Feasibility of an Early Warning Tool for Locust Outbreaks in East Africa

自2019年以来，东非国家一直面临着历史上最大的蝗虫疫情，破坏了数千公里的农田，大幅降低了作物产量。这种病虫害的爆发威胁着数百万人的粮食安全和生计。非洲小规模农户特别脆弱，因为他们主要依靠作物生产维持生计和获得粮食，此外，预计这种生态冲击的强度和频率将随着气候变化而增加，导致农业损失激增。这不仅将对非洲最贫困人口的粮食安全和营养健康造成严重后果，也将对全世界人民的粮食安全和营养健康造成严重后果。利用现有的地球观测数据和智能模型，可以预测包括蝗虫发生在内的农作物病虫害爆发。我们在Dtime计划开发一个移动的和基于网络的预警工具，除了深入了解农作物病虫害分布和生态之外，还将利用大数据和尖端机器学习算法提供农作物病虫害爆发的预警。该工具将向农民、农业官员、粮农组织等非政府组织、政府机构和价值链中的其他利益攸关方通报蝗虫出现和到达的概率、移动模式以及局部尺度（1- 10公里）和景观尺度（100- 500公里）的爆发情况。农民和其他利益相关者将能够预测和适应，以防止作物损失。该工具将帮助粮农组织和其他组织更有效地管理蝗虫，从而减少农药的使用，从而最大限度地降低虫害控制成本和环境影响。我们的团队正在使用卷积神经网络和随机森林分类等算法开发一系列智能预测模型。在这个阶段，我们专注于为东非建立模型。我们将使用蝗虫模型和数据管道来模拟其他毁灭性的害虫（例如，秋粘虫）和疾病（例如，我们将与保险部门、价值链利益相关者、商业农民和广告商合作，为该工具提供资金，因为我们希望该工具对小规模农民完全免费。我们还计划研究来自政府和粮农组织等区域组织的资金选择，因为该工具将显著降低粮农组织和世界银行的病虫害控制成本，2020年蝗虫控制成本超过600万美元。

Since 2019, countries in East Africa have been facing the biggest locust outbreaks in history, which damaged thousands of kilometres of cropland and drastically reduced crop yields. Such pest and disease outbreaks are threatening food security and livelihoods of millions of people. African small-scale farmers are particularly vulnerable as they are largely dependent on crop production for livelihoods and food.Besides, the intensity and frequency of such ecological shocks are projected to increase with climate change, resulting in an upsurge in agricultural losses. This will have severe consequences for food security and nutritional health for not only the poorest in Africa but also for people across the world.Crop pest and disease outbreaks including Locust occurrences can be predicted with the help of the existing earth observation data and intelligent models. We, at Dtime, plan to develop a mobile and web-based Early-Warning Tool that will provide warning of crop pest and disease outbreaks using big data and cutting-edge machine learning algorithms, in addition to a deep understanding of crop pest/disease distribution and ecology. This tool will inform farmers, agriculture officers, non-government organisations such as FAO, government bodies and other stakeholders in the value chain about the probability of locust emergence and arrival, movement patterns, and outbreak at local-scales (1-10km) and landscape scales (100-500km). Farmers and other stakeholders will then be able to anticipate and adapt in order to prevent crop losses. The tool will help FAO and other organisations manage locusts much more effectively resulting in reduced use of pesticides, hence minimising the costs of pest control and environmental impact.Our team is developing a range of intelligent forecast models using algorithms such as Convolutional Neural Networks and Random Forest Classification. At this stage, we focus on building models for East Africa. We will use the locust models and data pipelines to model other devastating pests (e.g., Fall armyworm) and diseases (e.g., Cassava mosaic disease) in east Africa and across the world.We will engage with the insurance sector, value chain stakeholders, commercial farmers and advertisements to fund the tool, as we want this tool to be entirely free for small-scale farmers. We also plan to look at funding options from governments and regional organisations such as the FAO as the tool will significantly reduce the costs of pest and disease control by the FAO and World bank which was over $ 6 million for Locust control in 2020\.