Monitoring Drought Stress

High-Resolution SAR Imaging

For assessing drought conditions at a local scale, such as a municipal or watershed level, high-resolution imagery from visible-infrared scanners and SAR systems can be used. The preceding discussion illustrated the potential of SARs for mapping drought-relevant parameters including soil moisture or plant stress. The latter may be manifest in reduced canopy moisture and ultimately in plant growth.

The use of radar for directly inferring drought stress conditions of crops has been limited. In a ground-based scatterometer study, radar backscat-ter differed between stressed and nonstressed crops, but backscatter was dependent on crop type and crop density (Brisco and Brown, 1990). The separability of the stressed crops was attributed to their stunted growth and lower biomass. Steven et al. (2000) concluded that Radarsat SAR data can be used to detect canopy dehydration under conditions of moisture stress.

The greatest potential for SAR data for drought monitoring is through the estimation of soil moisture. Dobson and Ulaby (1998) provided an overview of common modeling approaches on how to derive soil moisture from SAR data. At the scale of SARs (tens of meters), spatial patterns of surface roughness and both vegetation type and density leave a strong imprint on SAR imagery. Therefore it is not possible to infer soil moisture patterns by simple visual analysis of SAR images. More sophisticated retrieval approaches are required that account for surface roughness and, where vegetation is present, for the effect of vegetation on the SAR signal. Unfortunately, as previously discussed, models often fail to accurately describe backscatter from bare soil surfaces due to the complex and multi-scale structure of agricultural and other natural soil surfaces and the limited validity range of the models. Nevertheless, it has been demonstrated that relatively simple change detection approaches, which depend on the availability of a number of consecutive SAR acquisitions, may be successful in tracking soil moisture conditions within a watershed.

A number of approaches use the single-channel SAR data provided by ERS-1/2 (Crevier et al., 1996; Rotunno Filho et al., 1996; Moran et al., 2000; Le Hegarat-Mascle et al., 2002). Bare-soil fields are ideal targets for tracking soil moisture through time, but in an agricultural environment a field does not remain bare throughout the season. The approach of Le-Hegarat-Mascle et al. (2002), for a particular watershed, uses a time series of images from one year to develop an empirical relationship between the mean radar backscatter and mean soil moisture from a number of target sites. In subsequent years, through inversion of the empirical relationship, soil moisture can be directly estimated from the imagery rather than by time-consuming ground-based measurements. The target sites are fields of bare soil or fields with very little vegetation, which may change from image to image in the time series. However, on any one date sufficient sites are selected such that the mean value is representative of the watershed. As an alternative to bare soil, Crevier et al. (1996) and Rotunno Filho et al. (1996) proposed temporal measurements of semipermanent grass fields to monitor soil moisture status of a watershed. The semipermanent grass fields, due to their relative stability in surface characteristics both within and across years, offer a means to normalize the surface roughness and vegetation amounts. This idea is being further examined by Sokol et al. (2002).

The technique of Moran et al. (2000), developed in a rangeland system using ERS-1 data, may be useful in tracking a potential drought. A single image acquired under dry conditions was used to normalize for surface roughness and standing brown litter content in all other images in a time series. The simple subtraction of the dry image backscatter values from the backscatter values in every other image in the time series increased the relationship with soil moisture (r2 = 0.93) compared to using the values extracted directly from each image (r2 = 0.27). The green leaf area index (GLAI), the one-sided green leaf area per unit ground, within the rangeland system was <0.35 and thus could be ignored in terms of influencing soil moisture determination using ERS-1. However, a method was presented to correct for GLAI using a vegetation component derived from optical data.

Further progress in the use of radar can be expected with the launch of technically more advanced multipolar, multifrequency radar satellites. Moran et al. (1998) found, in alfalfa and cotton, that up to a GLAI value of 4, high-frequency Ku (VV) radar was sensitive to increases in GLAI, whereas the lower frequency C band (VV) was sensitive to soil moisture. Above a GLAI value of 4, the Ku band saturated and the sensitivity of C band to soil moisture decreased due to attenuation of the signal by the vegetation. Similarly, Prevot et al. (1993) found that the simultaneous use of X band (VV) which is adapted to biomass estimation and C-band

(HH) radar which is adapted to soil moisture estimation, enabled both soil moisture and LAI of wheat canopies to be estimated. The availability of L band would further increase the potential to estimate soil moisture under denser canopies.

Large-Scale Monitoring Using Scatterometers

Compared to SAR, the number of scientific studies investigating the potential of scatterometers for soil moisture and vegetation retrieval is limited. Most likely this is due to the low spatial resolution of scatterometers (tens of kilometers), which restricts their use to regional applications. Still, considerable progress has been made, and scatterometer-derived soil moisture data, which reflect the atmosphere-related, large-scale component of the soil moisture field (Vinnikov et al., 1999), have already been used as input to crop models to assess drought-induced yield reductions. Here we only review work done with the C-band (VV) scatterometer onboard ERS-1/2 but recognize that there have been impressive first studies using the Ku-band scatterometers onboard the QuikScat and Midori-2 satellites launched in 1999 and 2002, respectively.

Many of the initial ERS scatterometer studies focused on the retrieval of vegetation parameters because a substantial agreement between back-scatter and global vegetation index maps has been observed (Frison and Mougin, 1996). Consequently, several models capable of separating soil moisture from vegetation effects have been developed and applied to back-scatter time series. Surface roughness is less of a problem for the analysis of scatterometer time series given that, at regional scales, it can be considered to be time invariant. In recent years attention has shifted more and more to the retrieval of soil moisture given the greater than anticipated sensitivity of the C-band scatterometer to soil moisture. For example, application of the model developed by Woodhouse and Hoekman (2000) over a Mediterranean region (Spain) did not properly recover the seasonal vegetation signal but provided soil surface reflectivity values in agreement with the monthly precipitation records.

A soil moisture retrieval technique based on a change detection approach has been developed by Wagner et al. (1999b). The method is capable of separating the effects of soil moisture and vegetation phenology by exploiting the information content provided by the multiple-viewing capabilities of the ERS scatterometer. By comparing instantaneous ERS scatterometer measurements to the lowest and highest backscatter values in the ERS scatterometer time series, a relative measure of the moisture content of the surface soil layer (<5 cm) is obtained. The algorithm has been tested over different climatic regions with success, and the multiyear soil moisture data were derived from remotely sensed data (Wagner and Scipal, 2000; Scipal et al., 2002). The data set is available to other research groups on request and can be viewed on a Web site (IPF, 2003).

Microwaves sense only the first few centimeters of the soil, but for agri cultural applications mainly the soil moisture content within the reach of the plant roots is of interest (Jones et al., 2000). To estimate the water content at deeper levels using scatterometer data, Wagner et al. (1999b) proposed a two-layer water balance model that transforms the highly variable surface soil moisture series into a red-noise-like profile of soil moisture time series. When soil hydrologic properties (wilting level, field capacity, and total water capacity) are known, the water content available to plants can be estimated. Over the Ukraine, a comparison with gravimetric soil moisture data from the agro-meteorological station network showed that the soil moisture content in the 0-100-cm layer can be estimated with an accuracy of about 5% volumetric soil moisture. A comparison with in situ soil moisture data from a network of 20 TDR (time domain reflectory) probes in the Duero-Basin in Spain yielded an accuracy of better than 3% volumetric soil moisture for the same layer (Scipal et al., 2003).

Timely information about the regional soil moisture conditions from remote-sensing data is useful for assisting agro-meteorological analysis. The information content can be further enhanced by comparing the present year with the long-term mean and modeling the timing of crop water stress (Barron et al., 2003). Given the availability of nine years of ERS-1/2 scatterometer data, soil moisture anomalies can be easily calculated. Figure 8.2 shows how scatterometer-derived soil moisture anomalies for the months February and March 1999 over Southeast Asia (including India) compare to rainfall anomalies derived from a globally gridded precipitation data produced by the Global Precipitation Climatology Center (GPCC, 2003). In both data sets, anomalous dry conditions centered over southern China can be recognized.

To assess the timing of water availability during critical crop-growth stages, remotely sensed soil moisture data must be combined with crop-growth models. In a study conducted over Russia and the Ukraine, scatter-ometer-derived soil moisture data were used as input to a crop growth model WOFOST (Supit et al., 1994). This model simulates the daily growth of a specific crop using weather and soil data and following the hierarchical distinction between potential and limited production. The ratio between potential and limited production is the indicator of drought stress. The scatterometer-derived soil moisture is used to replace the soil moisture estimates normally derived using water budget models. The results of the study suggested that provincial yield assessment can be improved through the use of remotely sensed soil moisture information (Wagner et al., 2000).

Was this article helpful?

0 0
Swine Influenza

Swine Influenza

SWINE INFLUENZA frightening you? CONCERNED about the health implications? Coughs and Sneezes Spread Diseases! Stop The Swine Flu from Spreading. Follow the advice to keep your family and friends safe from this virus and not become another victim. These simple cost free guidelines will help you to protect yourself from the swine flu.

Get My Free Ebook

Post a comment