Testing a land model in ecosystem functional space via a comparison of observed and modeled ecosystem flux responses to precipitation regimes and associated stresses in a Central U.S. forest

Abstract

Testing complex land surface models has often proceeded by asking the question: does the model prediction agree with the observation? Such an approach has yet led to high-performance terrestrial models that meet the challenges of climate and ecological studies. Here we test the Community Land Model (CLM) by asking the question: does the model behave like an ecosystem? We pursue its answer by testing CLM in the ecosystem functional space (EFS) at the Missouri Ozark AmeriFlux (MOFLUX) forest site in the Central U.S., focusing on carbon and water flux responses to precipitation regimes and associated stresses. In the observed EFS, precipitation regimes and associated water and heat stresses controlled seasonal and interannual variations of net ecosystem exchange (NEE) of CO 2 and evapotranspiration in this deciduous forest ecosystem. Such controls were exerted more strongly by precipitation variability than by the total precipitation amount per se. A few simply constructed climate variability indices captured these controls, suggesting a high degree of potential predictability. While the interannual fluctuation in NEE was large, a net carbon sink was maintained even during an extreme drought year. Although CLM predicted seasonal and interanual variations in evapotranspiration reasonably well, its predictions of net carbon uptake were too small across the observed range of climate variability. Also, the model systematically underestimated the sensitivities of NEE and evapotranspiration to climate variability and overestimated the coupling strength between carbon and water fluxes. We suspect that the modeled and observed trajectories of ecosystem fluxes did not overlap in the EFS and the model did not behave like the ecosystem it attempted to simulate. A definitive conclusion will require comprehensive parameter and structural sensitivity tests in a rigorous mathematical framework. We suggest that future model improvements should focus on better representation and parameterization of process responses to environmental stresses and on more complete and robust representations of carbon-specific processes so that adequate responses to climate variability and a proper degree of coupling between carbon and water exchanges are captured.Overcoming this spaghetti problem of terrestrial models is an urgent need and a prerequisite for advancing Earth system science. Typically, model deficiencies at the system level are identified by comparing model GU ET AL. (2016), Testing a land model in ecosystem functional space via a comparison of observed and modeled ecosystem flux responses to precipitation regimes and associated stresses in a

Evaluation and uncertainty analysis of regional-scale CLM4.5 net carbon flux estimates

“…Studies which analyzed interannual variability in NEE detected the important role of temperature and precipitation as drivers of this variability (e.g., Keppel-Aleks et al, 2014). Gu et al (2016) found that for the Missouri Ozark AmeriFlux forest site, the interannual NEE variability is smaller in simulations by CLM than in the data. On the other hand, some field studies for pairs of monitoring sites (Kwon et al, 2006) or experimentation sites (with temperature increase and precipitation increase or decrease; Xu et al, 2016) found a limited impact of temperature and precipitation differences (or changes) on NEE.…”

Advancing Terrestrial Ecosystem Science With a Novel Automated Measurement System for Sun‐Induced Chlorophyll Fluorescence for Integration With Eddy Covariance Flux Networks

“…With this second objective, we wish to show how a robust SIF system can be designed for plug-and-play integration with existing EC systems for continuous, long-term synergistic observations with minimal maintenance in harsh environments. The third objective is to present initial findings from analyses of measurements made with the FAME prototype that was installed in September of 2016 at the Missouri Ozark AmeriFlux (MOFLUX) site (Gu et al, 2016). We will focus on the diurnal hysteresis patterns of SIF emission first reported with FAME.…”

Calibration of the E3SM Land Model Using Surrogate‐Based Global Optimization

“…In this study, the parameter values include the default parameter value and the parameter samples generated in the SG construction, which is described in detail in section 3. Wildfire is set to inactive in these simulations. The default model parameter values, simulation procedure, and structure are the same as those used in Gu et al (). For each model ensemble member, a 250 year accelerated decomposition and 250 year model equilibration are performed following Thornton and Rosenbloom ().…”

“…Wildfire is set to inactive in these simulations. The default model parameter values, simulation procedure, and structure are the same as those used in Gu et al (2016). For each model ensemble member, a 250 year accelerated decomposition and 250 year model equilibration are performed following Thornton and Rosenbloom (2005).…”

“…For both equilibration and transient simulations, we drive the model with continuously cycled half-hourly meteorological data as measured at the site from 2006-2014. Simulations are performed at the Missouri Ozark flux site, an AmeriFlux site (Baldocchi et al, 2001) that has been collecting eddy covariance data since 2004. The site consists of second-growth oak-hickory forests with a mean annual precipitation of 1,083 mm (Gu et al, 2016). The site is subject to frequent drought, most notably during the observation period in 2007 and 2012.…”