基于DNN算法的杭埠丰乐流域土壤有机碳空间分布制图研究  

Study on spatial distribution mapping of soil organic carbon in Fengle River of Hangbu based on DNN algorithm

作　　者：赵武 Zhao Wu(Lu’an Shucheng County Water Conservancy Bureau,Shucheng 231300,China)

机构地区：[1]安徽省六安市舒城县水利局,安徽舒城231300

出　　处：《吉林水利》2024年第6期74-78,共5页Jilin Water Resources

摘　　要：土壤有机碳(SOC)的管理处于温室气体减排的前沿,了解SOC精细化空间分布,对气候变化政策的规划和制定至关重要。本研究利用多源遥感数据和深度神经网络(DNN)算法,制取杭埠丰乐河流域SOC空间分布图。结果表明,该流域SOC变化范围介于3.8~135.90,呈现中度空间异质性;众多环境变量中以海拔、温度对SOC变化的影响最大;DNN通过逐层分布式数据学习,能有效拟合SOC随环境变量分布规律,其验证精度R2达0.76,RMSE为10.52。本研究不仅有助于深化对杭埠丰乐河流域SOC空间变异性的认识,还为遥感技术和深度学习在土壤科学研究中的应用提供了新思路。The management of soil organic carbon(SOC)is at the forefront of greenhouse gas emission reduction,and understanding the fine-grained spatial distribution of SOC is vital importane for climate change policy formulation and planning.This study used multi-source remote sensing data and deep neural network(DNN)algorithms to produce a spatial distribution map of SOC in the Fengle River basin in Hangbu.The results show that the SOC varies from 3.8 to 135.90,with moderate spatial heterogeneity;among the many environmental variables,elevation and temperature have the greatest influence on the SOC changes;DNN can effectively fit the distribution of SOC with the environmental variables through the layer-by-layer distributed data learning with the validation accuracy of R2 of 0.76 and the RMSE of 10.52.This study not only helps to deepen the understanding of the SOC distribution of Hangbu Fengle River,but also provide new ideas for the application of remote sensing technology and deep learning in soil science research.

关 键 词：土壤有机碳 多源遥感数据 深度神经网络 

分 类 号：P237[天文地球—摄影测量与遥感]