• Marc Ostermann Larsen
  • Rasmus Skole Jensen
4. term, Physical Geography, Master (Master Programme)
This project will examine ways of mapping floods from precipitation with radar satellites in the sizeable catchment area of Denmarks longest river: Gudenåen in the central part of Jutland. Pictures from radar satellites presents a number of advantages to other methods of monitoring floods, where an independence of weather and light from other sources are the main advantages. Floods are often a result of bad weather, and thus the extent of the floods can not be monitored with conventional monitoring metods in this scale.
The report will be split up in three parts. The first part will give the reader needed knowledge of how radar works, and also how radar on a satellite works.
The second part will introduce two monitoring concepts, which will include ways to determine the extent of a flooded area with a satisfactory accuracy. The first concept is the most simple and fastest of the two and will use different thresholds of when a cell in the satellite picture is water, and when it is not, to determine the flood extent.
The second concept will try to remove areas in the flood maps, that the radar satellites wrongly estimates as water.
These monitoring concepts will be validated with the method Confusion Matrix, which is a method commonly used for remote sensing data. In addition to this, another faster validation method will be used, to determine if this could be a way of getting results faster with simlar accuracies.
The third and last part will present which of the monitoring concepts best determines the flood extent and then if the faster validation method gives adequate results. It is found that the best total result from both monitoring concepts is mapping $\approx$ 47\% of the total flood extent with an accuracy of $\approx$ 80\%, and that neither of the concepts are decidedly better than the other. Additionally it is found that the faster validation method is adequate to use, for the first monitoring concept but not the other. Lastly, ways to get higher accuracies for the second concept are discussed, which presents great opportunities for future development of the concept.
Publication date10 Jun 2022
Number of pages160
ID: 472530345