A global gross primary productivity dataset of sunlit and shaded leaves via combining two-leaf light use efficiency model with random forest from 2002 to 2020

The TL-CRF model generated a global 0.05´0.05° product for eight-day gross primary productivity (GPP) of sunlit and shaded canopies from 2002 to 2020 by embedding the random forest (RF) submodule into the two-leaf light use efficiency (TL-LUE) model while considering the seasonal differences in the clumping index. The RF technique was used to integrate various environmental stress factors including meteorological, hydrological, soil properties, and elevation, thereby improving the overall scale of the complex environmental conditions to the maximum LUE. Eight-day GPP was then aggregated into monthly, seasonal, and annual GPP. This novel GPP product could support further research on spatial and temporal patterns of the carbon cycle and its association with climate change.TL-CRF模型将随机森林（Random Forest, RF）子模块嵌入两叶光能利用率（Two-leaf Light Use Efficiency, TL-LUE）模型，同时考虑聚集指数（Clumping Index）的季节差异，由此生成了2002—2020年阳生冠层与阴生冠层的8天总初级生产力（Gross Primary Productivity, GPP）全球0.05°×0.05°产品。该研究借助RF技术整合了气象、水文、土壤属性及高程等多类环境胁迫因子，进而实现复杂环境条件的尺度上推以适配最大光能利用率（Light Use Efficiency, LUE）。随后将8天尺度的GPP数据聚合为月尺度、季节尺度及年际尺度的GPP产品。这款新型总初级生产力产品可为碳循环时空格局及其与气候变化的关联研究提供支撑。