Assessing the time scales of flux tower-model agreement in relation to environmental condition

Understanding the extent of agreement between land surface models and observations can provide insight into theoretical advancements in our understanding of land-atmosphere interactions. In particular, understanding the conditions under which models perform particularly well or poorly is essential for identifying potential model limitations. Here, we use three eddy covariance towers, one annually burned C4 grassland, one abandoned agricultural site consisting of both C3 and C4 vegetation and a C4 grassland experiencing woody encroachment. We use these sites to assess the agreement with the Noah and Noah-MP ("Multi-Physics") land surface models as a function of environmental variables and regime (e.g. high soil moisture, high wind, etc.). The data spans 2007-2012 and encompasses both normal and drought conditions. The environmental regimes are isolated using self-organizing maps (SOMs) to diagnose the relative importance of the microclimatic factors (soil moisture, air temperature, humidity, solar radiation, wind-speed, etc.) on the resulting water and carbon fluxes. The temporal variability of model limitations is assessed with an information theory based wavelet technique within each environmental class. Discussion will focus on the role of predicting potential model biases as a function of environmental condition.