About UNCCD
The United Nations Convention to Combat Desertification is the only legally binding framework set up to address desertification and the effects of drought. There are 197 Parties to the Convention, including 196 country Parties and the European Union. The Convention – based on the principles of participation, partnership and decentralization – is a multilateral commitment to mitigate the impact of land degradation, and protect our land so we can provide food, water, shelter and economic opportunity to all people.Activities by UNCCD
Imagine 2040: An AI Crafted film Project
Africa, Asia and the Pacific, Europe and the Commonwealth of Independent States, Latin America and the Caribbean BrazilChinaSenegalUnited StatesAn AI-powered short film produced to inspire action on land restoration and climate resilience. The film merges youth storytelling, AI-generated visuals, and real-world conservation narratives. It is a proof of concept for using AI in sustainability storytelling while advocating for global land restoration policies.
Activity Type Awareness/AdvocacyScientific Report – Detection and mapping of abandoned mine sites
Africa LesothoSouth Africa
This pilot study uses AI-driven deep learning models and freely available multispectral satellite imagery to systematically monitor mining sites. The objective is to enhance large-scale predictions of surface mines and tailings dams, supporting environmental management and sustainable land use policies.
The trained AI model demonstrated strong predictive capabilities, correctly identifying 87% of mine points. However, it misclassified 5% of non-mine areas, necessitating manual corrections and highlighting the need for further refinement to reduce recurrent misinterpretations (e.g., agricultural areas, roads, rivers, and steep slopes). Transitioning from image classification to object detection (e.g., UNET models) could improve surface mining detection by identifying individual mines and tailings dams rather than classifying entire images. Additionally, separating mines from tailings dams in training data would further enhance the model’s precision. Incorporating training data from Australia—due to its similar landscapes, climate zones, and comprehensive mine datasets—could refine the AI’s detection capabilities. Additionally, using high-resolution satellite images like Sentinel-2 instead of lower-resolution Landsat images would enhance prediction accuracy, enabling better long-term monitoring of mining activities worldwide.