| 黄河流域山西段及矿区NPP时空动态和驱动机制分析 |
| 摘要点击 316 全文点击 13 投稿时间：2025-07-30 修订日期：2025-11-05 |
| 查看HTML全文
查看全文 查看/发表评论 下载PDF阅读器 |
| 中文关键词 净初级生产力（NPP） ACNN模型 时空变化 驱动机制 贡献分析 |
| 英文关键词 net primary productivity （NPP） ACNN model spatio-temporal variability driving mechanisms contribution analysis |
| DOI 10.13227/j.hjkx.202507394 |
|
| 中文摘要 |
| 黄河流域山西段生态系统植被净初级生产力的持续监测与精细评估，对实现该区域生态系统的可持续管理至关重要. 利用2003~2022年MOD17A3 HGF年度NPP产品数据，结合Theil-Sen Median趋势分析、Mann-Kendall 显著性检验及偏相关分析等方法，对该区域NPP时空变化进行综合分析. 同时，基于气温（TEM）、降水量（PRE）、太阳辐射（SR）以及归一化差值植被指数（NDVI）等多源数据，开发了一个ACNN深度学习模型（集成通道注意力机制的一维卷积神经网络）对NPP进行预测，并结合SHAP方法量化各驱动因子对NPP的贡献及其交互作用. 结果表明：①时间上，黄河流域山西段及矿区20 a NPP（以C计）皆呈波动上升趋势，整体NPP累计增长29.83%，矿区NPP累计增长31.14%. ②空间上，黄河流域山西段NPP呈上升趋势面积占95.93%，西北部地区主要呈强显著上升趋势. ③在NPP预测方面，ACNN模型能够有效从输入变量中提取与NPP相关的特征并且该模型准确性[R2=0.83，RMSE=0.041 4 kg·（m2·a）-1]高于1DCNN模型[R2=0.79，RMSE=0.044 2 kg·（m2·a）-1]. ④通过SHAP对ACNN模型进行解释，蒸散发（ET）和NDVI对NPP的贡献大于50%，与偏相关性分析结果一致，且它们之间存在显著的交互作用，如NDVI与蒸散发的协同效应显著促进了NPP的增长. |
| 英文摘要 |
| The continuous monitoring and refined assessment of the net primary productivity （NPP） of vegetation in the ecosystem of the Shanxi section of the Yellow River Basin are crucial for achieving the sustainable management of ecosystems in this region. In this study， MOD17A3 HGF annual NPP product data from 2003 to 2022 were used， combined with methods such as Theil-Sen Median analysis， Mann-Kendall significance test， and partial correlation analysis， to conduct a comprehensive analysis of the spatiotemporal changes in NPP in this region. Meanwhile， based on multi-source data including temperature （TEM）， precipitation （PRE）， solar radiation （SR）， and normalized difference vegetation index （NDVI）， a 1D convolutional neural network （1DCNN） deep learning model integrated with channel attention mechanism （referred to as the ACNN model） was developed to predict NPP. Additionally， the SHapley Additive exPlanations （SHAP） method was used to quantify the contribution of each driving factor to NPP and their interaction effects. The results show that： ①Temporally， the NPP （calculated in terms of carbon， C） in the Shanxi section of the Yellow River Basin and its mining areas both showed a fluctuating upward trend over the 20-year period. The overall NPP increased by 29.83% cumulatively， while the NPP in mining areas increased by 31.14% cumulatively. ② Spatially， the area with an upward trend of NPP in the Shanxi section of the Yellow River Basin accounted for 95.93%， and the northwestern region mainly showed a strongly significant upward trend. ③ In terms of NPP prediction， the ACNN model could effectively extract NPP-related features from input variables， and the accuracy of the proposed model [R2 = 0.83， RMSE = 0.041 4 kg·（m2·a）-1] was higher than that of the 1DCNN model [R2 = 0.79， RMSE = 0.044 2 kg·（m2·a）-1]. ④ Through the interpretation of the ACNN model using SHAP， the combined contribution of evapotranspiration （ET） and NDVI to NPP exceeded 50%， which was consistent with the results of the partial correlation analysis. Moreover， there was a significant interaction between them； for example， the synergistic effect of NDVI and evapotranspiration significantly promoted the increase in NPP. |
