Assessing the multi-resolution information content of remotely sensed variables and elevation for evapotranspiration in a tall-grass prairie environment N.A. Brunsell a, ⁎, J.M. Ham b , C.E. Owensby b a Department of Geography, University of Kansas, Lawrence KS, 66045-7613, USA b Department of Agronomy, Kansas State University, Manhattan KS, 66506, USA article info abstract Article history: Received 27 August 2007 Received in revised form 13 February 2008 Accepted 16 February 2008 Understanding the spatial scaling behavior of evapotranspiration and its relation to controlling factors on the land surface is necessary to accurately estimate regional water cycling. We propose a method for ascertaining this scaling behavior via a combination of wavelet multi-resolution analysis and information theory metrics. Using a physically-based modeling framework, we are able to compute spatially distributed latent heat fluxes over the tall-grass prairie in North-central Kansas for August 8, 2005. Comparison with three eddy-covariance stations and a large aperture scintillometer demonstrates good agreement, and thus give confidence in the modeled fluxes. Results indicate that the spatial variability in radiometric temperature (a proxy for soil moisture) most closely controls the spatial variability in evapotranspiration. Small scale variability in the water flux can be ascribed to the small scale spatial variance in the fractional vegetation. In addition, correlation analysis indicates general scale invariance and that low spatial resolution data may be adequate for accurately determining water cycling in prairie ecosystems. © 2008 Elsevier Inc. All rights reserved. Keywords: Konza prairie Latent heat Information theory Entropy Wavelets SVAT model Spatial heterogeneity MODIS 1. Introduction Grassland ecosystems occupy approximately 40% of the world's land surface, comprise one of the largest land cover types in the United States and account for significant amounts of grazing lands (Samson & Knopf, 1994). Understanding the spatial–temporal dynamics of water cycling throughout this region of the country has a direct impact on both climatic and economic processes across the globe. The importance of grasslands on the cycling of water and other fluxes has been recognized for quite some time, becoming the focus of the First ISLSCP Field Experiments (FIFE) experiments which focused on the inter-comparison of remote sensing and surface measurements at different scales (Sellers et al., 1995). In addition, the carbon and water cycles are intricately linked, and it is necessary to consider the flow of water, carbon and energy simultaneously (e.g. Ham & Knapp, 1998; Ham et al., 1995). Currently, the most promising method of measuring patch-scale (e.g. on the order of 100s of meters) dynamics of the water cycle is with the use of eddy-covariance (Baldocchi, 2003; Baldocchi et al., 1988). This has led to the formation of the FLUXNET international network of towers being established in order to monitor mass and energy cycling in different biomes (Baldocchi et al., 2001). These towers have led to reasonable estimates (i.e. on the order of 10% error) of net mass and energy transfer in grassland ecosystems as well as linking the seasonal and annual dynamics to variability in soil moisture, soil and air temperatures, and leaf area index (Barcza et al., 2003; Flanagan et al., 2002; Ham & Knapp,1998; Owensby et al., 2007; Suyker & Verma, 2001). Much of the research relating observed water and energy fluxes to environmental variables has resulted in empirical relationships that were developed from eddy-covariance data and do not consider spatial–temporal dynamics at larger scales. If we are to spatially extend beyond where the empirical relationships were established and develop methods for predicting future carbon exchange and assessing the impact of land-use management scenarios (Trumbore, 2006), we must couple the measurements with a mechanistic model. Linking the measurements from eddy-covariance systems to the responsible biophysical processes requires a linkage between tower data and a modeling framework (Braswell et al., 2005; Falge et al., 2005; Novick et al., 2004). The model framework usually adopted for small spatial scales is the Soil–Vegetation–Atmosphere-Transfer (SVAT) scheme coupled with an Atmospheric Boundary Layer (ABL) model. Such a scheme allows us to couple the physical mechanisms between a number of processes; such as linkages between the water and carbon cycles (e.g. Zhang et al., 2002). Ultimately, this will enable us to predict water, carbon and energy fluxes at the same scale used to make land management decisions, which for the prairie is the pasture or watershed extent (e.g. 50–300 ha). Some SVAT models also provide a framework for assessing the spatial variability in water and energy cycling by combining remotely sensed surface conditions with the model to quantify spatially variable fluxes. This allows a mechanistic framework for scaling Remote Sensing of Environment 112 (2008) 2977–2987 ⁎ Corresponding author. Tel.: +1 785 864 2021; fax: +1 785 864 5378. E-mail address: brunsell@ku.edu (N.A. Brunsell). 0034-4257/$ – see front matter © 2008 Elsevier Inc. All rights reserved. doi:10.1016/j.rse.2008.02.002 Contents lists available at ScienceDirect Remote Sensing of Environment journal homepage: www.elsevier.com/locate/rse