Spectral Reflectance as a Covariate for Estimating Pasture Productivity and Composition

estimating unsampled points for varying sampling intervals and densities was calculated (Webster et al., 1989). Pasturelands are inherently variable. It is this variability that makes The reflectance data measured in the Webster et al. sampling as well as characterizing an entire pasture difficult. Measurement of plant canopy reflectance with a ground-based radiometer (1989) study was optimal in the sense that it was a spatially offers an indirect, rapid, and noninvasive characterization of pasture correlated variable that was rapidly and densely colproductivity and composition. The objectives of this study were (i) lected. In this study, data of multispectral canopy reflecto determine the relationships between easily collected canopy reflectance were used for similar reasons. tance data and pasture biomass and species composition and (ii) to Multispectral reflectance measured with hand-held determine if the use of pasture reflectance data as a covariate imradiometers has been used to estimate many plant paproved mapping accuracy of biomass, percentage of grass cover, and rameters of interest. Reflectance has been correlated percentage of legume cover across three sampling schemes in a central with plant greenness in peanut (Arachis hypogaea L.) Iowa pasture. Reflectance values for wavebands most highly corre(Nutter, 1989; Aquino et al., 1992) and in maize (Zea lated with biomass, percentage of grass cover, and percentage of mays L.) (Ma et al., 1996). Reflectance measurements legume cover were used as covariates. Cokriging was compared with kriging as a method for estimating these parameters for unsampled were also found to be successful estimators of biomass sites. The use of canopy reflectance as a covariate improved prediction in alfalfa (Medicago sativa L.) (Mitchell et al., 1990), of grass and legume percentage of cover in all three sampling schemes peanut (Nutter and Littrell, 1996), and potato (Solanum studied. The prediction of above-ground biomass was not as consistent tuberosum L.) (Bouman et al., 1992). Seasonal biomass given that improvement with cokriging was observed with only one changes in tallgrass prairies were modeled by the norof the sampling schemes because of the low amount of spatial continumalized difference vegetation index (NDVI) along with ity of biomass values. An overall improvement in root mean square several other environmental variables (Olson and Cocherror (RMSE) for predicting values for unsampled sites was observed ran, 1998). Light reflectance before anthesis may be able when cokriging was implemented. Use of rapid and indirect methods to predict grain yield in corn (Ma et al., 1996), canopy for quantifying pasture variability could provide useful and convenient reflectance measurement at pod setting stage in soybean information for more accurate characterization of time consuming parameters, such as pasture composition. aided in early prediction of soybean [Glycine max (L.) Merr.] yield (Ma et al., 2001), and a good correlation was found between NDVI and millet total dry matter at harvest (Lawrence et al., 2000). S is the researcher’s best way to learn about Reflectance indices involving different wavelengths a population. When a pasture is considered to be a can also be used to discriminate between weed and crop population, the task is to determine where to sample species (Vrindts et al., 2002). Discriminant analysis in and how to assess the true variability as accurately as this study also resulted in 94% correct classification of possible. With only a limited number of observations broadleaved plants in test datasets of broadleaved plants attainable because of time and labor constraints, interpoand grasses (Vrindts et al., 2002). lation is necessary to estimate values at unsampled points. Measuring pasture variability through the use of a Matheron’s theory of “regionalized variables” (Mathground-based multispectral radiometer can be pereron, 1971) reported that field variables can be spatially formed quickly, nondestructively, and inexpensively. correlated, or coregionalized. Geostatistics is the field of Consequently, canopy reflectance data on a dense grid study that models spatial variability and is used to predict can be easily obtained. This dense data collection can be unknown values in space (Journel and Huijbregts, 1978). capitalized on through the use of geostatistics. Kriging is Consequently, spatial correlation within pastures is an a method of interpolation used when a variable displays opportunistic reality. Webster et al. (1989) capitalized spatial autocorrelation. Because reflectance values are on this observation by designing a sampling scheme for spatially correlated (Webster et al., 1989), kriging can ground-based radiometry measurements in both spebe used to predict reflectance at unsampled points. Cocies-poor and species-rich grassland and winter barley kriging is also an interpolation method used where there (Hordeum vulgare L.). By fitting a semivariogram to are two or more spatially interdependent variables. Ofradiation reflectance data sets, the error associated with ten, cokriging is used when one or more other properties have been extensively sampled in comparison to the A.B. Tarr, USDA-NASS, Des Moines, IA 50309; K.J. Moore, Dep. variable of interest (Oliver, 1987). Ideally, the densely of Agronomy and P.M. Dixon, Dep. of Statistics, Iowa State Univ., sampled variable, termed a covariate, secondary variAmes, IA 50011. Received 5 Jan. 2004. Forage & Grazing Lands. able, or subsidiary variable, is measured more cheaply *Corresponding author ([email protected]). and quickly than the property of interest, or target variPublished in Crop Sci. 45:996–1003 (2005). doi:10.2135/cropsci2004.0004 © Crop Science Society of America Abbreviations: NDVI, normalized difference vegetation index; RMSE, root mean square error. 677 S. Segoe Rd., Madison, WI 53711 USA 996 Published online May 6, 2005