Preprints
https://doi.org/10.5194/bg-2020-36
https://doi.org/10.5194/bg-2020-36
25 Feb 2020
 | 25 Feb 2020
Status: this preprint has been withdrawn by the authors.

Improving maps of forest aboveground biomass: A combined approach using machine learning with a spatial statistical model

Shaoqing Dai, Xiaoman Zheng, Lei Gao, Chengdong Xu, Shudi Zuo, Qi Chen, Xiaohua Wei, and Yin Ren

Abstract. Aboveground biomass (AGB) estimates at the plot level plays a major part in connecting accurate single-tree AGB measurements to relatively difficult regional-scale AGB estimates. However, complex and spatially heterogeneous landscapes, where multiple environmental covariates (such as longitude, latitude, and forest structure) affect the spatial distribution of AGB, make upscaling of plot-level models more challenging. To address this challenge, this study proposes an approach that combines machine learning with spatial statistics to construct a more accurate plot-level AGB model. The study was conducted in a Eucalyptus plantation in Nanjing, China. We developed, evaluated, and compared the accuracy and performance of three different machine learning models [support vector machine (SVM), random forest (RF), and the radial basis function artificial neural network (RBF-ANN)], one spatial statistics model (P-BSHADE), and three combinations thereof (SVM & P-BSHADE, RF & P-BSHADE, RBF-ANN & P-BSHADE) for forest AGB estimates based on AGB data from 30 sample plots and their corresponding environmental covariates. The results show that the performance indices RMSE, nRMSE, MAE, and MRE of all combined models are substantially smaller than those of any individual models, with the RF & P-BSHADE combined method giving the smallest value. These results demonstrate clearly that combined models, especially the RF & P-BSHADE model, can improve the accuracy of plot-level AGB models and reduce uncertainty on plot-level AGB estimates or even on large-forested-landscape AGB estimates. These research results are important because they reduce the uncertainty in estimates of the regional carbon balance.

This preprint has been withdrawn.

Shaoqing Dai, Xiaoman Zheng, Lei Gao, Chengdong Xu, Shudi Zuo, Qi Chen, Xiaohua Wei, and Yin Ren

Interactive discussion

Status: closed
Status: closed
AC: Author comment | RC: Referee comment | SC: Short comment | EC: Editor comment
Printer-friendly Version - Printer-friendly version Supplement - Supplement

Interactive discussion

Status: closed
Status: closed
AC: Author comment | RC: Referee comment | SC: Short comment | EC: Editor comment
Printer-friendly Version - Printer-friendly version Supplement - Supplement
Shaoqing Dai, Xiaoman Zheng, Lei Gao, Chengdong Xu, Shudi Zuo, Qi Chen, Xiaohua Wei, and Yin Ren
Shaoqing Dai, Xiaoman Zheng, Lei Gao, Chengdong Xu, Shudi Zuo, Qi Chen, Xiaohua Wei, and Yin Ren

Viewed

Total article views: 1,248 (including HTML, PDF, and XML)
HTML PDF XML Total Supplement BibTeX EndNote
861 348 39 1,248 194 36 38
  • HTML: 861
  • PDF: 348
  • XML: 39
  • Total: 1,248
  • Supplement: 194
  • BibTeX: 36
  • EndNote: 38
Views and downloads (calculated since 25 Feb 2020)
Cumulative views and downloads (calculated since 25 Feb 2020)

Viewed (geographical distribution)

Total article views: 1,146 (including HTML, PDF, and XML) Thereof 1,140 with geography defined and 6 with unknown origin.
Country # Views %
  • 1
1
 
 
 
 

Cited

Latest update: 28 Mar 2024
Download

This preprint has been withdrawn.

Short summary
This paper proposes a method to integrate the advantages of machine learning and spatial statistics, different datasets, and multiple environmental covariates to improve the accuracy of aboveground biomass estimation models, which provides a useful reference for climate change mitigation. This combined method can make full use of data from different sources, and realize the complementary advantages of machine learning and spatial statistics, which has important implications for other fields.
Altmetrics