Using geospatial data to improve targeting of agricultural projects in the context of climate change and resource scarcity

Darius A. Görgen


Who am I?

  • 🙋️ Darius A. Görgen

  • 🎓 M.Sc. Geography, B.Sc. Political Sciences

  • 🌍 Focus on Climate Change and Agriculture

  • Part of the MAPME Initiative since 2020

  • 🔬 Advocating for OpenSource and OpenScience


  • What is geospatial data?

  • From visual interpretation to automated analysis

    1. Counting trees by hand
    2. Deep Learning for field boundary delineation
    3. Satellite time series for crop type identification
    4. Data fusion for crop biophysical monitoring
  • Wrap-Up

Geospatial data

Geospatial data

Conceptualization of space in the dominant digital formats.
  • Landsat and Copernicus mission provide access to satellite archives for free
  • we can look back in time
  • we can produce valuable information by intersecting vectors with rasters
  • we can derive information for inaccessible areas

From visual interpretation to automated analysis

Visual interpretation

A tree plantation near the Jordan EcoPark.
  • aerial imagery can help monitoring progress
  • VHR images are usually quite expensive
  • sensible approach for small areas and simple information requirements, e.g. counting trees

Field boundary delineation

  • very often our area is too large for manual interpretation
  • we need tools that automate the interpretation of satellite imagery
  • Meta AI’s Segment Anything Network is already used in the agricultural sector
  • field boundaries can be used to inform about the area distribution of farms …
  • … but also they might be required for later analysis stages

Screenshot of agricultural boundaries produced by SAM.

Crop type indentification

We want something like this:

Example of a crop type classification map on the field level.

But how do we get there ..?

Satellite imagery timeseries …

Animation of a Sentinel-2 timeseries over an agricultural area.

… reveal distinct signatures of crops over time.

Temporal NDVI profiles of different crop types.

Mapping biopyhsical parameters

  • monitoring of crop development over time
  • derive key parameters such as:
    • crop health
    • growing cycle
    • water consumption
    • yield estimation
  • requires large amount of high-quality data collected on ground
  • requires the application of advanced statistical methods

Visualisation of an data fusion approach from Sentinel 1 and 2 for crop biophysical monitoring.


  • geospatial data can help to better target areas of intervention
  • remote sensing can deliver valuable insights, especially in data scarce regions
  • All models are wrong, but some are useful. (George Box)
  • evaluate the low-hanging fruits first
  • gold standards require large amounts of high-quality and thus expensive data
  • visual interpretation is a valid approach for small areas and simple problems
  • if the detail of required information increases, so do the data quality requirements
  • regional analysis can reveal key information:
    • area distribution of farms
    • distribution of crop types
    • yield estimation and other biophysical parameters

Thank you for your attention!

Extra Slides

Flooded areas

Animation of 2022 monsoon floods in Bangladesh.
  • support for immediate disaster responses
  • assess damages after flood events, e.g. to settle insurance claims
  • determine areas prone to flooding to improve land use planning

Climatological drought

  • SPI/SPEI to quantify intensity and duration of meteorological droughts
  • SPEI is preferable when temperature or \(ET_0\) data is available
  • gridded datasets allow drought analysis even in data scarce regions
  • CHELSA has good performance for complex terrains and data-scarce regions
  • includes climate projections for different CMIP6 scenarios

CHELSA animation of average precipitation between 1981-2010.

Comparison of SPI and SPEI to charachterize meterological drought.

Water accounting

  • Water accounting study in the Jordan River Basin by FAO
  • Uses remote-sensing based variables by FAO’s WAPOR
  • Differentiates between water generating (\(P > ET_a\)) and consuming (\(P < ET_a\)) land cover classes

(a) \(ET_a\)

(b) \(P\)

Figure 1: Contribution of landcover classes to \(ET_a\) (a) and precipitation (b) in the Jordan River Basin.

Difference between Precipitation (\(P\)) and Actual Evapotranspiration and Interception (\(ET_a\)).


FAO, and IHE Delft. 2020. “Water Accounting in the Jordan River Basin.” FAO WaPOR Water Accounting Reports.
Karger, Dirk Nikolaus, Olaf Conrad, Jürgen Böhner, Tobias Kawohl, Holger Kreft, Rodrigo Wilber Soria-Auza, Niklaus E. Zimmermann, H. Peter Linder, and Michael Kessler. 2021. “Climatologies at High Resolution for the Earth’s Land Surface Areas.” EnviDat.
Saab, David. 2003. “Conceptualizing Space: Mapping Schemas as Meaningful Representations.” PhD thesis.
Silva-Perez, Cristian, Armando Marino, and Iain Cameron. 2022. “Learning-Based Tracking of Crop Biophysical Variables and Key Dates Estimation from Fusion of SAR and Optical Data.” IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing 15: 7444–57.
Tariq, Aqil, Jianguo Yan, Alexandre S. Gagnon, Mobushir Riaz Khan, and Faisal Mumtaz. 2022. “Mapping of Cropland, Cropping Patterns and Crop Types by Combining Optical Remote Sensing Images with Decision Tree Classifier and Random Forest.” Geo-Spatial Information Science 0 (0): 1–19.
Vicente-Serrano, Sergio M., Santiago Beguería, and Juan I. López-Moreno. 2010. “A Multiscalar Drought Index Sensitive to Global Warming: The Standardized Precipitation Evapotranspiration Index.” Journal of Climate 23 (7): 1696–1718.