Monitoring Landslide Disturbance and Forest Recovery in Afromontane Landscapes: The Case of Nyungwe Forest Reserve, Rwanda

Abstract

This study examines landslide occurrence and post‑landslide recovery dynamics in Nyungwe Forest Reserve, Rwanda, using a combination of multi‑temporal landslide mapping, environmental and geomorphological predictor analysis, and an assessment of the Tropical Moist Forest (TMF) product. A landslide inventory covering the period from 2000 to 2024 was produced using high‑resolution optical imagery. This inventory represents the most complete record of landslide activity in Nyungwe so far and captures both the spatial distribution of failures and how their frequency has varied over nearly twenty‑five years.
Using Frequency Ratio (FR) and Binary Logistic Regression (BLR), the study found that steep slopes, rainfall, proximity to roads, profile curvature and exposure of slopes to the east are the main factors linked to landslide occurrence within the reserve. Together, the two approaches provided complementary insights: FR highlighted where landslides tend to cluster within specific variable classes, while BLR identified which predictors remain significant when considered simultaneously.
The evaluation of the TMF product showed that it can detect medium‑to‑large landslides reliably, but often misses small, localised failures. This is mainly due to Landsat’s 30-meter spatial resolution, temporal rules in the TMF classification system, cloud‑related gaps, and spectral confusion. Most disturbance trajectories were still ongoing by 2024, meaning many were right‑censored. Only a small number of landslides showed full disturbance‑to‑regrowth cycles, while others showed partial canopy recovery. Of note are the regrowth patterns in the reserve, with early vegetation return potentially mostly dominated by pioneer tree species and lianas which can alter long‑term recovery pathways.
Overall, this study demonstrates the value of combining landslide mapping, statistical modelling, and long‑term satellite‑derived forest disturbance data to understand forest ecosystems in tropical montane environments. The findings also highlight the limitations of medium‑resolution forest monitoring products for detecting small landslides and emphasise the need for future research that incorporates higher‑resolution optical and radar data, such as Sentinel‑2 and Sentinel‑1, to improve detection accuracy and better track post‑disturbance recovery.

Popular Abstract

Nyungwe Forest Reserve in Rwanda is one of Africa’s major mountain forests, characterised by steep terrain and rich wildlife. The reserve is highly prone to landslides, known to reshape landscapes and disrupt forest ecosystems. This research set out to understand landslide patterns and forest regrowth in the reserve.
Google Earth satellite images from 2000 to 2024 were used to map landslides across the reserve. This produced the most complete record to date of landslide activity, showing where failures occurred and how their frequency changed over time. Environmental factors were then examined to identify the conditions most strongly linked to landslide occurrence. Slope steepness, rainfall, proximity to roads, profile curvature and slope direction showed the greatest influence.
The research also assessed how well a global forest‑monitoring dataset, the Tropical Moist Forest (TMF) product, can detect landslides in this complex landscape. TMF uses long‑term satellite observations to track forest disturbance and recovery. Medium and large landslides were detected more reliably, while smaller, localised failures were often missed due to the coarse resolution of the imagery, cloud cover and the way forest change is classified.
TMF disturbance timelines were used to explore how the forest begins to recover after a landslide. Most scars had not fully recovered by 2024, suggesting that regrowth is still underway. Only a few showed a complete cycle from disturbance to regrowth, while many displayed partial recovery. Vegetation recovery within the reserve is likely influenced by dominant fast‑growing tree species and invasive but native lianas, which can shape long‑term forest structure and affect how quickly the ecosystem returns to mature forest.
By combining landslide mapping with long‑term satellite‑based forest monitoring, the research provides new insights into the drivers of landslides and the resilience of tropical mountain forests. The findings also highlight the need to incorporate higher‑resolution optical and radar satellite data, such as Sentinel‑1 and Sentinel‑2, to improve the detection of small landslides and to monitor forest recovery more accurately.

• author: Kintu, Ingrid Martha
• supervisor: Thomas Pugh
• organization: Department of Earth and Environmental Sciences (MGeo)
• course: GISM01 20261
• year: 2026
• type: H2 - Master's Degree (Two Years)
• subject: Earth and Environmental Sciences
• keywords: Geography, GIS, Remote sensing, Landslides, Nyungwe Forest Reserve, Tropical Moist Forest (TMF) product
• publication/series: Master Thesis in Geographic Information Science
• report number: 216
• language: English
• id: 9242813
• date added to LUP: 2026-06-23 08:57:16
• date last changed: 2026-06-23 08:57:16