Ecosystem models are often calibrated and/or validated against derived remote sensing data products (e.g. MODIS LAI). However, these data products are generally based on their own models, whose assumptions may not be compatible with those of the ecosystem model in question, and whose uncertainties are usually not well quantified. Here, we develop an alternative approach whereby we modify an ecosystem model to predict high spectral resolution surface reflectance (i.e. “hyperspectral†reflectance), which can then be compared directly against airborne and satellite data. Specifically, we coupled the two-stream representation of canopy radiative transfer in the Ecosystem Demography (ED2) model with a leaf radiative transfer model (PROSPECT 5). We then calibrated this model using airborne imaging spectroscopy (AVIRIS) and survey data from 54 temperate forest plots in the northeastern United States.
Results/Conclusions
The calibration successfully constrained the posterior distributions of model parameters related to leaf biochemistry and morphology (pigment and water contents; leaf mass per area) and canopy structure (leaf angle, canopy clumping, and leaf area index) for five plant functional types. However, comparisons of predicted spectra post-calibration against AVIRIS observations revealed biases in certain sparsely-vegetated canopies. Finally, we ran forward simulations of ED2, predicting surface reflectance at each time step and compared these predictions against Landsat time series. These simulations showed good agreement with Landsat during the growing season, suggesting that modeling surface reflectance is a promising avenue for applications of remote sensing time series to validation of vegetation dynamics in ecosystem models.