Rapid Operational Derivation of Surface Reflectance Values for Near-realtime Applications of Satellite Data

Modeling atmospheric radiative transfer is a highly complex and computationally expensive process. Simple empirical atmospheric correction methods such as dark object subtraction allow for quick image normalization but are difficult to fully automate. Analogous methods provided poor surface reflectance values for physical modeling and comparison with reflective spectra. Satellite surface reflectance values are required inputs to many operational uses of satellite data, including calculations for various evapotranspiration models. The Tasumi Atmospheric Correction Algorithm (TACA) uses atmospheric correction functions developed to require only general humidity data and a digital elevation model (Tasumi et al., 2008). These functions have a reduced structure to increase their operational applicability in routine calculation of instantaneous surface energy balances and to calculate normalized difference vegetation index values. This approach has the advantage that it does not require extensive ancillary data products to fully model radiative transfer processes. This increases its potential use by a broader range of agricultural scientists and engineers. This surface reflectance derivation procedure has been developed primarily for use with Landsat imagery, but additional analysis and preliminary testing has been conducted on Sentinel 2A and 2B imagery. We present results for analyses conducted with both Landsat 8 and Sentinel-2 data. To date, comparisons have been conducted for surface reflectances calculated using the TACA model against USGS Tier 1 Landsat surface reflectance data available for multiple Landsat scenes in California. Results to date indicate that the average difference in NDVI values calculated from the red and near infrared bands are within .01 to .03 (8.71% to 4.99%).