Agricultural Drought Monitoring
Focusing on the effect of drought on agricultural or naturally vegetated lands, a monitoring system can be based on either hydrological quantities affected by water shortage and related to growth and yield (i.e., soil moisture, actual evapotranspiration) or on indicators of the biomass amount or ecosystem productivity (i.e., vegetation indices, leaf area index, fraction of Absorbed Photosynthetically Active Radiation [fAPAR]) (Mishra and Singh 2010).
Soil moisture (0) is seen as one of the most suitable variables to monitor and quantify the impact of water shortage on plants. More specifically, drought events are commonly detected by means of soil moisture anomalies (deviation from the climatology) computed as z-score values (e.g., Anderson et al. 2012) for a given aggregation period (e.g., dekad or month):
where 0ik is the soil moisture for the i-th aggregation period at the k-th year, and q and are the long-term average and standard deviation for the i-th aggregation period, respectively. The use of z-scores is suitable to detect soil moisture conditions that are drier than usual according to a past climatology, which can be considered a good indicator of the occurrence of agricultural drought.
Often, hydrological model outputs are used to spatially reconstruct the temporal dynamic of soil moisture over a certain region. In EDO, the root zone soil moisture outputs (in terms of soil water suction, pF) of the LISFLOOD model (de Roo et al. 2000) are used to obtain dekadal (three approximately 10-day periods per month) anomalies over Europe on a 5-km grid. Near-real time runs of LISFLOOD from the European Flood Awareness System (EFAS, Thielen et al. 2009) are used for that purpose. At global scale, we are testing to combine different sources of soil moisture, including outputs from LISFLOOD and thermal and passive/active microwave remote sensing data for a merged global product at 0.1 degree resolution (Cammalleri and Vogt 2017a).
Applications over large areas have highlighted how under some circumstances the simple anomalies can be insufficient to detect negative effects on the plant cover, mainly in areas characterized by high water content values (i.e., low or null water deficit). For this reason, Cammalleri et al. (2016a) have developed a soil moisture-based Drought Severity Index (DSI) that accounts for both the rarity of a soil moisture status (derived from the z-score) and the actual magnitude of the vegetation water deficit. This drought index is computed as a geometric mean of two indicators:
where the term d represents a soil moisture-derived water deficit index and p represents a dryness probability index, the latter being related to the probability that d is drier than the mode of the reference climatology. The use of the geometric mean allows having a high value of DSI only when both d and p are high, and hence the soil moisture status is both rare and stressing for plants.
The plots in Figure 18.1 exemplify how d is directly derived from 0 by means of the s-shaped water stress curve proposed by van Genuchten (1987) and p is computed after fitting a beta probability distribution function (pdf) to the climatological data (Gupta and Nadarajah 2004). It is worth noting that d is > 0 only when 0 is greater than the critical value for which water stress starts to occur (Seneviratne et al. 2010) and p is > 0 only if d is significantly higher than the mode of the pdf.
An alternative approach for agricultural drought monitoring is to directly observe the variation in vegetation growth or greenness to detect areas affected by drought events. In this context, remote sensing-derived vegetation indices are very useful tools for such analysis over large areas (e.g., Ghulam et al. 2007; Peters et al. 2002). Among the quantities that can be derived from space, the absorbed photosynthetically active radiation (fAPAR) has been widely identified as a suitable proxy of the greenness and health status of
Schematic representation of the procedure to compute d and p factors. On the left plot 8 is converted in d through a water stress curve, in the central panel d is compared against the climatology to evaluate if it is drier than the mode, and on the right panel d is converted into p by accounting for the pdf of the climatology. More details can be found in Cammalleri et al. (2016a).
vegetation, thanks to its central role in both plant primary productivity and carbon dioxide absorption (Gobron et al. 2005a). The observed sensitivity of fAPAR to vegetation stress has suggested its use in drought monitoring (Gobron et al. 2005b), particularly using anomalies. Both EDO and GDO systems use long records (starting in 2001) of fAPAR maps derived from the Terra satellite MODerate-resolution Imaging Spectroradiometer (MODIS) 8-day standard product (MOD15A2); these maps are quality checked to ensure the use of only high-quality data, filtered and averaged to dekadal time scale, and used as input of Equation 18.1 to derive z-score values. Further studies tried to combine the analysis of the vegetation phenological cycle and fAPAR anomalies to improve the accuracy of drought detection (Cammalleri et al. 2016b).
fAPAR anomalies can also be related to a variety of other stress factors (e.g., heat and pests); hence, further information on water stress needs to be used to associate recorded anomalies with drought. Following these considerations, Sepulcre-Canto et al. (2012) developed the CDI to account for the cascade process from a shortage in precipitation to yield reduction through a soil moisture deficit. The authors investigated the relationship between three types of indices: (1) the n-month accumulation standardized precipitation index (SPI-n), (2) the soil moisture anomalies in terms of soil suction (pF), and (3) the fAPAR anomalies.
Figure 18.2 highlights the conceptual framework that constitutes the CDI; a watch status is issued when a significant precipitation deficit is observed (e.g., SPI-3 or SPI-1 < -1), which is then converted into a warning and then into an alert when a significant soil moisture deficit and negative fAPAR anomalies are observed as well. Two recovery classes are also added to track down the return to normal conditions of rainfall and vegetation status, respectively.
An example of the CDI output from EDO is shown in Figure 18.3.
Schematic representation of the stages of the idealized agricultural drought cause-effect relationship that inform the concept of the CDI and the associated warning levels that are outputs of the CDI. (Adapted from Sepulcre-Canto et al., Natural Hazards and Earth System Sciences 12:3519-3531, 2012.)
The European Drought Observatory (EDO). Example of the combined drought indicator (CDI) for September 21-30, 2016.