Ground and surface temperature variability for remote sensing of soil moisture in a heterogeneous landscape
At the Little River Watershed (LRW) heterogeneous landscape near Tifton Georgia US an in situ network of stations operated by the US Department of Agriculture–Agriculture Research Service-Southeast Watershed Research Lab (USDA–ARS-SEWRL) was established in 2003 for the long term study of climatic and soil biophysical processes. To develop an accurate interpolation of the in situ readings that can be used to produce distributed representations of soil moisture (SM) and energy balances at the landscape scale for remote sensing studies, we studied (1) the temporal and spatial variations of ground temperature (GT) and infra red temperature (IRT) within 30 by 30 m plots around selected network stations; (2) the relationship between the readings from the eight 30 by 30 m plots and the point reading of the network stations for the variables SM, GT and IRT; and (3) the spatial and temporal variation of GT and IRT within agriculture landuses: grass, orchard, peanuts, cotton and bare soil in the surrounding landscape. The results showed high correlations between the station readings and the adjacent 30 by 30 m plot average value for SM; high seasonal independent variation in the GT and IRT behavior among the eight 30 by 30 m plots; and site specific, in-field homogeneity in each 30 by 30 m plot. We found statistical differences in the GT and IRT between the different landuses as well as high correlations between GT and IRT regardless of the landuse. Greater standard deviations for IRT than for GT (in the range of 2–4) were found within the 30 by 30 m, suggesting that when a single point reading for this variable is selected for the validation of either remote sensing data or water-energy models, errors may occur. The results confirmed that in this landscape homogeneous 30 by 30 m plots can be used as landscape spatial units for soil moisture and ground temperature studies. Under this landscape conditions small plots can account for local expressions of environmental processes, decreasing the errors and uncertainties in remote sensing estimates caused by landscape heterogeneity.