CarbonEthics Presents at First International Young Environmental Scientist (YES) Conference by WRI Indonesia: Enhancing AI-Driven Land Cover Analysis

The CarbonEthics' team (second from right) presented at the YES conference hosted by WRI Indonesia

CarbonEthics participated in the 1st International Young Environmental Scientists (YES) Conference hosted by World Resources Institute Indonesia in 2025. At the conference, CarbonEthics presented its research titled “Enhancing Land Cover Classification Accuracy in Tropical Indonesian Regions Using Multi-Sensor Satellite Data and Advanced Machine Learning Techniques.”

The research highlights the importance of accurate Land Use/Land Cover (LULC) data in estimating carbon stocks and supporting the development of nature-based carbon projects. Reliable land cover classification plays a critical role in establishing baselines, assessing ecosystem conditions, and informing carbon accounting processes.

CarbonEthics' Team speaking during the presentation session

However, developing accurate LULC data for tropical regions presents several challenges. These include limited availability of multi-temporal land cover datasets to support baseline development, limitations in the adaptability of multi-model approaches to complex tropical landscapes, and inconsistencies in the accuracy of global land cover platforms when applied to tropical environments.

To address these challenges, CarbonEthics developed three methodological approaches. First, the research utilizes a robust artificial intelligence (AI) model that integrates complementary Earth observation datasets from the Sentinel-1 and Sentinel-2 satellites using Top-of-Atmosphere (TOA) reflectance data. This multi-sensor approach enables the model to capture diverse land surface characteristics and improve classification performance.

Second, the AI model is specifically calibrated for complex tropical landscapes. The model was trained using data from 291 locations representing approximately 106 million hectares across Indonesia. In addition, rule-based temporal analysis was incorporated to improve classification consistency across different land cover dynamics.

Third, the research aims to produce highly consistent land cover classification at a 10-meter spatial resolution, enabling a more accurate representation of heterogeneous tropical landscapes.

See more: CarbonEthics at ABAC Meeting I 2026: Building a Credible Carbon Market in Geoeconomic Landscape

This work contributes to CarbonEthics’ broader research agenda to support data-driven screening for carbon project feasibility, particularly during the pre-feasibility study stage. Accurate land cover analysis plays an important role in identifying suitable project areas and improving the reliability of early-stage carbon project assessments.

Three Methodological Approaches to LULC Challenges

As part of the next phase of development, the research team is continuing to refine the model by testing additional rule-based approaches and machine-learning combinations to enhance its performance and accuracy further. CarbonEthics is also preparing this research for submission to a Scopus Q1-indexed journal, while exploring opportunities to expand land cover processing and analysis across Southeast Asia.

CarbonEthics continues to advance science-driven carbon project development.

Explore our carbon project at:  carbonethics.co/nature-based-carbon-project 

---

About CarbonEthics

CarbonEthics is a tech-enabled Ecosystem Restoration Company specializing in carbon project development and blue carbon ecosystems. We bridge ecosystem restoration, climate finance, and regulatory expertise to deliver high-integrity carbon solutions that create real impact for local communities while preserving biodiversity.