Construction of Reference Value Indicators for Low-Elevation Hillside Ecosystems in Central Taiwan

Abstract

Global environmental governance is transitioning toward a "Nature Positive" future. In response to the risks posed by biodiversity loss, directing capital into ecological conservation has become a prominent international trend. However, ecological restoration solely relying on carbon sink methodologies for carbon credit certification does not contribute to the enhancement of overall biodiversity. This two-year project aims to address this trend by being the first to introduce the internationally credible "Verra Nature Framework 1.0," focusing on the low-elevation hill conservation corridors in Central Taiwan. By integrating historical vegetation data with supplementary field surveys, and employing machine learning algorithms (such as GLM, Random Forest, etc.) combined with environmental variables like topography and climate, we will construct Taiwan's first visualized spatial prediction map of the "Best Reference Value (Rv)". The first year focuses on the Dadu Plateau for indicator construction and model training. The second year will extrapolate the model to complete the mapping for the four major low-elevation hill conservation corridors in Central Taiwan.
Expected outcomes include: (1)Constructing scientific forest ecological baselines and producing spatial mapping of the best reference states for low-elevation hills. (2)Establishing a quantitative measurement tool for "Nature Credits" aligned with international standards. (3)Facilitating corporate implementation of the TNFD framework and "Nature Positive" goals.
The results of this study can serve as a basis for the future matchmaking platform of the Forestry and Nature Conservation Agency, guiding corporate capital into areas of high ecological conservation value, thereby creating a win-win-win scenario for enterprises, ecosystems, and local development.

Methodology and Implementation Steps

This project is divided into three major phases:

1. Plot Surveys and Historical Data Inventory
   We will compile historical vegetation data from over 300 plots across the Dadu Plateau, Huoyanshan, Baguashan, Douliu Hills, Dakeng, and adjacent areas. Additionally, at least 30 supplementary field surveys will be conducted in areas with spatial data gaps to establish reference ecosystem datasets.
2. Quantification and Construction of Ecological Condition Indicators
   Following the guidelines of the Verra "Nature Framework 1.0," we will quantify the "Quality Hectares (Qha)" of the ecosystem, which includes the following 5 indicators:
   1. Composition Indicators (3 items):
      1. Species Richness: Calculating the number of native target species characteristic of the target ecosystem within a specified area, excluding invasive species.
      2. Stratification and Association Diversity: Assessing forest stratification and corresponding species composition across different life-form niches.
      3. Threatened Species Abundance: Identifying threatened plants by integrating buffer zone spatial data of occurrence points from the Taiwan Biodiversity Network (TBN) Red List.
   2. Structure Indicators (2 items):
      1. Canopy Cover: Estimating canopy cover percentages across different forest ages using remote sensing or field measurements.
      2. Stand Diameter Class Structure: Evaluating the number of trees across different size classes per unit area to reflect forest age structure and dynamics.
3. Spatial Predictive Mapping and Data Analysis
   By integrating plot indicator data with high-resolution environmental variables (such as topography, climate, and soil), we will run ensemble models using R packages (biomods2 or sdm), combining Generalized Linear Models (GLM), Random Forest, and Artificial Neural Networks. Models will be evaluated and weighted through repeated cross-validation. Ultimately, the point data will be converted into a comprehensive spatial prediction map of the "Best Reference Value ($Rv$)" for the entire region, which will be overlaid with existing forestlands to screen and identify potential investment hotspots with the highest net biodiversity outcomes.

Innovation and Cross-Disciplinary Collaboration

Closely aligned with the global "Nature Positive" and TNFD trends, this project is the first in Taiwan to introduce the "Verra Nature Framework 1.0". By integrating vegetation ecology, remote sensing, and spatial information science, and utilizing machine learning with environmental variables (GLM, Random Forest, etc.), we will construct the nation's first visualized, spatialized prediction map of the "Best Reference Value" for low-elevation hills. This addresses the challenge of measuring and standardizing ecological restoration effectiveness. The resulting visualization of potential areas for acquiring "Nature Credits" echoes TSMC’s Ecological Harmony Program. It can directly serve as an evaluation basis for the Forestry and Nature Conservation Agency's matchmaking platform, acting as a scientific bridge for industry, government, academia, and society, while guiding corporate capital precisely into low-elevation hills to achieve mutual prosperity.

Expected Results and Contributions

Based on the Verra Nature Framework 1.0, this project lays the research foundation for Taiwan's nature credit market: (1)Constructing Scientific Baselines for Forest Ecological Restoration: Analyzing the four major low-elevation hill corridors using machine learning to produce the first "Best Reference Value" map, guiding hotspot prioritization. (2)Developing an Internationally Aligned Quantification Tool for Nature Credits: Mapping the Best Reference Value ($Rv$) to Verra's "Quality Hectares (Qha)" formula to provide a credible habitat restoration measurement standard, ensuring that funded investments are converted into verifiable nature assets. (3)Facilitating Corporate Implementation of the TNFD Framework and Nature Positive Goals: Assisting enterprises in the "Locate" and "Evaluate" phases of the LEAP framework, thereby driving progress toward Nature Positive targets.