Can Land Surface Models Capture the Observed Soil Moisture Control of Water and Carbon Fluxes in Temperate- To- Boreal Forests?

The reservoir of soil water is exchanged with the atmosphere through evaporative processes, which are mediated via vegetation through transpiration. Carbon uptake is closely coupled to transpiration, and most process- based models link transpiration and photosynthesis explicitly. Here, we evaluate the simulation of water and carbon fluxes in forested areas of North America using point- based simulations of the Community Land Model version 5 (CLM5) with Soil Moisture Active/Passive (SMAP) satellite derived soil moisture measurements and in situ measurements at eight Fluxnet and two United States Climate Research Network (CRN) sites. Compared to observations, there is a broad site- to- site variability in simulated soil moisture, with some sites exhibiting wet biases and others dry biases. The bias sign does not depend on ecosystem or other environmental drivers such as radiation and temperature. Compared to Fluxnet latent heat (LH) and gross primary production (GPP) flux tower observations, simulated LH flux biases at deciduous broadleaf forests are linked with soil moisture biases, and the model captures the observed seasonal cycle and parabolic seasonal relationship with soil moisture. The parabolic shape is driven by high soil moisture and low LH fluxes in June, peak LH and drier soil conditions in July, followed by further soil moisture drawdown in August. GPP is underestimated at most sites, and the model exhibits a linear relationship between soil moisture and GPP. Because the photosynthesis parameterizations are similar in most Earth system models, further model development that incorporates observations and observed relationships is needed to accurately capture the GPP- soil moisture relationship.Key PointsCommunity Land Model (CLM) latent heat flux biases at deciduous sites are related to soil moisture biases, and capture the observed seasonal cycle and soil moisture relationshipsCLM modeled gross primary production is underestimated compared to Fluxnet and SMAP L4 data and is uncoupled from latent heat seasonal changesIndividual environmental variables (temperature, radiation, and soil moisture) do not provide a mechanistic explanation of site- to- site variability, suggesting complex canopy processesPeer Reviewedhttp://deepblue.lib.umich.edu/bitstream/2027.42/167528/1/jgrg21863_am.pdfhttp://deepblue.lib.umich.edu/bitstream/2027.42/167528/2/jgrg21863.pd