Forest Resilience, Precipitation, and Ecosystem Service Value: A Correlation and Trend Analysis

Laure Berti‐Équille, Pius Nnamdi NwachukwuGlobal 2025 · DOI

TL;DR

This study investigates the complex relationship between forest health, rainfall, and economic value (ESV) across the globe from 2000 to 2024. By leveraging machine learning and Earth Observation data, the researchers found that while vegetation health and precipitation are closely linked in regions like Asia and North America, the economic valuation of these ecosystems remains weakly correlated with environmental metrics, suggesting that policy and human factors play a dominant role in how we value nature.

What Problem Does This Paper Address?

Ecosystem Service Valuation (ESV) is the process of assigning monetary value to the benefits humans derive from nature—such as carbon sequestration, flood control, and biodiversity. While forests cover 30% of the Earth's land and absorb over a third of anthropogenic carbon, they are under increasing stress from climate change, wildfires, and human depletion.

The challenge lies in the "valuation gap." We know forests are ecologically resilient (or struggling), but how does that physical reality translate into economic data? Traditional models often fail to account for the time-dependent nature of climate variability. This paper seeks to bridge that gap by using remote sensing indicators to see if forest resilience and precipitation patterns actually drive the trends seen in global ESV databases.

Methods

The researchers employed a multi-step data science pipeline:

  1. Proxy Selection: They used kNDVI (Kernel Normalized Difference Vegetation Index) as a proxy for forest resilience and health. kNDVI is considered more robust than traditional NDVI for monitoring dense forest canopies.
  2. Correlation Analysis: They performed Pearson correlation and p-value analysis to identify relationships between precipitation, kNDVI, and ESV across different continents and forest biomes.
  3. Machine Learning: To predict ESV trends, they compared several models:
  4. Random Forest (RF)
    • XGBoost
    • Gradient Boosting
    • Stacking Ensemble: A meta-model that combines the predictions of RF and XGBoost.

Datasets

The study integrated three primary data sources spanning 2000 to 2024:

  • MOD13A3 V6.1 (MODIS): Monthly vegetation data at 1 km resolution used to compute kNDVI.
  • CHIRPS: The Climate Hazards Group InfraRed Precipitation with Station data, providing global monthly precipitation means at a 0.05° resolution.
  • ESVD (Ecosystem Services Valuation Database): A comprehensive database of economic valuation studies. The authors filtered this to focus on 2,801 specific forest sites.

Key Insights and Ideas

  • The Ecological-Economic Disconnect: One of the most striking "aha moments" is that while kNDVI and precipitation are strongly correlated in many regions (like Asia at +0.76 and North America at +0.70), ESV shows very weak or even negative correlations with these environmental factors. This suggests that the "value" assigned to a forest in economic databases is often detached from its immediate biological health or the local climate.
  • Policy as a Driver: The researchers noted a significant spike in ESV in Europe between 2014 and 2018. Since this didn't align with a massive ecological shift, they attribute it to potential policy changes and increased environmental awareness in the region.
  • Regional Variability: Forest dynamics are not uniform. For instance, correlations between vegetation and rain were strong in Asia and North America but actually negative in Oceania (-0.41), highlighting the complexity of different forest biomes.
  • Ensemble Superiority: For predicting ESV, the stacking ensemble model outperformed single-algorithm approaches, suggesting that the non-linear relationships in ecosystem valuation require more complex, layered modeling techniques.

Key Results

  • Best Model: The Stacking Ensemble (RF + XGBoost) achieved the best performance for ESV prediction with a Mean Absolute Error (MAE) of .0047 and an R² score of .2082.
  • Vegetation vs. Rain: Strongest correlations for kNDVI and precipitation were found in Asia (0.71) and North America (0.76).
  • ESV Trends: Asia saw a high peak in valuation in 2019, while Africa and South America showed smaller, more sporadic peaks (e.g., Africa in 2002 and 2015).

Limitations and Future Work

The authors acknowledge that ESV is influenced by a myriad of factors beyond just greenness and rain. The current model's R² score (.2082) indicates that while the model is heading in the right direction, there is still a significant amount of variance in ecosystem valuation that remains unexplained.

Future directions include: * Granularity: Refining the spatial and temporal resolution of data collection. * Broader Indicators: Incorporating more environmental and socio-economic indicators (e.g., soil health, local legislation, or biodiversity indices) to improve model robustness. * Policy Application: Developing these models into tools that can directly assist policymakers in making data-driven conservation decisions.

Summary updated

Abstract

Ecosystem service value (ESV) is critical for understanding ecosystems' economic benefits and their responses to environmental change. This study uses Earth Observation (EO) data, statistics, and machine learning methods to evaluate ESV trends across multiple continents from 2000 to 2024. Three key datasets were used: MODIS NDVI for vegetation monitoring, CHIRPS precipitation data, and the ESVD database. Data preprocessing includes data cleaning, feature engineering, and outlier detection. We compared Random Forest, XGBoost, and ensemble stacking models to predict key variables and their relationships, such as kNDVI, as a proxy of forest resilience and ESV trends, precipitation patterns, and biome-specific variables. Our results show that kNDVI changes across continents reveal various patterns in vegetation dynamics that are in-line with precipitation patterns and are weakly correlated with ESV changes for forest biomes. Our study emphasizes the significance of time interdependence and climate variability in ESV predictive modelling. Future efforts focus on refining the time/space granularity of data collection and aggregation techniques and incorporating more environmental indicators to improve model robustness and application in policy-making.
BibTeX
@inproceedings{bertiequille2025forest,
  author = {Laure Berti‐Équille and Pius Nnamdi Nwachukwu},
  title = {{Forest Resilience, Precipitation, and Ecosystem Service Value: A Correlation and Trend Analysis}},
  booktitle = {Global},
  year = {2025},
  doi = {10.30955/gnc2025.00053},
}
Figure