Constraining canopy biophysical simulations with daily MODIS reflectance data ensuring pixel-target adequacy

Modern vegetation models incorporate ecophysiological details that allow for accurate estimates of carbon dioxide uptake, water use and energy exchange, but require knowledge of dynamic structural and biochemical traits. Variations in these traits are controlled by genetic factors as well as growth stage and nutrient and moisture availability, making them difficult to predict and prone to significant error. Here we explore the use of daily MODIS optical reflectance data for constraining key canopy- and leaf-level traits required by forward biophysical models. A multi-objective optimization algorithm is used to invert the PROSAIL canopy radiation transfer model against MODIS optical reflectance observations. PROSAIL accounts for the effects of leaf-level optical properties, foliage distribution and orientation on canopy reflectance across the optical range. Inversions are conducted for several growing seasons for both soybean and maize at multiple sites across the Central US agro-ecosystem. These inversions provide estimates of seasonal variations, and associated uncertainty, of variables such as leaf area index (LAI). The inversion-derived canopy properties are used to examine the ability of MODIS data to characterize seasonal variations in these states relative to field observations. The canopy properties are then used as inputs into the MLCan biophysical model to conduct forward simulations. MLCan characterizes the ecophysiological functioning of a plant canopy at a half-hourly timestep, and has been rigorously validated for both C3 and C4 crops against observations of canopy CO2 uptake, evapotranspiration and sensible heat exchange. By utilizing the inverted canopy states to drive MLCan over several growing seasons, we are able to assess the impact of uncertainty in the MODIS inversion procedure on uncertainties in forward model flux estimates. This work requires the use of instant (non-composited) observations obtained at a daily frequency from both Terra and Aqua platforms. As a whiskbroom imaging instrument, MODIS has a complex viewing geometry which affects its spatial response, i.e. the way the electromagnetic radiation reflected from the surface is ultimately encoded in the remotely-sensed image. A model of this spatial response is used here to ensure that the footprint of the satellite observations matches adequately with the coupled model simulations of the target fields. The relationship between the purity of the remote sensing observation, with respect to the target field, and the quality of the biophysical variable inversion is also investigated.