A multimodel random forest ensemble method for an improved assessment of Chinese terrestrial vegetation carbon density

Abstract

1. Assessing the terrestrial vegetation carbon density (TVCD) is crucial for evaluating the national carbon balance. However, current national-scale TVCD assessments show strong disparities, despite the good estimation method of their underlying models. Here, we attribute this contradiction to a flaw in the methods of using multimodel simulation results, which ignore the connections between results, leading to an overoptimistic evaluation of the multimodel ensemble mean (MMEM) method.2. Thus, using the state-of-the-art multimodel random forest ensemble (MMRFE) method to integrate the results of 10 models, we reproduced Chinese TVCD data during 1982-2010.3. Compared with the nationally averaged TVCD field investigation data (27 ± 26 Mg C/ha), we found that the results of five models were overestimated by 7.4%-85.2%, and the remaining models were underestimated by 3.7%-77.8%. The MMEM TVCD method produced an overestimation of 2%, but the MMRFE method produced an underestimation of only 0.2%. Additionally, the summary Taylor diagrams of the TVCD at the national and ecosystem (forest, shrub, grass and crop ecosystems) scales all showed that the MMRFE TVCD produced the smallest standard deviations and root mean square deviations and the highest correlation coefficients. Furthermore, the MMRFE TVCDs were all significantly positively correlated with the normalized difference vegetation index (NDVI), and they had the same increasing trend, but an opposite variation trend from the MMEM TVCD and NDVI. This result implied that the spatiotemporal variation modes of the MMRFE TVCD were consistent with those of the NDVI. The results suggested that compared with the traditional MMEM method, the MMRFE TVCD and its spatiotemporal variation modes were more similar to the real TVCD. 4. In conclusion, the MMRFE method can effectively improve the accuracy of nationalscale TVCD estimation, and effectively reduce the uncertainty of large-scale terrestrial vegetation carbon estimation processes. Notably, we provide a new method thatThis is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.

Learning ensembles of process-based models for high accurately evaluating the one-hundred-year carbon sink potential of China’s forest ecosystem

“…To train and test the new model, we use 5-year (2011–2015) field-investigated plot vegetation carbon (labeled as ObscVeg) and soil carbon (labeled as ObscSoil) density datasets from Tang et al [ 26 ]. These data have been used to estimate the carbon pool of China’s terrestrial ecosystem [ 20 , 21 , [27] , [28] , [29] ]. In this study, using a bilinear interpolation technology, the ObscVeg and ObscSoil values of 7800 forest plots were interpolated to a grid surface with a spatial resolution of 0.01 * 0.01° in 2010, respectively.…”

“…Notably, advanced machine learning (ML) methods not only enhance the accuracy of carbon estimation of terrestrial ecosystems [ [14] , [15] , [16] ], but also enrich the technical methods [ 17 , 18 ]. Researchers have explored various ML models such as random forests [ [19] , [20] , [21] ], artificial neural networks [ 22 ], and support vector machines [ 23 , 24 ] to estimate carbon components. Additionally, studies have investigated the use of remote sensing data, including LiDAR and satellite imagery, in combination with ML algorithms for carbon mapping [ 19 , 21 ].…”

“…To accurately estimate carbon stock, and scientifically judge their carbon source/sink status of the CSP of China’s forest ecosystem, as a method of learning ensembles of process-based models for effectively improving the estimation accuracy of terrestrial ecosystem carbon stock [ 20 , 21 ], we employ a multi-model random forest ensemble (MMRFE) model (see detailed information in Section 2.4 ) for evaluating CSP of China’s forest ecosystem in the historical (1982–2021) and future (2022–2081) periods. Meanwhile, we also proposed a new carbon sink index method that takes into account the characteristics of the continuous stability of carbon sink function of ecosystem for scientifically identifying the carbon sink status of the forest ecosystem.…”

Realising the promise of large data and complex models

“…However, off-the-shelf algorithms may not be sufficient or may be too limiting, as described by Wang et al (2023), so additional developments may be required for ecological applications. For example, it will generally be important to incorporate known ecological processes within the data analysis.…”

“…The challenges that arise regarding scalability due to large (and new) datasets are also an opportunity for the development and use of machine learning algorithms. However, off‐the‐shelf algorithms may not be sufficient or may be too limiting, as described by Wang et al (2023), so additional developments may be required for ecological applications. For example, it will generally be important to incorporate known ecological processes within the data analysis.…”