Objective The spatiotemporal evolution patterns and driving factors of vegetation net primary productivity (NPP) in the Yangtze River delta region from 2001 to 2023 were explored in order to provide a scientific basis for formulating rational ecological protection strategies in this region. Methods Based on long-terrn series MODIS NPP data, the spatiotemporal evolution characteristics of vegetation NPP in the Yangtze River delta and its influencing factors were analyzed using trend analysis, Hurst index analysis, partial correlation analysis, residual analysis, random forest regression, and SHAP interpretable analysis. Results From 2001 to 2023, vegetation NPP in the Yangtze River delta showed an increasing trend, with an average annual increase of 3.39 g/(m²·a) (calculated by carbon). Its spatial distribution was characterized by high values in the south and low values in the north, among which the mean NPP of Zhejiang Province was significantly higher than that of other provinces. At the grid scale, it was found that the average Theil-Sen value of the study area was 2.71 g/(m²·a), with vegetation NPP showing an overall increasing trend. The increase in vegetation NPP was especially pronounced in areas along both banks of the Yangtze River and around rivers and lakes. The average Hurst index of regional vegetation NPP was 0.72, indicating that the overall NPP in the Yangtze River delta would continue to increase in the future. Climatic factors contributed more to vegetation NPP variation than human activities and exerted a positive effect. Furthermore, random forest and SHAP driving factor analysis showed that precipitation, vegetation type, and elevation had relatively strong positive effects on vegetation NPP values in the study area, and that the effects of all influencing factors exhibited obvious nonlinear characteristics. Conclusion From 2001 to 2023, vegetation NPP in the Yangtze River delta generally maintained a steady growth trend, with ecological environmental factors playing a crucial role in NPP variation. In the future, continuous efforts should be made in the improvement of ecological environment of the Yangtze River delta to achieve sustainable carbon sink enhancement.
文献参数: 吴达, 邵光成, 陈广兵, 等.长江三角洲地区植被净初级生产力时空演变及其驱动因素[J].水土保持通报,2026,46(3):435-446. Citation:Wu Da, Shao Guangcheng, Chen Guangbing, et al. Spatiotemporal evolution and driving factors of vegetation net primary productivity in Yangtze River delta region [J]. Bulletin of Soil and Water Conservation,2026,46(3):435-446.
植被NPP的变化是气候因素和人为因素共同作用的结果,本文以气温、降水作为气候变化因子,根据气候因子的变化设定和计算植被NPP预测值(NPP C ),利用实际植被NPP值与根据气候因子计算得到的预测值的差值作为人为因素对植被NPP的影响值(NPP H )。
如图6所示,2001—2023年长三角地区植被NPP C 和NPP H 年均值均呈波动上升趋势。NPPC由2001年的513.24 g/(m2·a)上升至2023年的558.19 g/(m2·a),年均增长量为2.04 g/(m2·a)。NPPH由2001年的10.11 g/(m2·a)上升至2023年的35.19 g/(m2·a),年均增长量为1.14 g/(m2·a)。
2.4.3 NPP C 和NPP H 空间变化特征
如图7a所示,2001—2023年长三角地区植被NPPC年变化的变化趋势在-1.85~5.43 g/(m2·a),平均值为1.32 g/(m2·a),变化趋势呈现增加趋势的面积比例为80.52%,主要分布在安徽省南部、江苏省南部、上海市与浙江省北部,变化趋势呈现减少趋势的区域主要分布在安徽省北部、江苏省北部与浙江省南部。分布情况表明长三角植被NPP变化长期受到气候因素的正向影响,并且气候变化带来的植被NPP变化呈现缓慢增加的态势。如图7b所示,2001—2023年长三角地区植被NPP H 年变化的变化趋势在-36.13~34.37 g/(m2·a),平均值为1.25 g/(m2·a),变化趋势呈现增加趋势的面积比例为68.97%,主要分布在安徽省北部、安徽省中部、江苏省北部、江苏省南部等区域,变化趋势呈现减少趋势的区域主要分布在安徽省西南部、江苏省东南部、上海市与浙江省西南部。分布情况表明长三角植被NPP变化长期受到人为因素的正向影响,并且人为因素带来的植被NPP变化呈现快速增加的态势。长三角地区植被NPP变化受到气候因素与人为因素的综合影响,在两者的综合作用下不断增长,趋势与NPP Theil-Sen趋势分析结果相同。气候因子的影响较人为因素更为显著。
2.5 植被NPP变化驱动因素分析
2.5.1 机器学习建模
NPP的变化与NPP C 和NPP H 因素密切相关,NPP C 和NPP H 分别受到不同生态环境和人类活动因素的综合影响。为探究深层次生态环境和人类活动对植被NPP的非线性复杂影响关系,以NPP取值的变化量作为因变量Y,以11个生态环境和人类活动指标为自变量X,利用机器学习算法构建回归模型。使用的机器学习算法包括决策树、随机森林、XGBoost、支持向量机,基于Python平台构建模型,将指标的80%作为训练模型的训练样本,剩余20%作为测试样本。计算测试样本的精度,分析精度的实际情况与模型的参数情况计算因子重要性。对各类机器学习算法在本研究集中的精度进行对比,模型模拟结果详见表3。从统计学角度来看,4种评价指标的性能均符合模型假设条件,表明机器学习模型对这些数据具有较好的适用性。其中,随机森林回归模型在训练集与测试集上均展现出良好的预测效能,优于其他机器学习模型,整体泛化能力较强,是最适合模拟各因素对植被NPP影响分析的模型。具体而言,随机森林模型训练集的预测效果略优于测试集,训练集的决定系数(R²)为0.957,测试集为0.723;R²差值为0.234。训练集的RMSE为28.056,测试集的RMSE为72.385;RMSE差值为44.329。
(3) 2001—2023年长三角地区植被NPP C 和NPP H 年均值均呈上升趋势,NPPC由2001年的513.24 g/(m2·a)上升至2023年的558.19 g/(m2·a),NPP H 由2001年的10.11 g/(m2·a)上升至2023年的35.19 g/(m2·a)。长三角地区植被NPP在人类活动和气候因子的共同影响下不断升高,气候因子对NPP升高的影响更为显著。
LinBoqiang, LiuXiying. China’s carbon dioxide emissions under the urbanization process: Influence factors and abatement policies [J]. Economic Research Journal, 2010,45(8):66-78.
WangFeng, WuLihua, YangChao. Driving factors for growth of carbon dioxide emissions during economic development in China [J]. Economic Research Journal, 2010,45(2):123-136.
[5]
ShiQinren, ZhengBo, ZhengYixuan, et al. Co-benefits of CO2 emission reduction from China’s clean air actions between 2013—2020 [J]. Nature Communications, 2022,13:5061.
[6]
SongJian, WanShiqiang, ShilongPiao, et al.Global change and China’s terrestrial carbon sink: A quantitative review of 30 years’ ecosystem manipulative experiments [J]. Ecological Monographs, 2025,95(1):e70005.
[7]
MuWenbin, ZhuXingyuan, MaWeixi, et al. Impact assessment of urbanization on vegetation net primary productivity: A case study of the core development area in central plains urban agglomeration, China[J]. Environmental Research, 2023,229:115995.
TaoBo, LiKerang, ShaoXuemei, et al. Temporal and spatial pattern of net primary production of terrestrial ecosystems in China [J]. Acta Geographica Sinica, 2003,58(3):372-380.
[10]
SongShu, NiuJun, SinghS K, et al. Projection of net primary production under changing environment in Xinjiang using an improved wCASA model [J]. Journal of Hydrology, 2023,620:129314.
ChaoQingchen, ZhangYongxiang, GaoXiang, et al. Paris agreement: A new start for global governance on climate [J]. Climate Change Research, 2016,12(1):61-67.
[13]
ZhangJunping, WangJia, ChenYuhan, et al. Spatiotemporal variation and prediction of NPP in Beijing-Tianjin-Hebei region by coupling PLUS and CASA models [J]. Ecological Informatics, 2024,81:102620.
[14]
JiaLu, YuKunxia, LiZhanbin, et al. Spatiotemporal pattern of NPP and its response to climatic factors in the Yangtze River economic belt [J]. Ecological Indicators, 2024,162:112017.
[15]
WangMingrui, SunKun, JiaJunjie, et al. Climate change drove the decline in Yangtze Estuary net primary production over the past two decades [J]. Environmental Science & Technology, 2024,58(43):19305-19314.
[16]
BaiWenqian, HeLi, HeZhengwei, et al. Analysis of spatial and temporal variation of vegetation NPP in Daning River basin and its driving forces [J]. International Journal of Remote Sensing, 2023,44(19):6194-6218.
[17]
MaZhuoran, WuJianjun, YangHuicai, et al. Assessment of vegetation net primary productivity variation and influencing factors in the Beijing-Tianjin-Hebei region [J]. Journal of Environmental Management, 2024,365:121490.
LiQianqian, YangYanrong, ZhangWenyi, et al. Effects of forest management and drought disturbance on forest net primary productivity in three northeastern provinces from 2000 to 2022 [J]. Bulletin of Soil and Water Conservation, 2024,44(6):308-317.
ZhouYanyan, ZhuMinxiang, GuoXiaojuan, et al. Relative effects of climate change and human activities on net primary productivity in Shule River basin [J]. Acta Ecologica Sinica, 2019,39(14):5127-5137.
[22]
WangZiyu, ZhouYoufeng, SunXinyu, et al. Estimation of NPP in Huangshan district based on deep learning and CASA model [J]. Forests, 2024,15(8):1467.
[23]
ChenAn, ZhongXuzhen, WangJinliang, et al. Spatiotemporal patterns and driving forces of net primary productivity in south and southeast Asia based on Google Earth Engine and MODIS data [J]. Catena, 2025,249:108689.
[24]
FernandesR, LeblancS G. Parametric (modified least squares) and non-parametric (Theil-Sen) linear regressions for predicting biophysical parameters in the presence of measurement errors [J]. Remote Sensing of Environment, 2005,95(3):303-316.
[25]
LiuYanxu, WangYanglin, PengJian, et al. Correlations between urbanization and vegetation degradation across the world’s metropolises using DMSP/OLS nighttime light data [J]. Remote Sensing, 2015,7(2):2067-2088.
JiangTianhan, DengLiantang. Some problems in estimating a Hurst exponent: A case study of applicatings to climatic change [J]. Geographical Science, 2004,24(2):177-182.
YanMin, LiZengyuan, TianXin, et al. Remote sensing estimation of gross primary productivity and its response to climate change in the upstream of Heihe River basin [J]. Chinese Journal of Plant Ecology, 2016,40(1):1-12.
YiLang, RenZhiyuan, ZhangChong, et al. Vegetation cover, climate and human activities on the Loess Plateau [J]. Resources Science, 2014,36(1):166-174.
[32]
PalM. Random forest classifier for remote sensing classification [J]. International Journal of Remote Sensing, 2005,26(1):217-222.
[33]
BelgiuM, DrăguţL. Random forest in remote sensing: A review of applications and future directions [J]. ISPRS Journal of Photogrammetry and Remote Sensing, 2016,114:24-31.
[34]
ChenHao, YangNi, SongXuanhua, et al. A novel agricultural drought index based on multi-source remote sensing data and interpretable machine learning [J]. Agricultural Water Management, 2025,308:109303.
[35]
XuLichang, NingShaowei, XuXiaoyan, et al. Comparative analysis of machine learning models and explainable AI for agriculture drought prediction: A case study of the Ta-pieh Mountains [J]. Agricultural Water Management, 2024,306:109176.
FangHe, FanGaofeng, WangKuo, et al. Spatiotemporal variation characteristics and driving factors of vegetation NPP in the Yangtze River Delta from 2000 to 2022 [J]. Environmental Science, 2025,46(9):5788-5799.
FangHe, WangKuo, LiZhengquan, et al. Dynamic characteristics of vegetation ecological quality in the Yangtze River Delta[J].China Environmental Science,2025,45(8):4513-4523.
LiuZequn. Estimation of vegetation net primary productivity in coastal cities of Yangtze River Delta based on multi-source data fusion [D]. Nanjing, Jangsu: Nanjing Normal University, 2021.
ZhangYanting, TangDiwei, TaoWen, et al. Spatiotemporal variation and influencing factors of vegetation gross primary productivity in the Yangtze River Delta urban agglomeration [J]. Journal of Hubei MinZu University (Natural Sciences Edition), 2025,43(4):593-600.