📄 earth_science

Monitoring Forest Landscape Restoration Success Using Sentinel-2 NDVI and Field Measurements in Timor-Leste

This study demonstrates that integrating Sentinel-2 NDVI data with field measurements effectively quantifies the successful vegetation recovery of a 148-hectare Forest Landscape Restoration site in Timor-Leste, providing a scalable framework for monitoring restoration outcomes in tropical drylands.

Original authors: Albino da Silva Barbosa, Adalfredo do Rosario Ferreira, José Ronaldo Oqui Ronaldo Oqui Fernandes, Bartolomeu de Jesus Soares, Adelino Rojario, Pramudya Bumishambara

Published 2026-06-24

📖 4 min read☕ Coffee break read

CC BY 4.0

Original authors: Albino da Silva Barbosa, Adalfredo do Rosario Ferreira, José Ronaldo Oqui Ronaldo Oqui Fernandes, Bartolomeu de Jesus Soares, Adelino Rojario, Pramudya Bumishambara

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Big Picture: Healing a "Bald" Landscape

Imagine a patch of land in Timor-Leste that is like a bald spot on a head. For a long time, this 148-hectare area (called Balak) was stripped of its trees and plants due to farming, grazing, and fires. The soil was exposed, rocky, and vulnerable to erosion. It was a "degraded" landscape that couldn't support much life.

To fix this, a team planted trees and started a "Forest Landscape Restoration" project. The big question was: Did it actually work?

This paper is the report card on that project. The researchers wanted to see if the land was truly getting its "hair" (vegetation) back, or if the trees were just struggling to survive.

The Two-Part Detective Work

To get a clear answer, the researchers used two different tools, like a detective using both a satellite photo and a magnifying glass.

1. The Satellite "Bird's-Eye View" (Sentinel-2 & NDVI)
Instead of walking every inch of the land, the team used a satellite called Sentinel-2. Think of this satellite as a giant camera in space that takes pictures of the Earth every few days.

They used a special filter called NDVI (Normalized Difference Vegetation Index). You can think of NDVI as a "greenness meter."

Low NDVI: The land looks brown or gray (bare soil, rocks).
High NDVI: The land looks lush and green (healthy plants).

They took a picture in May 2022 (before the trees were fully established) and another in May 2026 (after four years of growth). By comparing the two, they could measure exactly how much the "greenness" increased across the whole area.

2. The Ground-Level "Magnifying Glass" (Field Measurements)
Satellites are great for big pictures, but they can't see small details. So, the researchers also went to the ground. They set up 29 specific squares (like small garden plots) across the forest.

In these squares, they counted:

How many trees were still alive? (Survival rate)
How tall were the trees? (Growth)

This was their way of checking if the "green" seen from space was actually real, healthy trees, or just a trick of the light.

What Did They Find?

The results were very positive. Here is the breakdown:

The "Greenness" Jumped: The average "greenness score" (NDVI) went from 0.288 in 2022 to 0.475 in 2026. That is a 65% increase.
- Analogy: Imagine a dimly lit room where the lights were turned up significantly. The whole area became much brighter and greener.
Most of the Land Recovered: About 77.6% of the land showed moderate to significant improvement. Only a tiny sliver (2.3%) got worse.
The Trees on the Ground: When they checked the 29 plots, 62.1% were considered "high recovery" (meaning most trees survived and grew well). Only 6.9% were "low recovery."
Height Matters: They found a connection between how tall the trees grew and how many survived. It's like a sports team: the teams with taller, stronger players tended to have more players on the field. Taller trees generally meant a more successful planting spot.

Why Was This Hard?

The paper notes that this wasn't an easy place to grow trees. It's a tropical dryland with:

Long, hot dry seasons.
Shallow, rocky soil.
Frequent droughts.

Usually, planting trees here is like trying to grow a garden in a desert with a leaky hose. Many projects fail because the seedlings die. But in this case, the combination of choosing the right tree species (like Casuarina junghuhniana, which is tough and loves dry soil) and good management helped the trees survive.

The Main Takeaway

The study proves that you can heal a "bald" landscape even in tough, dry conditions if you use the right methods.

The researchers also showed that mixing satellite photos with ground checks is a smart, cheap, and effective way to monitor these projects. You don't need to hire an army of people to walk every inch of the forest; you can use satellites to see the big picture and spot-check a few areas on the ground to confirm the results.

In short: The trees planted in Balak, Timor-Leste, are growing well, the land is turning green again, and the method used to measure this success works perfectly for places where data is hard to find.

Technical Summary: Monitoring Forest Landscape Restoration Success in Timor-Leste

Problem Statement
Forest Landscape Restoration (FLR) is a critical strategy for reversing land degradation in tropical drylands, yet quantitative evaluations of restoration effectiveness remain scarce in Timor-Leste. The country faces severe ecological challenges, including steep terrain, a dry monsoonal climate, and persistent human disturbances such as shifting cultivation and uncontrolled grazing. These factors have left extensive areas devoid of plant cover, characterized by poor soil fertility and high erosion susceptibility. While international donor-supported initiatives, such as the AFoCO-funded "Re-greening the Bare Lands" project, have implemented large-scale reforestation, there is a lack of rigorous, outcome-based assessments that integrate satellite monitoring with field observations. Existing evaluations often focus on implementation metrics rather than measurable vegetation recovery, making it difficult to determine if restored areas are progressing toward functional ecosystems, particularly in harsh environments where seedling mortality is high.

Methodology
This study employed an integrated "before-and-after" assessment framework to evaluate vegetation recovery in a 148-hectare FLR site in Balak, Manatuto, Timor-Leste. The research combined multi-temporal remote sensing with ground-based ecological measurements:

Remote Sensing: Sentinel-2 Level-2A imagery from May 2022 (pre-restoration) and May 2026 (post-restoration) was utilized. The Normalized Difference Vegetation Index (NDVI) was calculated using the Near Infrared (B8) and Red (B4) bands. Vegetation change was quantified via NDVI differencing ( $\Delta$ NDVI = NDVI $_{2026}$ - NDVI $_{2022}$ ). NDVI values were classified into five vegetation cover categories, and $\Delta$ NDVI values were categorized into five ecological recovery classes ranging from "Severe Vegetation Loss" to "Significant Vegetation Recovery."
Field Measurements: In 2025, 29 systematic monitoring plots (20 m × 20 m) were surveyed. Data collected included the total number of planted seedlings, surviving seedlings, survival rates, and average plant height.
Statistical Analysis: Pearson correlation and linear regression analyses were conducted using R to examine the relationship between vegetation structural development (average plant height) and restoration performance (seedling survival rate).

Key Results
The study found substantial evidence of vegetation recovery across the restoration site:

NDVI Trends: The mean NDVI increased from 0.288 in 2022 to 0.475 in 2026, representing a 65% increase in vegetation greenness. Spatial analysis revealed a shift from low vegetation cover (dominant in 2022) to moderate and high vegetation cover classes by 2026.
Recovery Classification: Based on $\Delta$ NDVI analysis, 77.6% of the site experienced moderate to significant vegetation recovery. Only 2.3% of the area showed vegetation decline, and less than 0.1% experienced severe loss.
Field Performance: Of the 29 monitoring plots, 62.1% were classified as "High Recovery" (survival rate >80%), and 31.0% as "Moderate Recovery" (50–80%). Only 6.9% were classified as "Low Recovery."
Species Performance: Casuarina junghuhniana demonstrated the highest establishment success, with survival rates often exceeding 90% and average heights between 1.1 and 3.1 meters. In contrast, Anacardium occidentale experienced complete establishment failure in the tested plots.
Statistical Correlation: A significant positive relationship was observed between average plant height and seedling survival rate ( $r = 0.423$ , $p = 0.022$ ). However, the linear regression model explained only 17.9% of the variation in survival rates ( $R^2 = 0.179$ ), indicating that factors beyond plant height, such as soil depth and micro-topography, significantly influence outcomes.

Significance and Claims
The paper asserts that the integration of Sentinel-2-derived NDVI indicators with field measurements provides a practical, cost-effective, and scalable framework for monitoring FLR outcomes in data-limited regions. The findings offer empirical evidence that the AFoCO-supported reforestation program has effectively accelerated vegetation establishment and improved ecosystem conditions within a degraded tropical dryland landscape in a relatively short timeframe (four years).

The study claims that the observed positive NDVI trajectory reflects genuine ecological improvement rather than short-term fluctuations, aligning with international evidence from similar dryland restoration programs in China, Ethiopia, and the Sahel. The authors conclude that while NDVI is a robust indicator of vegetation greenness and canopy development, it should be interpreted as a measure of vegetation recovery rather than a comprehensive assessment of biodiversity or ecosystem functionality. The research supports the continued implementation of FLR strategies in Timor-Leste and highlights the value of combining remote sensing with ground truthing to support restoration planning and policy development.

Limitations Noted by the Authors
The authors acknowledge several limitations:

NDVI primarily reflects vegetation greenness and may saturate under dense conditions; it does not directly measure biodiversity, species composition, or soil conditions.
Field data were collected in 2025, one year prior to the final satellite observation in 2026, creating a minor temporal gap.
The study lacked an untreated control site to fully isolate restoration effects from interannual environmental variability.
Future assessments would benefit from longer monitoring periods, control sites, and the inclusion of biodiversity and soil indicators.