missions include the series of Landsat missions, ESA’s Sentinels, NASA’s Visible Infrared Imaging Radiometer Suite (VIIRS) and GPM, GRACE Follow‐On, and GOES‐R. The process of reconstructing the time series introduces another source of uncertainty to drought modeling. To provide an estimate of uncertainty associated with remote sensing products, a number of different statistical techniques can be used, including data assimilation (Massari et al., 2015), triple collocation analysis (TCA; W. B. Anderson et al., 2012), generalized triple collocation analysis (GCA; Dong & Crow, 2017), and spectrum analysis (Kumar et al., 2018). W. B. Anderson et al. (2012) used TCA to estimate the total observation error variance of the combined three different soil moisture products: thermal remote sensing by atmosphere‐land exchange inverse (ALEXI), microwave AMSR‐E, and simulations from physically based models. The TCA was validated for the 2010–2011 Horn of Africa drought and showed promising results.
When it comes to data processing and analysis of satellite imageries, different algorithms can help in distinguishing pixels and identifying objects, such as deep learning methods. There are some atmospheric features, however, that act as a barrier for certain optical and infrared satellite instruments and result in data inconsistencies. Optical‐based vegetation indicators are error prone when the area studied has atmospheric effects, cloud cover, aerosols, and water vapor (Andela et al., 2013). Moreover, optical satellite observation only reflects information from the top of the canopy. These problems can be resolved using microwave sensors that provide the opportunity to monitor carbon cycling during drought episodes over the long term. A unique approach would be to combine the vegetation optical depth (VOD; Owe et al., 2001) with optical based methods (i.e., NDVI) for a complementary analysis that considers both canopy top greenness and biomass. Combination of microwave, optical, and lidar observations provides an opportunity to monitor ecosystem response to drought that often continues even after drought recovery (C. D. Allen et al., 2010). Recent studies indicate that some variables such as snow and relative humidity can be integrated into drought monitoring models for improving estimations of drought recovery and detection of its onset, respectively (AghaKouchak et al., 2014; Rott et al., 2010).
Another challenging issue with remote sensing observations is the process of preserving large historical records, as it requires large and costly infrastructure and help of professional to store these data. Climatic data records can be merged together to produce longer records that would be appropriate for assessment of drought and monitoring (AghaKouchak & Nakhjiri, 2012). For example, several attempts have been made to generate NDVI from observations of multiple satellite missions including AVHRR and MODIS (Beck et al., 2011; Pinzon & Tucker, 2014; Tucker et al., 2005).
A change in satellite sensors, such as a follow‐up mission, is introduction of a great deal of uncertainty in modeling drought, and these uncertainties are often unquantified (Mehran et al., 2014). Therefore, an ideal way to tackle the problem is to provide uncertainty bounds along with raw observations. This uncertainty and the structural and parameter uncertainty resulting from model‐based simulations can be merged together to help decision making in operational applications (Sadegh, Ragno, et al., 2017). Such models and indicators are now being used more frequently and they quantify the uncertainty associated with satellite observations (AghaKouchak & Mehran, 2013; Entekhabi et al., 2010; Gebremichael, 2010). Therefore, the more remote sensing data are tailored for drought assessment, the more decision makers and drought experts can be engaged with remote sensing data.
1.6. CONCLUSION
Remote sensing offers a new way to monitor drought and develop drought models that would consider multiple variables at a global scale. Limitations of measurements in situ, such as nonuniformity and lack of measurement, are now resolved by multisensor remote sensing frameworks. Moreover, the introduction of multi‐index and composite drought monitoring has enhanced drought detection capabilities. This further extends drought analysis by allowing scientists to investigate the extent of effects of drought on other natural processes after the period of drought recovery. This chapter highlights different variables that contribute to formation of drought and discusses numerous satellite products that can offer valuable data as input for different categories of drought models. The critical role that precipitation as snow plays in the occurrence of drought is discussed, including that its consideration into drought modeling has been hindered due to the lag between its occurrence and changes in surface water. Recently, several composite drought‐monitoring frameworks have been proposed to address this issue and have shown promising results. It is argued that recent shifts in rainfall patterns and increases in temperature happen in parallel with increasing severity and frequency of concurrent heatwaves and droughts. Anthropogenic hydrological change has altered the recurrence of extremes and this hinders monitoring systems, which are developed based on stationary assumptions of natural variables. Therefore, more uncertainty is introduced into modeling drought‐monitoring systems if the stationarity assumption is to be retained. In addition, there is a growing need to assimilate satellite‐retrieved information into land surface, hydrological, and climate models. In order to do so, the uncertainty associated with satellite observations should be quantified in such a way that it can be used in modeling. Moreover, large database records should be designed to be accessible to both the scientific and the operational communities.
REFERENCES
1 Adler, R.F., Huffman, G.J., Chang, A., Ferraro, R., Xie, P.P., Janowiak, J., et al. (2003). The version‐2 global precipitation climatology project (GPCP) monthly precipitation analysis (1979–present). Journal of Hydrometeorology, 4(6), 1147–1167.
2 AghaKouchak, A., Cheng, L., Mazdiyasni, O., & Farahmand, A. (2014). Global warming and changes in risk of concurrent climate extremes: Insights from the 2014 California drought. Geophysical Research Letters, 41(24), 8847–8852. https://doi.org/10.1002/2014GL062308
3 Aghakouchak, A., Farahmand, A., Melton, F.S., Teixeira, J., Anderson, M.C., Wardlow, B.D., & Hain, C.R. (2015). Remote sensing of drought: Progress, challenges, and opportunities.