Sub-Saharan West Africa Land Cover Change

Overview

Development stage

In development or acquisitionSource: a) Pre-deployment – The use case is in a development or acquisition status.

High-impact designation

Not high-impactSource: c) Not high-impact

Impact justification

Not reported

Start date

Not reported

Withheld from public reporting

NoSource: a) No

Mission

Topic area

Science

Operational functions

Image and video analysis Derived by keyword rules; see methodology

Problem the AI is intended to solve

In recent decades, Sub-Saharan West Africa has seen rapid and ongoing land cover change fueled by population growth and subsequent agricultural expansion and intensification. These changes have led to negative impacts including a decrease in land productivity, loss of local biodiversity, and a general degradation of ecosystem services, resulting in debate over whether policies to discourage this type of transformation should be introduced. However, moderate resolution satellite data and traditional remote sensing methods are insufficient at resolving land cover land use change in this region, which is dominated by small, dispersed patches of savanna-woodlands and smallholder agriculture systems (< 3 ha.) that consist of highly dynamic and often ill-defined field boundaries. In Senegal, extreme latitudinal gradients in phenology, limited availability of cloud-free wet season imagery, and widespread burnt area during the dry season add further complexity in identifying sub-hectare land cover and change.

Expected benefits

Thanks to the growing availability of very high resolution (VHR) imagery (< 3 m GSD) through commercial vendors and the increased accessibility of high-performance computing resources such as GPUs, we are now able to perform computationally-expensive, deep learning-based predictions on thousands of VHR observations for mapping fine-scale land cover over large areas. We have leveraged these enhanced capabilities by developing a series of deep learning models for land cover classification with WorldView 8-band imagery (2 m GSD), and have performed inference on all data available over the study domain. To assess the cost-benefit of this effort, we implemented a simple spatial resolution experiment at select locations in Senegal by pansharpening and resampling 2 m WorldView multispectral imagery and training data to alternate spatial resolutions (0.5 m, 5 m, 10 m, and 30 m) for training and inference using our deep learning models.

System outputs

Presented here, our quantitative results evaluating the impact of spatial resolution on the accuracy of mapping agricultural expansion and tree/shrub cover in Senegal provide insight into the optimal input parameters for mapping land cover with deep learning applications.

Technology

AI classification

Classical / predictive MLSource: Classical/Predictive Machine Learning: Models trained on data to make predictions or classifications based on identified patterns or relatio…

System name(s)

Not reported

Custom-developed code

Not reported

Public source code

Not reported

Sourcing

How it was built

Not reported

Vendor (as reported)

Not reported

Vendors (standardized)

None on the standard list

A vendor is the supplier named by the agency. It does not identify the underlying model or AI technology.

Data and privacy

Involves PII

Not reported

Privacy Impact Assessment

Not reported

Authorization to Operate

Not reported

Demographic features

Not reported

Training and evaluation data

Not reported

Federal Data Catalog

Not reported

Governance

0 of 8 minimum-practice questions answered. A blank answer means the agency reported nothing; it does not mean the practice is absent.

Potential impacts and how they were identified

Not reported