Skip to main content
ARS Home » Pacific West Area » Boise, Idaho » Northwest Watershed Research Center » Research » Publications at this Location » Publication #140625

Title: Incorporation of remote sensing data in an upscaled soil water model

item Seyfried, Mark

Submitted to: CRC Press
Publication Type: Book / Chapter
Publication Acceptance Date: 1/20/2003
Publication Date: 1/20/2003
Citation: Seyfried, M.S., CRC Press LLC., Incorporation of Remote Sensing Data in an Upscaled Soil Water Model. no. 18, p. 309-341.

Interpretive Summary: There is considerable interest in developing computer models that simulate vegetation production and water use on rangelands. These models are useful for estimating management impacts (e.g., from prescribed fire) and potential vegetation production on rangelands. A key component of these models is the relationship between soil-water content, plant growth and stream-flow. In order to be relevant to management needs, simulation model estimates must be capable of estimating soil water content changes across the landscape, and must cover relatively large areas. We documented a model that uses fairly simple parameters that can be extrapolated across the landscape at large spatial scales. The model also incorporates readily available remote sensing data as input. We described how these data can be used to provide spatially distributed model estimates, and showed that the model provides reasonably accurate estimates of soil water content. This modeling effort showed that readily available information derived from soil survey and remote sensing data sources can provide information at sufficient detail to describe major variations in soil moisture across the landscape.

Technical Abstract: There is great potential utility for relatively large-scale, spatially distributed soil water models. Application of these models requires consideration of scale and spatial variability effects on model parameters and ground measurements. We considered the following two aspects of spatial variability: extent, which determines the sources of variability, and aggregation density, which determines the nature of variability. These considerations should be an explicit part of model description in the context of the modeling approach used. Remote sensing is currently the only means of obtaining the spatially extensive data required for parameterization and testing of these models. The limitations of remote sensing data must be considered in the model design and interpretation. Some limitations are that remote sensing data: (i) are retrospective, (ii), represent a snapshot in time, (iii) have a resolution that is greater than the standard measurements, (vi) have a resolution that may not be appropriate relative to the landscape spatial variability, and (v) are not direct measures of the parameters of interest. We described field data and a modeling approach that uses definable, measurable parameters that scale linearly. We found that spatial aggregation at the level of the soil-mapping unit is large enough to provide a basis for stochastic description of small-scale spatial variability of soil water content while delineating, deterministically, critical larger-scale variability. Remote sensing of vegetation was shown to provide accurate information concerning relatively subtle changes in vegetation in the semiarid landscape. In addition, leaf area index in these sparsely vegetated areas can be accurately described with a vegetation index, although, the relationship is different from that used for other types of vegetation. Remote sensing data, specifically Landsat imagery, is spatially and functionally compatible with the modeling approach. Soil mapping units, used in the model to delineate critical deterministic variability of soil water content, aggregate Landsat pixels sufficiently that vegetation-cover-type and vegetation-index are effectively described within mapping units while also delineating differences among them. Functional compatibility was demonstrated by the correlation between model-calculated soil water stress and a remote-sensing-derived vegetation index.