Continuous anomaly detection using remote sensing to monitor on-farm restoration in sub-Saharan Africa

Land degradation poses a significant threat to ecosystem health and food security, particularly in the global South. Given the severity of land degradation globally, land restoration is urgently needed to recover degraded ecosystems through, for example, tree planting and (farmer-managed) natural regeneration (FMNR). In this study we monitor the impacts of farmer-managed land restoration using satellite time series data through a Continuous Anomaly Detection after Intervention (CADI) approach. We also propose ways that this approach can be used to generate insights to help design future land restoration interventions. Data was collected for 127,782 restoration plots in seven sub-Saharan countries through the use of the “Regreening App”, which was designed for citizen science data collection. For each plot, a reference NDVI was modelled based on multiple years prior to restoration interventions which was compared to the actual NDVI to quantify the restoration impact. A comparison between our CADI approach and the residual trend (RESTREND) method was done based on the visual interpretation of 645 validation points. The CADI analysis proved better able to detect greening compared to RESTREND (F-score: 0.84 vs 0.79) and it performed better in arid regions (F-score: 0.88) than in dry sub-humid ecosystems (F-score: 0.75). FMNR was predominantly preferred in arid regions where higher greening was observed, indicating FMNR as a powerful and cost-effective option for future land restoration initiatives. To stimulate further use by policy makers and practitioners, the CADI analysis has been made available as an online tool here.

---

• View source

---

This work is licensed under CC-BY 4.0

---

DOI:

https://doi.org/10.1016/j.rsase.2025.101644

---

Altmetric score:

Dimensions Citation Count:

---

Export citation:

RIS (.ris)

Endnote (.ciw)