A Deep Learning Approach to Estimate Gross Primary Productivity in Ethiopia

Understanding the impact of climate change on ecosystems is essential for effective environmental planning and resource management, particularly in sensitive tropical regions. This study explored the impact of climate change on Gross Primary Productivity (GPP), a key indicator of the carbon cycle, in a tropical region with high biodiversity and climate sensitivity in Ethiopia. By leveraging deep learning long short-term memory (LSTM) networks, we captured the complex nonlinear interactions between climate variables and vegetation productivity. We integrated historical climate data, satellite-derived GPP estimates, and projected the Coupled Model Intercomparison Project (CMIP6)-based Shared Socioeconomic Pathways (SSPs) and carbon dioxide (CO₂) emission scenarios (SSP1-2.6, SSP2-4.5, SSP3-7.0, SSP5-8.5) under six Earth System Models (ESMs). The growth of GPP per decade was expected to increase by 12.34 ± 8.01, 36.71 ± 15.79, 65.48 ± 21.61, and 89.79 ± 27.28 percent (%) under the SSP1-2.6, SSP2-4.5, SSP3-7.0, and SSP5-8.5 scenarios, respectively, by 2100, derived from the Ensemble Mean (MEM) of ESMs. The spatial distribution of GPP in the study area revealed higher values in the southwestern mid-latitude (7-11 degree) and lower values in the northeast high-latitude (11-15 degree) zones. The approach and findings of this study would respond to the anticipated climate change, putting into practice an integrated sustainable management plan to create a green economy in Ethiopia and contribute to the net-zero emissions and climate resilient mission of global sustainable development goals (SDGs).

探明气候变化对生态系统的影响，对于开展高效的环境规划与资源管理至关重要，在敏感热带区域尤为如此。本研究以埃塞俄比亚一处生物多样性丰富且气候敏感性强的热带区域为研究对象，探讨了气候变化对碳循环关键指标——总初级生产力（Gross Primary Productivity，GPP）的影响。本研究借助深度学习长短期记忆（long short-term memory，LSTM）网络，捕捉了气候变量与植被生产力之间复杂的非线性相互作用，整合了历史气候数据与卫星反演的GPP估算结果，并基于6种地球系统模式（Earth System Models，ESMs），对基于耦合模式比较计划（Coupled Model Intercomparison Project，CMIP6）的共享社会经济路径（Shared Socioeconomic Pathways，SSPs）及二氧化碳（CO₂）排放情景（SSP1-2.6、SSP2-4.5、SSP3-7.0、SSP5-8.5）进行了预估。通过地球系统模式的集合平均（Ensemble Mean，MEM）结果可知，到2100年，在SSP1-2.6、SSP2-4.5、SSP3-7.0及SSP5-8.5情景下，GPP的十年增长率预计将分别提升12.34±8.01、36.71±15.79、65.48±21.61及89.79±27.28个百分点。研究区GPP的空间分布特征显示，其高值区位于西南中纬度（7°~11°）区域，低值区则分布在东北高纬度（11°~15°）地带。本研究的方法与研究结果可为应对预期中的气候变化提供支撑，助力埃塞俄比亚落地一体化可持续管理方案以构建绿色经济，并为全球可持续发展目标（Sustainable Development Goals，SDGs）的净零排放与气候韧性使命作出贡献。