Thermal imaging and vegetation detection through UAV survey for large scale hazard monitoring of river levee

Levees are linear structures that can be thousands of kilometers long and play a very important role in flood protection. They are usually monitored by traditional direct survey techniques, such as CPTU or coring, or piezometers, which provide high accuracy, but are localized and performed in predetermined locations.

As a result, long distances between investigated sections limit the detailed analysis of the entire structure. In addition, predetermined locations may not cover areas of actual potential weakness.

Recently, new survey technologies from aerial media (drones) have been successfully applied to obtain a first level of levee investigation in order to identify the location of possible weak areas or potential locations of levee failure, so as to plan further local investigations in those areas.

Usually, levee failures are localized in the presence of:

(i) concrete structures passing the levee;

(ii) large trees, which can be dangerous because their roots are a preferred route for water infiltration and, therefore, potential seepage pipes. In addition, at higher erosion levels of the river bank, large trees can promote bank collapse due to their weight;

(iii) sandy soils, which are characterized by high permeability. From previous experience, we have noticed that levee failures have occurred at sections previously vegetated by reeds. Reed canes usually grow on sandy soils and, in addition, are characterized by very deep and large roots, possible routes of localized infiltration through the body of the levee. From these observations comes the idea of using reedbeds as indicators of sandy soils and possible weak levee sections;

(iv) sections where unfavorable conditions of the levee body, such as soils with high permeability or the presence of animal burrows crossing the levee or obstructed drains, prevent proper drainage and bring the phreatic surface close to the levee surface.

Thus, the idea is to test different innovative UAV-supported survey approaches on the same test area, in combination with local on-site surveys, to compare and combine the obtained results. Firstly, we would test the possibility of using vegetation maps as an indicator of weak sections of the embankment. Up to now, a first drone survey data has been performed and the obtained RGB orthophotos have been elaborated to determine the Green Red Vegetation Index (GRVI), in order to acquire a vegetation cover map of the embankment. The obtained data have been calibrated with on-site surveys conducted by vegetation experts. To facilitate the identification of reedbeds, the campaign has been carried out in winter, when reedbeds are yellowish in color, unlike short grass. In areas identified as reedbed vegetated, the soil has been sampled by coring and fully classified in the geotechnical laboratory to check if reedbed can effectively be an indicator of sandy soils. Further characterization may be carried out in order to investigate the relationship between reedbeds and soil characteristics.

The final aim is to develop an innovative method of low-cost aerial monitoring of levee structures that can provide an initial state of information and identify areas in need of further direct investigation in order to define the necessary maintenance works, decreasing associated risks.