Land Cover Classification for the CSIR

Aug 31, 2009

Mission: Test the aerial surveying technique for a project which develops classification techniques to update the current National Land Cover (NLC), and which will produce maps which abide with international classification standards
Date: 31 August 2009
Requesting organisation: Council for Scientific & Industrial Research (CSIR)
Location: Centurion and Vereeniging, Gauteng
Pilot: Tony Kent and Rob Osner

In early August we were approached by Nic Badenhorst of the Satellite Applications Centre (SAC), of the Council for Scientific & Industrial Research (CSIR).  Nic gave us a detailed motivation for a flight request to test the aerial surveying technique for a project which develops classification techniques to update the current National Land Cover (NLC), and which will produce maps which abide with international classification standards.  This mission took place at the end of August and was flown by Tony Kent and Rob Osner. Hadley Remas, a Remote Sensing Specialist with the CSIR, was one of the passengers on the flight and he has provided a detailed scientific report of the mission, extracts from which appear below.  Anyone wishing to see the full report should please contact us at  info@bateleurs.org to request a copy.

The Land Cover Classification team. 
From the left:  Bateleurs pilots Tony Kent and Rob Osner, with Hadley Remas and Paida Mangara. (CSIR SAC), at Grand Central Airport.

A most demanding mission
by pilot Tony Kent

“The CSIR’s Satellite Applications Centre approached The Bateleurs requesting a flight for two separate surveys in the Centurion and Vereeniging areas. Sounded easy enough, and Rob Osner and I volunteered for this mission using his C210. The date was set for Monday 31st August.  But once they started to feed me information of the required area and the tracks to be followed, I began to realise this was not going to be one of those easy, relaxed missions.  I started pin-pointing the co-ordinates and began to wish we had not put up our hands for this particular flight!

As can be seen from the dotted line on the Jeppesen 1 : 250 000 chart, we would be traversing O.R. Thambo, Waterkloof, Grand Central, Lanseria and Wonderboom airspaces – all at around 500′ AGL.  Clearance was duly authorised by the Central Air Management Unit (CAMU) of Air Traffic & Navigation Services (ATNS).  For accuracy, the tracks were then down-loaded onto the GPS’s.  Accurate flying was specifically requested, since the object of the exercise was to take heaps of photos and compare them to satellite images of exact locations.

We ferried ourselves over to Grand Central Airport and met up with Hadley and his colleague, Paida Mangara.  A quick brief followed and then it was time to take off as per the filed Flight Plan, at 11h00 LT.  After take-off we turned east to start from the Olifansfontein area on our first track westwards heading towards Lanseria. Johannesburg Tower proved to be a breeze and we were soon out of their airspace, heading into Lanseria’s airspace. We left their airspace on re-positioning for the next line-up, heading eastwards and entering Waterkloof’s airspace.  Initially this was not a problem, but the third track took us over Swartkops and then Waterkloof Air Force bases.  CAMU had merely given them the co-ordinates of the outermost points of the survey area and they were unaware of our zig-zag routings.  After a lengthy exchange over the radio with a senior Controller we were allowed to continue.  The only other complication was keeping clear of prohibited airspaces such as the Voortrekker Monument and Iscor HQ.

Each leg took about 7 minutes, and the re-positioning turn for the next leg took about 3 minutes, so we completed the Centurion mission in approximately 50 minutes and could relax a bit as we headed off to Vereeniging.

The Vereeniging mission was exactly the same size as the Centurion ‘block’, and other than avoiding the Iscor/Mital factory, it was far less complicated, airspace-wise.  Around two-and-a-half hours later we touched down back at Grand Central to drop off the CSIR crew, bidding them farewell after what they professed was a successful flight.

Over a bite and welcome coldrink at Grand Central’s cafeteria, Rob and I took a couple of deep breaths and agreed this was the most demanding and difficult mission we had ever flown.  We were both so tired that the ferry flight back to home-base seemed to be an unusual challenge!  We trust the CSIR achieved their goal and while the flight was a toughie, with unusual challenges, this was yet another rewarding flying experience for the two of us sitting ‘up front’.  Thank you, again, to The Bateleurs!”

Super-sites for the CSIR
by passenger Hadley Remas

Project Background
The CSIR Satellite Application Centre (SAC) has embarked on a project with the Chief Directorate: Surveys and Mapping (CDSM) to develop a land cover field guide for a national mapping initiative for the Republic of South Africa.  Over the past few decades the classification of remote sensing satellite images has proved to be the most essential and cost-effective way for initiatives like this.  The CSIR SAC receives, archives and distributes satellite imagery and performs value addition and thematic information extraction as a service to the government and private industry.  Aerial photographs are also useful in the classification process especially since they have traditionally provided higher resolution images than satellite images.  The CDSM is the custodian for aerial photographs in South Africa.

Project
Eight areas (supersites from here on) with 32 diverse land cover classes have been identified in South Africa for this project. Two of these supersites are within the Gauteng Province – Vereeniging and Centurion.  Each of these areas are measured in 15′ x 15′ blocks.  The Vereeniging area was identified mainly because of its summer crops, mining activities and heavy industries, whereas the Centurion area was identified mainly for its metropolitan urban classes.  Other land cover classes do exist within these areas

The flight plan for the Centurion block.

Land_Cover02
  • Facebook
  • LinkedIn
  • Twitter
  • Pinterest

 

The flight plan for the Vereeniging block.

Land_Cover03
  • Facebook
  • LinkedIn
  • Twitter
  • Pinterest

Ground truthing
In order for the effective analysis of satellite imagery, ground truthing is necessary to compare a pixel on a satellite image with what is happening on the ground, i.e. it is a method to assess the accuracy of classified satellite images.  In the ground truthing process, the investigator is usually equipped with a camera and a GPS to get the co-ordinates of the features on the ground, and later compares it to the classified image.  Points to be visited are randomly selected and stratified and are within 100m of major, secondary and other roads.  For relatively small areas, investigators usually cover the area of interest by foot, while bigger areas are covered by car, aeroplane or helicopter.

Based on previous experience, a 15′ x 15′ block, can be covered in 2 to 4 days by car.  This is a very time consuming exercise, taking into account that four photos must be taken in sequence, in all major directions – north, south, east and west.  Besides time and route planning, road traffic, the accessibility of points and general safety are other factors that must to be taken into account when doing ground truthing by car.  Ground truthing by means of air transport was selected for accuracy of assessment purposes and for a much more time efficient coverage of the supersites.

The flight plan was based on the 15′ x 15′ blocks of the two areas of concern.  Five horizontal flight lines, 3 minutes apart, were selected in order to get a full coverage of the area.

SAC staff were equipped with the following cameras:  a Canon EOS 1000D (10 Megapixel), a  Sony Cybershot (7.2 Megapixel), and a Canon Cybershot (3.2 Megapixel), plus a GARMIN GPS and a hard copy flight plan (SPOT 5 2.5m satellite image) of the selected areas.  We were unable to track satellites with the GPS at the back of the aircraft and therefore the flight path could not be traced as initially anticipated.  However, points were marked on the flight plan and Tony Kent’s knowledge of the Gauteng area came in very handy when he communicated with us via the headphones.  Photos were taken from both sides (left and right) of the aircraft and were noted on the hard copy flight plan by the SAC staff. 

Project outcome
Overall, this part of the project (collecting airborne reference points) was regarded as a great success.  Approximately 1000 photos were taken with the three cameras and inaccessible areas such as mines and quarries, and areas where safety was a concern – such as  informal settlements – were traversed effortlessly from the air.  This information will aid the rest of the project, especially with ground referencing and accuracy assessment.”

Pin It on Pinterest