Mapping of Groundwater-Dependent Ecosystems in Denmark utilizing remotely sensed indices and topography in unsupervised clustering

Groundwater-Dependent Ecosystems (GDE) can be broadly categorized as ecosystems where 
the vegetation utilizes groundwater for a significant part of transpiration and depends on 
groundwater access for maintaining plant health. The use of remotely sensed data for GDE
detection has evolved considerably in the past decade. Especially areas with a distinct dry
season have received much attention, as GDEs remain ‘greener’ during dry periods which 
makes dry-season NDVI an excellent indicator for GDE presence. However, for temperate 
GDEs, where no distinct dry season occurs, indicators suitable for GDE identification are 
currently lacking.  
Denmark is characterized by a temperate climate, which challenges existing GDE detection 
methods. To overcome this, we introduce two NDVI-based GDE indicators. Initially, NDVI 
dynamics of known GDEs were compared with surrounding ecosystems in a well-studied river 
valley containing cultivated and pristine peatlands with shallow groundwater. It was found 
that GDEs have a later onset of the growth season, due to soils being water-logged. To derive 
this NDVI-based GDE indicator, the average relative difference of NDVI between March and 
July from 2018 to 2024 was calculated for each cell. The second method uses the difference 
in responses to occasional summer droughts. The drought year 2018 resulted in large-scale 
wilting of vegetation in Denmark, but GDEs, being able to utilize groundwater, were more 
resilient. Thus, the summer of 2018 could be used as a pseudo dry season, and the difference 
of NDVI between 2018 and the average of the following 5 years was calculated for each cell as 
the second NDVI-based GDE indicator. Sentinel-2 at 10m resolution was sourced for 
calculating the NDVI indicators. The high-spatial resolution of the Sentinel data was critical, 
as the Danish GDEs are often small (below 1 ha), and found in narrow river valleys with 
considerable heterogeneity in land use and land cover. The two NDVI-based GDE indicators 
were applied together with topography-based indicators in different classification approaches 
to map GDEs. The tested classification approaches were based on a manual scoring routine 
and an unsupervised clustering. Their results were evaluated against more than 10,000 
polygons spanning ~110 km2 with GDE information derived from field surveying. It was found 
that incorporating the two NDVI indicators together with topography and depth to the 
groundwater table resulted in a very satisfying classification. The derived spatial patterns of 
the classification could largely be linked to land use, i.e. drainage of peat soils in the river 
valleys for cultivation or grazing.