Evapotranspiration (ET)

In addition to green biomass amount, as sampled by standard VIs, another indicator of vegetation health is the rate at which plants consume and transpire water. As available soil moisture in the root zone depletes toward the permanent wilting level, plants reduce their transpiration rates to conserve remaining water. This reduces evaporative cooling of the leaf surfaces, resulting in a detectable thermal signal of elevated canopy temperature that can be measured from space using thermal infrared sensors (Moran 2003). In such cases, the exposed soil surface is typically dry and also elevated in temperature—further enhancing the composite thermal signature of drought (Anderson et al., 2008, 2012; Kustas and Anderson, 2009). Land surface temperature (LST) is, therefore, a valuable remote sensing diagnostic of drought conditions and their impacts on vegetation health. This fact was exploited in the empirical VHI described in Section 10.5, which is a linear combination of scaled VI and LST anomalies contributing with opposite sign, with negative VI and positive LST departures interpreted as a signal of drought stress (Kogan 1995). Karnieli et al. (2010) demonstrated, however, that under energy-limited circumstances (e.g., at higher latitudes and elevations), positive LST departures can be a sign of beneficial plant growth conditions and therefore a more physically based interpretation of LST anomalies may be useful for more definitively attributing causal factors to stress.

One approach has been to use remotely sensed LST to diagnose evaporative fluxes using a physical model of the land surface energy balance (see, e.g., review by Kalma et al. 2008). Such models estimate the evaporative cooling required to keep the land surface at the observed temperature given the radiative load (solar plus atmospheric radiation) prescribed over the modeling domain. The derived evapotranspiration (ET) includes water extracted from the soil profile and transpired by the vegetation as well as water evaporated directly from the soil and other surfaces, and is therefore a valuable metric of both soil moisture status and vegetation health.

One example of an ET-based drought indicator is the ESI, describing standardized anomalies in the ratio of actual-to-potential ET (fPET = ET/PET) computed with the Atmosphere-Land Exchange Inverse (ALEXI) energy balance model (Anderson et al. 2007, 2012). ALEXI uses time changes in LST, obtained from geostationary satellites or day-night polar orbiter overpasses, to estimate time-integrated fluxes of daytime sensible and latent heating. Anderson et al. (2012) showed that ESI in general agrees well with spatiotemporal patterns in standard precipitation-based drought indicators and with the USDM, but can be generated at significantly higher spatial resolution (meter to kilometer scale) by using the LST proxy for rainfall information. Recent work reveals that Landsat-scale ESI can be effectively used to separate moisture response from different land cover types (e.g., crops, forest patches, and surface water bodies), thereby better capturing agricultural drought impacts over heterogeneous surfaces. Otkin et al. (2016) showed the LST inputs to ESI convey effective early warning of rapid stress development during flash drought events, with the crops showing an elevated thermal signal several weeks before significant changes in VI can be detected from space.

Figure 10.2 shows the time evolution of ESI, developed at 4-km resolution using GOES-E and W TIR imagery, during the flash drought that affected the US Corn Belt in 2012. In addition, maps of temporal changes in the ESI (AESI) convey valuable information about the rate at which vegetation and soil moisture conditions are deteriorating and recovering during periods of drought intensification and abatement (Anderson et al. 2012; Otkin et al. 2013). Early signals of significant drought intensification were visible in ESI and AESI in late May/early June—well before the USDM recorded extreme drought in the region starting in mid-July. Real-time ESI products are generated daily on an 8-km grid over most of North America as part of NOAA's GOES ET and Drought (GET-D) Information system (http://www.ospo.noaa.gov/Products/land/getd/). Efforts are underway to transition a prototype 5-km global ESI product, generated using MODIS or VIIRS day-night LST differences, to operational status.

The Famine Early Warning Systems Network (FEWSNET) also produces an ET anomaly product, generated with the Simplified Surface Energy Balance Operational (SSEB-op) modeling system, which is a simplified LST-based ET retrieval methodology designed specifically for operational applications (Senay et al. 2013). Primary remote sensing inputs include time-composited MODIS LST (8-day) and NDVI (16-day) products, as well as topographic information from the Shuttle Radar Topographic Mission (SRTM). SSEB-op ET anomaly products are generated routinely in several climate-sensitive global regions to support early detection of regional crop failure and threats to food security. Domestically, SSEB-op ET products are used within the USGS Water for Sustaining and Managing America's Resources for Tomorrow (WaterSMART) program for accounting of regional water use and availability.


Monthly evolution of ESI (1-month composite), DESI (1-month difference), and USDM drought classifications during the 2012 US flash drought event. ESI detected signals of intensifying stress in the central United States in late May. DESI highlights areas where drought intensification (red) or recovery (green) is most significant.

ET diagnosed using remotely sensed LST via energy balance provides complementary yet independent information in comparison with estimates derived using prognostic land surface or hydrological models using water balance constraints, such as the NLDAS suite of soil moisture and ET indices (Xia et al. 2012). Hain et al. (2015) demonstrated that LST-based ET can inherently reflect ancillary sources of moisture besides local rainfall, including water applied through irrigation or extracted phreatically by vegetation growing over shallow water tables. These moisture sources may help mitigate drought impacts locally, but are difficult to capture in water balance models without extensive a priori knowledge of the complete hydrologic system, including human manipulations. Senay et al. (2012) report similar advantages in a comparison of the SSEB framework with a related water balance approach to ET mapping (forced by precipitation data); however, they note the latter may be more useful in some hydrologic applications that require detailed information on the temporal variability in soil moisture and runoff.

Remotely sensed actual ET is also a complement to drought indices based on potential ET such as the evaporative demand drought index (EDDI) (Hobbins 2016; McEvoy 2016), which describes the desiccating power of the local atmospheric conditions. High evaporative demand can be an effective early indicator of rapid drought onset, although it does not always result in actual drought impacts materializing on the ground, for example, because of amelioration by ancillary moisture sources. Taken together, there is much to be learned about ecosystem resilience and susceptibility to drought by comparing actual ET anomalies derived through the energy and water balance with anomalies in potential evapotranspiration (PET) deduced from meteorological conditions. A multi-index early warning system following signal progression from EDDI to ESI could be used to track if/how atmospheric precursors of drought evolve into vegetative stress that can negatively impact crop yields or rangeland condition. See further discussion of this topic in Chapter 11.

< Prev   CONTENTS   Source   Next >