.png){kind=icon}
返回
Vegetation height estimation based on machine learning model driven by multi-source data in Eurasian temperate grassland
.png){kind=icon}
PRE
AI
DOI：10.1016/j.ecolind.2024.113013.png)
.png){kind=icon}
原文求助.png){kind=icon}
分享.png){kind=icon}
收藏
摘要
En 中文
Eurasian Steppes, the world's largest grassland ecosystem, significantly contribute to our ecological environment. Grassland height is a crucial biophysical factor, offering insights into vertical vegetation structure and regional ecosystem dynamics. Therefore, investigating the trends in changes of grassland height in Steppes is vital for sustainable land use and ecological balance. This study utilized machine learning models such as Random Forest (RF), AdaBoost, BP-Neural Network (BPNN), and Stacking Ensemble, combining them with topographic and meteorological data, MODIS reflectance data, and grassland height measurement data. The first Eurasian temperate grassland vegetation height dataset was finally generated. The results highlighted the superior performance of the Random Forest with an R2 of 0.54, RMSE of 7.39 cm, and MPAE of 49.25 %. Spatial and temporal analyses revealed a discernible spatial distribution in grassland heights across Eurasia, characterized by a consistent decrease from west to east and north to south. Throughout the year of 2001-2021, there were minor fluctuations in changes of grassland heights, ranging from 15.93 to 17.99 cm, with an overall incremental trend of 0.027 cm annually. Significance tests indicated that approximately 4.40 % of the surveyed regions, primarily concentrated in northern Central Asia, experiencing a noteworthy increase in grassland height dynamics. Conversely, about 2.10 % of the areas, predominantly located in Inner Mongolia, China, and southern Central Asia, have witnessed a reduction in grassland height. This study reveals both temporal and spatial dynamics by offering a comprehensive dataset on vegetation height in Eurasian temperate grassland. These discoveries bear significant ramifications for land management strategies and the advancement of ecological sustainability objectives. Additionally, they contribute significantly to improve the accuracy of estimates of aboveground biomass and carbon stocks in grassland ecosystems.
Keyword:
Grassland height
Multi-source data
Machine Learning
Eurasian steppe
期刊
IF:
7.4
论文数:
1.3W
被引数:
7.1W
机构
引用论文
ERA5-Land: a state-of-the-art global reanalysis dataset for land applicationsERA5-Land: 最先进的全球土地应用再分析数据集
OAAI
OAAI分享收藏
Comparisons of heuristic, general statistical and machine learning models for landslide susceptibility prediction and mapping滑坡敏感性预测和制图的启发式，通用统计和机器学习模型的比较.png)
CATENA
CATENAIF5.7
PREAI
分享收藏Zinc Hydroxyfluoride (ZnOHF) Nanostructure as Photoelectrode of Quantum Dot-Sensitized Solar Cells
PREAI
分享收藏
查看更多内容