// Land · Agriculture · Environment · Data
We use satellite observations, computational modeling, and machine learning to understand how land, water, and climate interact — from individual fields to the entire planet.
Mapping crop types, irrigated areas, and agricultural dynamics from space using multi-temporal satellite imagery (Landsat, MODIS, Sentinel). Developing methods that work at field scale and across continents.
Quantifying irrigation water use, evapotranspiration, and land-atmosphere feedbacks. Applications span the US, Middle East, Central Asia, and Sub-Saharan Africa. Contributing to OpenET and related data products.
Developing data-driven and model-based methods to estimate crop yields at field and regional scales, combining satellite observations with machine learning and process-based crop models.
Detecting forest disturbance, post-Soviet farmland abandonment, and land cover transitions across large regions using Landsat time series and advanced classification algorithms including Support Vector Machines.
Assessing climate change impacts on crop productivity, snow water availability, and watershed hydrology. Studies span Turkey, Ethiopia, the Mekong Delta, Sri Lanka, and the US Midwest.
Applying deep learning, Random Forests, and Google Earth Engine to large-area remote sensing problems. Developing generalizable classifiers that operate across diverse environments with limited training data.
Director, CULTIVATED Lab
Associate Professor
Dept. of Forest & Wildlife Ecology · Nelson Institute for Environmental Studies · UW–Madison
Dr. Özdoğan's research focuses on the application of satellite remote sensing to understand interactions among land use, water resources, climate, and agriculture. His work spans scales from individual farm fields to global mapping efforts, with active projects in Africa, Asia, Europe, and the Americas. He has published over 50 peer-reviewed articles and led numerous NASA, World Bank, and CGIAR-funded projects.
He teaches remote sensing, environmental modeling, and GIS at UW–Madison, and has trained researchers across four continents through field workshops and international training programs.
Selected graduate advisees (* denotes student co-authorship on publications)
A generalized Landsat-based classifier for mapping cropland extent across large, diverse regions with limited labeled training data. Applied across Europe, the Middle East, Russia, and Central Asia.
→ View on GitHubFramework for field-level crop yield estimation combining Landsat imagery with DSSAT crop model outputs through a hierarchical Bayesian data assimilation scheme.
→ View on GitHubA phenology-guided, training-independent algorithm for time-series winter wheat detection at large scales. Published in IEEE TGRS 2026.
→ View paper and codeTemporal unmixing code using Independent Component Analysis to derive sub-pixel crop type fractions from MODIS time series. Described in RSE 2010.
→ Contact for accessAdditional code and scripts are available on request. Please email Dr. Özdoğan at [email protected].
Global 30-m resolution cropland extent map derived from Landsat time-series data for 2015 using multiple machine learning algorithms on Google Earth Engine. Published as USGS Professional Paper 1868.
Field-scale ET estimates for irrigated agriculture across the western United States, generated from an ensemble of satellite-based ET models. Supports water rights and irrigation management decisions.
Time-series analysis of tree cover dynamics in Senegal spanning 1990–2020, derived from Landsat archives. Published as a CGIAR Professional Paper.
Multi-year, conterminous US crop type distribution product derived from MODIS phenology, covering major commodity crops at 500-m resolution.
Large-area cropland extent maps at Landsat resolution for Europe, the Middle East, Russia, and Central Asia, generated with Random Forest classifiers in Google Earth Engine.
Annual irrigated area maps for the CONUS derived from multi-temporal MODIS data and USDA ancillary information. Widely used as benchmark for land surface model validation.