Satellite embedding‑based mapping of grassland mowing frequency and its response to drought in Denmark

Mowing frequency is a primary indicator of grassland management intensity, forage productivity, and biogeochemical dynamics. Yet, despite a growing body of literature, operational frameworks capable of mapping cutting regimes at a regional scale remain computationally demanding and lack independent benchmarking. Here, we evaluate the potential of Google AlphaEarth annual satellite embeddings to map grassland mowing frequency across Denmark (2017–2023) at 10-m resolution. Using farmer logs across multiple years (n = 346), Random Forest emerged as the top-performing classifier, achieving an overall accuracy of 0.73 and a macro F1-score of 0.65. Our workflow substantially outperformed the operational Copernicus Land Monitoring Service (CLMS) Grassland Mowing Events (GRAME) benchmark (overall accuracy of 0.24). Latent embedding dimensions strongly correlated with macro-climatic and vegetation drivers (|ρ| > 0.8). Once deployed across the full nation extent, under 6 min of computing time, the model predictions proved ecologically consistent across pedological and agronomic gradients. We further applied Dirichlet regression to quantify how hydrological drought, indexed by a 6-month scale Standardised Precipitation-Evapotranspiration Index (SPEI-6), affects national mowing patterns. The extreme 2018 drought triggered a landscape-wide management de-intensification, shifting the intensive (≥4 cuts) regime into a 2 cuts system. During moderate droughts, regional responses diverged sharply: western regions exhibited agronomic resilience, whereas drier eastern islands repeatedly reverted to lower-intensity management. These results demonstrate that satellite foundation models can underpin scalable, automated agronomic monitoring and reveal grassland management’s acute sensitivity to drought, insights essential for adaptive land-use planning and national-scale environmental reporting.

Authors

Institutions

Publication Details

Journal
International Journal of Applied Earth Observation and Geoinformation
Published
2026-11-18
DOI
https://doi.org/10.1016/j.jag.2026.105687
Primary Topic
Remote Sensing in Agriculture
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Satellite embedding‑based mapping of grassland mowing frequency and its response to drought in Denmark

Klaus Butterbach‐Bahl, Michele Meroni, Diego Ábalos, Jaber Rahimi et al.
International Journal of Applied Earth Observation and Geoinformation
Remote Sensing in Agriculture
article

Satellite embedding‑based mapping of grassland mowing frequency and its response to drought in Denmark

Klaus Butterbach‐Bahl, Michele Meroni, Diego Ábalos, Jaber Rahimi, Francesco Pietro Fava, Mogens Humlekrog Greve, Rodolfo Ceriani, Lars Uldall-Jessen, Søren Kolind Hvid
article en

Abstract

Mowing frequency is a primary indicator of grassland management intensity, forage productivity, and biogeochemical dynamics. Yet, despite a growing body of literature, operational frameworks capable of mapping cutting regimes at a regional scale remain computationally demanding and lack independent benchmarking. Here, we evaluate the potential of Google AlphaEarth annual satellite embeddings to map grassland mowing frequency across Denmark (2017–2023) at 10-m resolution. Using farmer logs across multiple years (n = 346), Random Forest emerged as the top-performing classifier, achieving an overall accuracy of 0.73 and a macro F1-score of 0.65. Our workflow substantially outperformed the operational Copernicus Land Monitoring Service (CLMS) Grassland Mowing Events (GRAME) benchmark (overall accuracy of 0.24). Latent embedding dimensions strongly correlated with macro-climatic and vegetation drivers (|ρ| > 0.8). Once deployed across the full nation extent, under 6 min of computing time, the model predictions proved ecologically consistent across pedological and agronomic gradients. We further applied Dirichlet regression to quantify how hydrological drought, indexed by a 6-month scale Standardised Precipitation-Evapotranspiration Index (SPEI-6), affects national mowing patterns. The extreme 2018 drought triggered a landscape-wide management de-intensification, shifting the intensive (≥4 cuts) regime into a 2 cuts system. During moderate droughts, regional responses diverged sharply: western regions exhibited agronomic resilience, whereas drier eastern islands repeatedly reverted to lower-intensity management. These results demonstrate that satellite foundation models can underpin scalable, automated agronomic monitoring and reveal grassland management’s acute sensitivity to drought, insights essential for adaptive land-use planning and national-scale environmental reporting.

International Journal of Applied Earth Observation and GeoinformationVol. 155
Karlsruhe Institute of Technology (DE), University of Milan (IT), Aarhus University (DK), Knowledge Centre for Agriculture (DK)
Aarhus Universitet, European Commission, Danmarks Grundforskningsfond, Novo Nordisk
Life on land
Remote Sensing in Agriculture
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon