Why is it important?

The results of this study provides the basis for using satellite data sources and Machine Learning methods for a nationwide automated classification and quality assurance of topographic maps in the GeoDanmark dataset.

A further development of the methods could make a nationwide screening of all lakes in Denmark possible.

Project highlights:

Using Sentinel time series and Machine Learning to validate lakes in existing topographic maps

Providing quality managers with a cost-effective tool to prioritise efforts with quality assurance of existing and future data sets

Showing the possibility of a nationwide screening of all lakes in Denmark in the future

In more detail..

A study area covering parts of Northern Jutland was analysed using relevant Sentinel 1 and 2 satellite imagery from 2017.

The temporal resolution of Sentinel-1 is high, and it therefore has the advantage of providing large amounts of data throughout the year. However, it is limited in its spatial resolution which makes the classification of relatively small objects difficult e.g. small lakes, since they do not show on the images.

Sentinel-2, on the other hand, has the advantage of having a significantly better spatial resolution, which enables the classification of small lakes. Here, the limitation is that clouds and shadows limit the temporal resolution.

Combining data from the two satellites, would therefore give a better data coverage.

We found that a relatively large portion of the mapped lakes in the GeoDenmark dataset, from a Sentinel-based point of view, either cannot be described as lakes or may be deviating from the existing categories.

The results of this study indicates that there are good prospects in using the Sentinel-based data sources and methods for a nationwide automated classification and quality assurance of topographical maps, such as the mapped lakes in the GeoDenmark data set.

In this study, we focused on previously mapped lakes in the GeoDenmark data set, but it is expected that the same method could be used to map lakes that do not already exist in the dataset.

Based on a limited further investigation of the methods used, it is estimated that a nationwide screening of all lakes in Denmark, approx. 180.000 in total, could be automatically categorized and figure in the GeoDenmark dataset for lakes.

The Danish Agency for Data Supply and Efficiency:

The agency provides the public and private sector with high-quality data, enabling them to make important community decisions based on the best available information.

The agency is part of the Energy, Supply and Climate Ministry in Denmark.

DHI GRAS A/S

gras@dhigroup.com
+45 4516 9100

Agern Alle 5,
2970 Hørsholm,
Denmark

CVR: 25621646