SCIENCE@home

Comparing different methods for biomass modeling over tropical region based on Landsat data

Manyao Li

Published 2023 in International Conference on Geographic Information and Remote Sensing Technology

ABSTRACT

Forest biomass is an important ecological parameter for forest research, which is greatly important to the understanding and analysis of global carbon cycle. Landsat Time Series (LTS) data, as medium optical resolution data, have the merits of long time–span and appropriate spatial resolution. Forest biomass research based on this data can better guide human understanding of forest ecosystem. In this study, LTS data and sample biomass data were used to study the forest biomass modeling. This study explored the effects of forest biomass modeling based on features extracted from two LTS processing methods and three algorithms including random forest (RF), extremely randomized trees (ERT), and eXtreme Gradient Boosting (XGBoost). Among these models, the accuracy of the models based on the features extracted by Best Available Pixel (BAP) image composite and three algorithms is generally higher. Among the models constructed by combining feature groups extracted from BAP image composite and three algorithms, the models constructed by ERT were stable in most situations from the perspective of coefficient of determination (R2), root mean square error (RMSE), mean absolute error (MAE), and bias. However, the XGBoost model that combines all variables performed best in perspective of R2 and RMSE, where R2 can reach 0.87, and the RMSE, MAE, and bias were 34.84 Mg/ha, 19.26 Mg/ha, and 2.29 Mg/ha, respectively. The textures and Tasseled Cap (TC) indices also show favorable performances in different models in the feature importance analysis.

PUBLICATION RECORD

  • Publication year

    2023

  • Venue

    International Conference on Geographic Information and Remote Sensing Technology

  • Publication date

    2023-02-10

  • Fields of study

    Engineering, Environmental Science

  • Identifiers
    DOI 10.1117/12.2667877
  • External record

    Open on Semantic Scholar

  • Source metadata

    Semantic Scholar

CITATION MAP

EXTRACTION MAP

CLAIMS

  • No claims are published for this paper.

CONCEPTS

  • No concepts are published for this paper.

REFERENCES

Showing 1-24 of 24 references · Page 1 of 1

CITED BY