Desertification Detection with Deep Learning and Satellite Data

Local Chapter Gaborone, Botswana Chapter

Coordinated byBotswana ,

Status: Completed

Project Duration: 29 May 2022 - 09 Aug 2022

Open Source resources available from this project

Project background.

In terms of desertification, Botswana is one of the most seriously affected countries in Southern Africa.

Land provides valuable ecosystem services for human well-being, but land degradation leads to a reduction in the provision of these services with significant social and economic costs to the country. The decline of ecosystem services can take different forms, including decline in food availability, soil fertility, carbon sequestration capacity, wood production, groundwater recharge, among others

Land degradation is the result of human-induced actions which exploit land, causing its utility, biodiversity, soil fertility, and overall health to decline as explained by UNCCD. It is simply the deterioration of the economic productivity of the land – such as the ability to farm the land for commercial or subsistence purposes.

Although it takes a long time to restore degraded land, it is essential to start now. A recent global assessment on land degradation shows that for Botswana the returns on taking action against land degradation versus inaction are estimated at 6 USD for every dollar invested in reverting degraded land,'” underlining the economic solid incentives for bold actions on achieving LDN

The problem.

In Botswana, the total annual cost of land degradation is estimated at 353 million United States Dollars (USD) — this is equal to 3.2% of the country’s Gross Domestic Product (GDP). 
Desertification increases the likelihood of droughts and prolonged dry periods and will increase soil erosion

Project goals.

AI has proven to provide more and more accurate forecast results in recent years, allowing the formulation of solutions in a faster and more agile way than before. Here we’ll work to harvest this technological advancement to help predict the areas and regions that could fall victim to desertification in the upcoming years in Botswana.That is why for 4 weeks** **our goal will be to produce a forecasting model to predict the status of different land covers in Botswana. This will include working on the following:- Collect free and publicly available Satellite data that covers Botswana over the years and in different seasons. - Using Supervised and Unsupervised learning algorithms to classify different land type covers. - Analyze the loss of green, degradation of lands in Botswana over the years (using NDVI, NDWI, and other indices), and build a forecast model based on that information. - Build a dashboard visualizing the areas affected and the future prediction using Streamlit or other freely available tools.

Project plan.

    Learning outcomes.

    1. Geospatial and Satellite data that will include learning the basics and advanced techniques to process satellite data.

    2. Implement supervised and unsupervised machine learning and deep learning algorithms to classify different land types.

    3. Building a dashboard to visualize and present our results in an impactful way.

    4. The project will include conducting multiple workshops on the above topics

    Share project on: