Identifying Urban Creep with a UAV Survey

It is quite well known amongst the industry that multi-spectral data can be used to identify impermeable areas, aiding the development of drainage solutions and flood prevention measures.

Existing data is out of date and lacks precision = inaccurate models

Richard Allitt Associates believes UAVs are now an essential tool in producing accurate and up-to-date models of our landscape and preventing flooding problems.

Most of the current data for models has historically been collected from large fixed wing aircraft and although it has proven to be useful for a wide range of different modelling work, most of this data was collected in 2010-2011, so while still useful for simple modelling projects, there are some circumstances when it falls short of the mark.

Satellite data is another set of data that is available to hydraulic modellers. It is collected every 2-3 weeks at a ground resolution of 30m, however, while it is up-to-date it does not offer the kind of precision that is sometimes needed. This has led us to consider a new source of data which addresses both the issue of timeliness and detail.

Taking account of urban creep

Last year Richard Allitt Associates were given the task by Wessex Water of remodelling a catchment at Bishops Lydeard, in order to take account of urban creep. Urban creep occurs when historically permeable areas are redeveloped into driveways, car parks or any other types of impermeable area. Fixed wing data for this location was found to be over 4 years old and thus would not have accounted for recent redevelopment. Furthermore, satellite data although up-to-date, did not possess high enough resolution in order to classify impermeable areas accurately. The presence of urban creep was consequently causing errors when interrogating the models, and so an up-to-date detailed reclassification was needed. 

 UAVs are the perfect survey solution to get up-to-date high resolution imagery.

 

Figure 1 – Skyjib Super 6 Ti-QR Hex-rotor Platform

One of the company’s pilots collected new data using a small multi-rotor platform carrying a multispectral camera, as shown above. The platform is flown in a ladder formation collecting images from pre-programmed waypoints. An average flight takes in the region of 10-15 minutes to complete, and the project surveyed an area totalling 13.4 hectares from just 7 flights. Once collected, the data is processed using a photogrammetry program, and an accurate georeferenced 2D map is constructed of the location in both visible (VIS) and near infra-red (NIR) light.

Below you can see a comparison of the raw imagery from a car park that was introduced. The image on the left is off-the-shelf data collected from a full sized, fixed-wing aircraft, and the image on the right is produced from data collected from an UAV. As well as identifying the car park being concreted over, there is also quite clearly an improvement in the resolution of the imagery. This means that smaller permeable areas can be identified accurately. Once geo-referenced false colour composite imagery has been produced, the final stage involves the classification of permeable vs impermeable areas, using the Normalised Differential Vegetation Index (NDVI)*

(a)                             (b)

Figure 2 – Two near infra-red false colour composite images derived from images collected by (a) full sized fixed wing aircraft in 2010 (Left), and (b) a UAV in 2014 (right)

Once processed, thresholding techniques are used to classify areas which are covered in vegetation versus those which are not. There is a high correlation between permeability of the ground and vegetation cover. The main exceptions are those areas which have recently been ploughed, and so a manual inspection step is introduced to identify these regions.

Once the whole process has completed, a full vectorised model can be produced. The Figure below shows the finished model. All those areas which are shaded are classified as impermeable areas. Those areas shaded in solid red signify areas which are classified impermeable purely from the full sized fixed wing data set. Those in blue represent areas which were classified using the UAV data set, and areas in green represent those classified by both the data sets. Black regions are buildings, roads, rivers and pavements, as derived from the ordinance survey master map data set.

Figure 3 – Map of Impermeable Areas -classified from:

  • the full sized fixed wing data set (red)
  • from the UAV data set (blue)
  • and by both data sets (green) 
  • This is complimented with road, building, river and pavement data obtained from OS master map (black).

Urban creep can be identified

Urban creep is represented by those areas highlighted in blue, since they were classified permeable using the 2010 fixed wing data set, and now they are classified as impermeable using the 2014 UAV data set. Areas highlighted in red identify the regions in which impermeable areas have been changed back to permeable areas. These areas are a lot less noticeable and it can be seen that the net change results in urban creep.

UAVs make this data affordable

Without the easy application afforded by a UAV, this kind of work would be too expensive to conduct and thus the accuracy of the models being produced would be perennially compromised.  This may result in localised flooding close to recently paved over areas, or other undesirable effects. Richard Allitt Associates believes UAVs are now an essential tool in producing accurate and up-to-date models of our landscape and preventing flooding problems.

To find out more about updating your existing data and models please call us on 01444 401840 

                                      

*NDVI is a graphical indicator which can be used to determine whether an area contains live green vegetation. It works from the logic that plants absorb visible light which is used for photosynthesis; however they are unable to absorb NIR light. By looking at the differences between the reflected values of NIR and VIS light, the vegetative state of each pixel can be processed using the standard NDVI equation:

NDVI = NIR-VIS/NIR+VIS