Research Outputs

Innovative components

Understanding the seasonal variation of low floodplain vegetation (from 0.5 up to 3 m high) is crucial for flood modeling, ecosystem services, and ecological applications. However, changes at the centimeter scale are not easy to monitor with traditional aerial photos. At a plot scale, we tested the performance of the so-called unmanned airborne vehicle or drones to derive a digital terrain level and to record both the vegetation growth and greenness changes over a growing season (see right figure). As a reference for the novel imagery derived from the drone sensor, we manually took a measure of vegetation height for every plot and seasonal period. Thereby, we processed the imagery into point clouds with color attributes for each time step. From the differences between the point clouds, we calculated per plot the terrain and vegetation height as well as vegetation greenness.

Findings and implications to practice

We studied a 116 ha floodplain, known as Breemwaard that is located on the southern bank of the Dutch river Waal in the (see figure above). Simultaneously with each drone survey, we collected data on vegetation height in 28 plot for six times over a growing season (February, April, June, September, November 2015 and January 2016). Results show high potential for increasing the classification accuracy of low floodplain vegetation (see right figure).

Moreover, the accuracy of the digital terrain model derived from the drone survey varied over a year and was most accurate during winter. The vegetation height was derived in a simple way (via linear regression) from the multitemporal imagery. We also identified temporal greenness and height profiles: The growth of herbaceous vegetation (0.5m up to 2.5m) follows a loop according to the seasonal stage whereas such loop was not clearly identified in grassland vegetation (maximum height of 0.5m).