Remote estimation of soil moisture availability and fractional vegetation cover for agricultural fields

Carlson, T.N., Perry, E.M. and Schmugge, T.J., ! 990. Remote estimation of soil moisture availability and fractional vegetation cover for agricultural fields. Agric. For. Meteorol., 52: 45-69. The purpose of this paper is to present a method for using remote measurements to estimate vegetation fraction, surface energy fluxes and the root zone and soil surface water contents for partial vegetation canopies. The primary tools are a boundary layer model with vegetation and substrate components and two image products: the variation of surface radiometric temperature vs. normalized difference vegetation index (NDVI), and the standard deviation of radiometric surface temperature vs. radiometric surface temperature. The method is based on determining: ( 1 ) asymptotic values of two radiometric surface temperatures for sunlit bare soil and for dense sunlit vegetation; and (2) a relationship between NDVI and surface temperature, which we call the axis of variation. The method is illustrated using aircraft and surface measurements made at Lubbon during the French HAPEX field experiment (1986), I N T R O D U C T I O N Models for estimating the surface turbulent energy fluxes and the soil moisture generally depend on a sensitivity of the surface radiometric temperature to soil water content. Over bare soil, variations in radiometric surface temperature tend to be highly correlated with variations in surface water content (Jackson et al., 1977; Schmugge, 1978; Jackson, 1982 ). Various models have been constructed to exploit the relationship between surface temperature and soil moisture. When used in conjunction with remote measurements of surface temperature, such as determined from a satellite, these models yield estimates of the surface moisture availability, the surface energy fluxes and the thermal inertia (Carlson and Boland, 1978; Carlson et al., 1981; Price, 1982; 0168-1923/90/$03.50 © 1990 Elsevier Science Publishers B.V. 46 T.N. CARLSON ET AL. Hatfield et al., 1984; Raffy and Becker, 1985; Taconet et al., 1986; Wetzel and Woodward, 1987; Lagouarde and Choisnel, 1990). Some of these models also take into account a layer of vegetation (Taconet et al., 1986; Wetzel and Woodward, 1987; Lagouarde and Choisnel, 1990). The relationship between soil moisture and surface temperature is vastly more complex over vegetation than over bare soil. Over vegetation there is a considerable amount of temperature variability owing to the structure of the vegetation canopy and particularly to the amount of bare soil viewed by the radiometer and exposed to the direct solar beam. Leaves tend to be cooler than the exposed bare soil because the intercellular airspaces are nearly saturated with water vapor, which is drawn from a relatively deep soil layer, the root zone. Bare soil temperatures reflect the soil moisture only over the top one or two centimeters (Idso et al., 1975 ). Emitted surface radiance constitutes a blend of radiances emitted from either shaded or unshaded bare soil and vegetation. Thus, radiometric surface temperature variations over vegetated surfaces may be the result of variations in the amount of bare soil visible to the radiometer. The situation is complicated further by the problems introduced by canopy architecture, which includes the variation in solar elevation angle and viewing angle of the radiometer. A significant step in the direction of modeling sparse or patchy vegetation cover was taken by Shuttleworth and Wallace (1985 ) who adapted the Penman-Monte i th equation (Monteith, 1975 ) to account for energy partitioning between crop and soil. In order to circumvent the inconsistency of using a one-dimensional model to represent horizontally inhomogeneous surfaces, Shuttleworth and Wallace introduced the idea of two asymptotic temperature limits, one for bare soil and the other for vegetation. Their model was expressed as separate bare soil and vegetation components with identical ambient conditions above the crop. Latent heat fluxes from each of the components were combined using weighting factors, which were expressed in terms of combinations of soil and plant resistances determined from a knowledge of crop height and leaf area index (LAI). Shuttleworth and Wallace recognized that the intractable complexity of a vegetation canopy required that they make simplifications that allow them to avoid confronting the aspects of the three-dimensional canopy. While not dismissing the effects of vegetation structure on the interception of solar radiation, they proposed that much of the fine-scale detail of the three-dimensional canopy would tend to cancel over time, leaving the first-order effects to be represented by a one-dimensional model. Stated differently, the fluxes over bare soil are relatively independent of the adjacent vegetation (and vice versa), but are linked via common substrate and atmospheric properties. In such a geometry, LAI has two values, one for the ensemble of vegetation clumps and one for the ensemble of vegetation REMOTE ESTIMATION FOR AGRICULTURAL FIELDS 47 clumps and contiguous bare patches. The one-dimensional model operates separately in each regime; bare soil and vegetation. Total fluxes for the combined bare soil and vegetation are related through an additional parameter, the fractional vegetation cover. We propose a similar approach to that of Shuttleworth and Wallace, in which a boundary layer model is used in conjunction with remote measurements of surface temperature and normalized difference vegetation index (NDVI) to calculate fractional vegetation cover, LAI, substrate water content in two layers, and surface energy fluxes. The method is illustrated using remote measurements made during the French HAPEX/MOBILHY field experiment. THE BOUNDARY LAYER MODEL Like Shuttleworth and Wallace, we will ignore the intractable aspects of the three-dimensional canopy, and imagine a simple structure in which clumps of vegetation (which may consist of a single plant ) are interspersed with bare soil patches (Fig. 1 ). Within the vegetation clumps, no direct solar radiation reaches the ground and no bare soil is visible to the radiometer. Uniformly distributed sparse vegetation and small dense clumps of vegetation are separated, perhaps randomly, by bare soil. The fractional part of the surface covered by the vegetation clumps is fv and that covered by bare soil is ( 1 fv) . No distinction is made between sunlit and shaded vegetation or between sunlit and shaded soil visible to the radiometer, although shaded bare soil may be much cooler than unshaded bare soil. Vegetation fraction also depends on the viewing angle of the radiometer; it will effectively increase with decreas-