Evaluation of a Data Assimilation System for Land Surface Models Using CLM4.5

Abstract

The magnitude and persistence of land carbon (C) pools influence long-term climate feedbacks. Interactive ecological processes influence land C pools and our understanding of these processes is imperfect so land surface models have errors and biases when compared to each other and to real observations. Here we implement an Ensemble Adjustment Kalman Filter (EAKF), a sequential state data assimilation technique to reduce these errors and biases. We implement the EAKF using the Data Assimilation Research Testbed coupled with the Community Land Model (CLM 4.5 in CESM 1.2). We assimilated simulated and real satellite observations for a site in central New Mexico, United States. A series of observing system simulation experiments allowed assessment of the data assimilation system without model error. This showed that assimilating biomass and leaf area index observations decreased model error in C dynamics forecasts (29% using biomass observations and 40% using leaf area index observations) and that assimilation in combination shows greater improvement (51% using both observation streams). Assimilating real observations highlighted likely model structural errors and we implemented an adaptive model-variance-inflation technique to allow the model to track the observations. Monthly and longer model forecasts using real observations were improved relative to forecasts without data assimilation. The reliable forecast lead-time varied by model pool and is dependent on how tightly the C pool is coupled to meteorologically driven processes. The EAKF and similar state data assimilation techniques could reduce errors in projections of the land C sink and provide more robust forecasts of C pools and land-atmosphere exchanges. Plain Language Summary The amount of carbon stored in vegetation and soils is an importantcontrol on how much carbon dioxide is in the atmosphere, and that influences future climate. Land surface models are used to simulate where this carbon is, but they are imperfect and there are often differences between model predictions and observations of the carbon stores. Here we describe a system that combines model predictions and observations and updates the modeled carbon stores so they are closer to the observations, considering uncertainty in both the model and the observations. We test our system at a location in New Mexico, United States, where we use observations from satellites of the amount of leaves on the vegetation and the amount of carbon stored in the vegetation. When we combine these observations with our land surface model there are large changes in the predicted amounts of stored carbon and the times of the year when the vegetation has the most leaves. These changes persist in the model after we stop updating it with observations, improving the model forecast.FOX ET AL.

Assimilation of Remotely Sensed LAI Into CLM4CN Using DART

“…There are three ways for improving the model performance by assimilating observations into the CLM (Fox et al, ). The first and most direct way is to update modeled LAI based on the calculated Kalman gain.…”

“…Although DART have provided accesses to multiple assimilation algorithms (e.g., the EnKF and particle filter), the EAKF algorithm is the most mature technique developed for land surface DA within DART/CLM. Many studies using DART coupling with CLM have used the EAKF algorithms (Fox et al, ;Zhang et al, ; Zhao et al, ), which is a fully deterministic algorithm for estimating model forecast error statistics based on observation uncertainty (Anderson, ). Comparison with other algorithms should be conducted in the future work, for example, particle filters may provide a means to capture non‐Gaussian errors (Moradkhani et al, ).…”

“…Rüdiger et al (2010), Albergel et al (2010Albergel et al ( , 2017Albergel et al ( , 2019, and Barbu et al (2011) have assimilated satellite-derived vegetation and soil moisture products based on a Simplified Extended Kalman Filter and have found a strong impact on LAI itself, as well as significant improvements for river discharge, land ET, and gross primary production (GPP). Similar studies were conducted in the fields of agriculture (Bao et al, 2015;Dong et al, 2013) and hydrology, and most were conducted at a single site or on regional scales (Fox et al, 2018;Li et al, 2017;Montzka et al, 2012;Pauwels et al, 2007;Sabater et al, 2008).…”

“…There are three high LAI regions located in the tropics, including the Amazon, central Africa, and some islands in Southeast Asia. The region with large LAI values in the high latitudes of the Northern Hemisphere in July does not There are three ways for improving the model performance by assimilating observations into the CLM (Fox et al, 2018). The first and most direct way is to update modeled LAI based on the calculated Kalman gain.…”

Constraining estimates of terrestrial carbon uptake: new opportunities using long‐term satellite observations and data assimilation

“…LUE), leading to the largest reduction in model ensemble error. Importantly, DA using EAKF as well as other common ensemble filters requires observational data with well‐characterized uncertainty, thus reducing the likelihood of DA failure due to divergence between the model ensemble and the observations (Fox et al ., ). Looking forward we suggest close collaboration between remote sensing and DA experts to rapidly improve our understanding of CFE and enable more accurate forecasts of terrestrial carbon uptake.…”