Integrating Sentinel-2 Land-Cover Classification with Peatland GHG Assessment in Latvia
ラトビアにおける泥炭地GHG評価とSentinel-2土地利用分類の統合 (AI 翻訳)
Maksims Feofilovs, Linda Gulbe-Viluma, Andrei Grishanov, Ilze Barga, Amrutha Rajamani, Nidhiben Patel, Claudio Rochas, Francesco Romagnoli
🤖 gxceed AI 要約
日本語
本研究は、衛星画像（Sentinel-2）と教師付きランダムフォレスト分類器を用いて、ラトビアの泥炭地の土地利用分類を行い、排出係数と組み合わせることで温室効果ガス（GHG）排出量を推定するワークフローを提案する。分類精度は高層湿原や採掘地で高く、不均一なクラスで低かった。この手法は現地データが限られた地域でのスケールアップに有効だが、不確実性や時間変動の分析は未実施である。
English
This study presents a proof-of-concept workflow integrating Sentinel-2 land-cover classification (Random Forest) with emission factors to upscale peatland GHG estimates in Latvia. Classification accuracy was high for spectrally distinct classes like raised bogs and extraction areas, but lower for heterogeneous classes. The approach supports regional GHG accounting when field data is scarce, though uncertainty and temporal variability are not assessed.
Unofficial AI-generated summary based on the public title and abstract. Not an official translation.
📝 gxceed 編集解説 — Why this matters
日本のGX文脈において
ラトビアの泥炭地を対象とした手法だが、日本でも北海道などに泥炭地が存在し、土地利用変化に伴うGHG排出評価に応用可能性がある。ただし、日本のSSBJや有報などの開示枠組みに直接関係するわけではない。
In the global GX context
This paper demonstrates a scalable, low-cost remote sensing method for peatland GHG accounting, relevant to global efforts in land-use-based emissions reporting (e.g., under UNFCCC). It highlights the integration of satellite data with inventory-style emission factors, a common approach in national GHG inventories, though it does not yet meet the rigor required for corporate disclosure standards like TCFD or ISSB.
👥 読者別の含意
🔬研究者:Provides a reproducible workflow for coupling land-cover classification with GHG emission factors, useful for regional upscaling studies.
🏛政策担当者:Shows how satellite data can support spatially explicit GHG inventories for peatlands, aiding land-use planning and climate reporting.
📄 Abstract（原文）
Draining peatlands for peat extraction converts them into significant sources of greenhouse gas (GHG) emissions. Quantifying GHG emissions at the regional scale remains challenging because direct field measurements are spatially limited, while GHG accounting for land-use planning requires spatially explicit information. Building on the advances in remote sensing (RS) as a scalable low-cost emission accounting tool for large areas, this study presents a proof-of-concept workflow that integrates satellite-based land-cover classification with an emission-factor (EF) approach to support spatial upscaling of peatland GHG estimates. Using Sentinel-2 imagery and a supervised Random Forest classifier, peatland-related land-cover classes were mapped for selected sites in Latvia. The classification results show higher accuracy for spectrally distinct classes such as raised bogs and active peat-extraction areas, while more heterogeneous classes exhibited lower performance. The study provides an overview of how to utilize the RS approach to generate accurate land-cover maps, which can be used to upscale GHG estimation in Latvia when field data is limited. The study does not include calibration against site-level flux measurements, uncertainty propagation, or temporal variability analysis; therefore, the emission results are illustrative and consistent with current EF-based inventory practice rather than validated site-specific fluxes.
🔗 Provenance — このレコードを発見したソース
- crossref https://doi.org/10.3390/land15050766first seen 2026-05-14 23:56:03
🔔 こうした論文の新着を逃したくない方は キーワードアラート に登録（無料・3キーワードまで）。
gxceed は公開メタデータに基づく研究支援データセットです。要約・翻訳・解説は AI 支援で生成されています。 最終的な解釈・検証は利用者が原典資料に基づいて行うことを前提とします。