Improving Dynamic Vegetation Modeling in Noah‐MP by Parameter Optimization and Data Assimilation Over China's Loess Plateau. Issue 19 (8th October 2022)

• Record Type: Journal Article
• Authors: Shu, Zunyun; Zhang, Baoqing; Tian, Lei; Zhao, Xining
• Is Part Of: Journal of geophysical research. Volume 127:Issue 19(2022)
• Page Start: n/a
• Page End: n/a
• Publication Date: 2022-10-08
• Subjects: Noah‐MP -- vegetation dynamics -- parameter optimization -- data assimilation; Atmospheric physics -- Periodicals; Geophysics -- Periodicals; 551.5
• DOI: 10.1029/2022JD036703
• Languages: English
• ISSNs: 2169-897X

Abstract

Abstract: Accurate modeling of vegetation dynamics is needed to improve our understanding of and ability to predict the impacts of vegetation changes on terrestrial water‐energy‐carbon cycles. Parameter optimization (PO) and data assimilation (DA) are widely used to improve the performance of dynamic vegetation modules in land surface models (LSMs). However, their effectiveness is unclear. Here we analyze their impacts on the performance of the dynamic vegetation module of the Noah with multiparameterization options (Noah‐MP) LSM over the Chinese Loess Plateau, which is an ideal study case because it is a large region that has undergone dramatic vegetation change. We first optimize these parameters that strongly affect the predicted vegetation dynamics based on the results of sensitivity analysis using PO. In addition, we evaluate the effect of DA by assimilating leaf area index (LAI) remote sensing data into Noah‐MP without PO. Finally, we investigate the effect of applying PO and DA together. PO increases the predicted rates of carbon assimilation and turnover and thus reduces the underestimation of LAI and the lag in vegetation seasonality. DA has a weaker impact than PO: it only reduces the root mean squared error (RMSE) of the predicted LAI in around 49.76% of the studied region and is mainly beneficial in the growing phase. Combining PO and DA compensate the limitations of each other, and gives the most significant reduction in RMSE (median: −0.24 m 2 /m 2 ) and increase in R 2 (+0.44). The improved vegetation dynamics with different optimization methods thus improve the modeling of water and carbon cycle processes. Plain Language Summary: The performance of dynamic vegetation modules in land surface models still has considerable uncertainty. Two common methods, for example, parameter optimization (PO), and data assimilation (DA), have been widely used to improve model performance, while there are still limitations of each method. Here we first screen out the key vegetation‐related parameters based on the results of sensitivity analysis, and calibrate these parameters to identify how model parameters influence vegetation dynamic over the Chinese Loess Plateau. In addition, the satellite vegetation observations are assimilated into the default model to investigate the impacts of DA on the performance of dynamic vegetation module. Finally, the PO and DA are combined, and compared with each independent method. We find that the PO provides a more accurate representation of vegetation seasonality, and the DA largely reduces the bias of LAI magnitude during the growing phase. Generally, the combination of PO and DA takes full advantages of each method and presents the most significant reduction of model errors. The improved vegetation dynamics further provide beneficial impacts on simulating water and carbon cycle processes. This study is helpful for guiding dynamic vegetation module development and maximizing the effectiveness of DA especially over the areas with sparse vegetation. Key Points: Parameter optimization improves the modeling of vegetation dynamics by increasing the predicted rates of carbon assimilation and turnover Leaf area index (LAI) assimilation is more efficient for reducing the bias of the LAI magnitude in the growing phase Combining parameter optimization with data assimilation gives the best performance and mitigates the limitations of each method