School of Environmental and Municipal Engineering, Qingdao University of Technology, Qingdao 266033, China.
School of Information and Control Engineering, Qingdao University of Technology, Qingdao 266033, China.
Int J Environ Res Public Health. 2023 Jan 31;20(3):2513. doi: 10.3390/ijerph20032513.
The development of ecological restoration projects is unsatisfactory, and soil erosion is still a problem in ecologically restored areas. Traditional soil erosion studies are mostly based on satellite remote sensing data and traditional soil erosion models, which cannot accurately characterize the soil erosion conditions in ecological restoration areas (mainly plantation forests). This paper uses high-resolution unmanned aerial vehicle (UAV) images as the base data, which could improve the accuracy of the study. Considering that traditional soil erosion models cannot accurately express the complex relationships between erosion factors, this paper applies convolutional neural network (CNN) models to identify the soil erosion intensity in ecological restoration areas, which can solve the problem of nonlinear mapping of soil erosion. In this study area, compared with the traditional method, the accuracy of soil erosion identification by applying the CNN model improved by 25.57%, which is better than baseline methods. In addition, based on research results, this paper analyses the relationship between land use type, vegetation cover, and slope and soil erosion. This study makes five recommendations for the prevention and control of soil erosion in the ecological restoration area, which provides a scientific basis and decision reference for subsequent ecological restoration decisions.
生态恢复项目的发展不尽人意，水土流失仍然是生态恢复区的一个问题。传统的土壤侵蚀研究大多基于卫星遥感数据和传统的土壤侵蚀模型，无法准确描述生态恢复区（主要是人工林）的土壤侵蚀状况。本文利用高分辨率无人机（UAV）图像作为基础数据，可以提高研究的准确性。考虑到传统土壤侵蚀模型无法准确表达侵蚀因素之间的复杂关系，本文应用卷积神经网络（CNN）模型来识别生态恢复区的土壤侵蚀强度，可以解决土壤侵蚀的非线性映射问题。在研究区域内，与传统方法相比，应用改进的 CNN 模型进行土壤侵蚀识别的精度提高了 25.57%，优于基线方法。此外，基于研究结果，本文分析了土地利用类型、植被覆盖度和坡度与土壤侵蚀之间的关系。本文针对生态恢复区的土壤侵蚀防治提出了五点建议，为后续的生态恢复决策提供了科学依据和决策参考。
