Improving model performance in mapping black-soil resource with machine learning methods and multispectral features
Authors:
Open Access
- 7 January 2025
- journal article
- research article
- Published by Springer Nature in Scientific Reports
- Vol. 15 (1) , 1-15
Abstract
Accurate information on the distribution of regional black-soil resource is one of the important elements for the sustainable management of soils. And its results can provide decision makers with robust data that can be translated into better decision making. This study utilized all Sentinel-2 images covering the study area from April to July in 2022. After masking clouds, all images were synthesized monthly. Based on the revised random forest classification algorithm, model performance using different feature combination programs were evaluated to search for an efficient, high-precision method for mapping black-soil resource. The impact on model performance of adding data from temperature, precipitation and slope geographic covariates was analyzed. And the robustness of the model was verified using Landsat-8 data with lower spatial resolution. The results showed that (1) the model based on multi-temporal ensemble features for mapping black-soil resource shows the best performance, with an OA of 94.6%; (2) adding temperature covariate can effectively improve the accuracy of black-soil resource mapping; (3) compared to the sentinel data, the performance of the model based on Landsat-8 data is reduced but still plausible, verifying the robustness of the model. This study provides a robust method to improve model performance for rapid mapping of black-soil resource.Keywords
Funding Information
- Program of China Geological Survey (DD 20243191, DD20220862, DD 20243191, DD20220862)
- National Key Research and Development Program of China (SQ2023YFD1500029, SQ2023YFD1500029)
- cience and Technology Innovation Foundation of Command Center of Integrated Natural Resources Survey Cente (KC20220005, KC20220005)
This publication has 56 references indexed in Scilit:
- Spatiotemporal variation and GeoDetector analysis of NDVI at the northern foothills of the Yinshan Mountains in Inner Mongolia over the past 40 yearsHeliyon, 2024
- Union With Recursive Feature Elimination: A Feature Selection Framework to Improve the Classification Performance of Multicategory Causes of Death in Colorectal CancerLaboratory Investigation, 2023
- A remote sensing-based strategy for mapping potentially toxic elements of soils: Temporal-spatial-spectral covariates combined with random forestEnvironmental Research, 2023
- Balanced fertilization over four decades has sustained soil microbial communities and improved soil fertility and rice productivity in red paddy soilScience of The Total Environment, 2021
- High-resolution digital mapping of soil organic carbon and soil total nitrogen using DEM derivatives, Sentinel-1 and Sentinel-2 data based on machine learning algorithmsScience of The Total Environment, 2020
- Incorporating limited field operability and legacy soil samples in a hypercube sampling design for digital soil mappingJournal of Plant Nutrition and Soil Science, 2016
- Remote estimation of soil organic matter content in the Sanjiang Plain, Northest China: The optimal band algorithm versus the GRA-ANN modelAgricultural and Forest Meteorology, 2016
- Soil organic‐matter stocks and characteristics along an Alpine elevation gradientJournal of Plant Nutrition and Soil Science, 2010
- Discriminative Methods for Multi-labeled Classification2004
- A soil-adjusted vegetation index (SAVI)Remote Sensing of Environment, 1988