|Mission with Chimanimani National Reserve|
Conducting Change Detection in Chimanimani National Preserve, Mozambique
By Kerry Brooks & Madyo Couto
This project was commissioned by the Transfrontier Conservation Areas Unit of Ministry of Tourism in
Overview of Processing
The overall project objective was simple: examine imagery from 2005 and 2008 to identify areas where illegal gold panning operations may be occurring, using existing data sets. The main issue in carrying out this simple task turned out to be that the initial imagery had several problems, including a fairly large misalignment between dates. It was also the case that not all areas of the Reserve were covered by the original data, though the core central mountain area was covered.
Figure 1: Location of Chimanimani National Reserve
Several not-very-successful attempts were made to obtain alternative or additional data, (including being selected in the Digital Globe's 8-Band Challenge). Finally, a wonderful, cloud-free 2011 Quickbird dataset was donated by Digital Globe. That imagery was not only up-to-date, but also covered almost all of the Reserve except for a small area in the extreme NW corner. This meant that a workflow that first identified all potential panning/mining areas that existed in 2011 was applied and then compared the 2011 delineations to the 2005 imagery in areas that had 2005 coverage. That is how it was determined whether the 2011 mining areas existed in 2005 or were new since then and that accomplished the change-detection goal of the project.
Some Processing details
The resolution of the imagery was fairly high, 0.6m for 2005 and 0.5m for 2011, however the data included only three visible bands, necessitating a mostly human-interpretive approach as opposed to a digital image processing approach. That said, the 2011 imagery was geo-referenced but not enhanced, meaning that it needed to be enhanced prior to interpretation. This was one of the more challenging processing phases, because the apparent mining sites (confirmed by sending images to the Mozambique partners) included bright/sandy soils that tended to become over-saturated when ‘standard’ ArcGIS (or PCI) image enhancements were applied. Therefore, a methodology was developed that 1) extracted the geo-referencing information from the raw images, 2) adjusted the images in Photoshop CS5, including creating a new background mask and 3) re-applied the geo-referencing information to the new, adjusted images.
Once the imagery was adjusted, all the image tiles were systematically examined and delineated any (potential) mining sites with circular features that encompassed the sites and provide a general indication of the existence of the sites. This method was chosen because it was difficult to exactly delineate the boundaries of these ‘messy’ areas by for example drawing a polygon. It is important to note that these areas indicate the general not exact locations of the mining activity and should be used as such. Indicators of mining activity included ‘pock-marks’ from apparent mine excavation, disturbed soil and even trail patterns. Some areas appeared to be sediments washed downstream from the mining activities, so they were not delineated unless there were other indicators of disturbance.
It should be noted that since almost all this activity seems to be along rivers, it would be good to eventually obtain watershed boundaries for the rivers and then make maps that show which watersheds/sub-watersheds have mining activity.
The map scale used for the general examination ranged from 1:5000 to 1:4000 - -in suspect areas usually zoomed to about 1:1200-1:2000 for a closer look and for creating the delineation feature. It's not clear if the additional processing created technical issues with the data, but a screen refresh at the higher zoom levels in ArcMap 10 literally took 5 minutes or more, despite building multiple pyramid levels. Consequently the 1:4000 scale was determined as the preferred working scale. Figure 2 shows an example of a mining area and the circular delineation.
Figure 2: Example of Mining Areas Delineation
After examining all of the 2011 imagery the 2005 imagery was also examined to determine which sites existed then. A site that existed in 2005 was also found but seems to have disappeared by 2011. This comparison could not cover the exact same area as the 2011 because 1) the images were not available and 2) some cloud cover in the 2005 prevented comparison. The attribute table in the Potential Mining feature class contains attributes that represent all of these cases: Exists in 2011 and 2005, Exists in 2011 but not 2005, Exists in 2005 not 2011, Exists in 2011 but could not be checked for 2005.
Next, the examination in the NW corner of the Reserve began. No possible mining sites were found in that area, but many groups of structures and other evidence of settlements such as agricultural fields were found there. A feature class that delineated those areas was then created. It was also possible to discern trails and roads. However, those were not delineated for this project but it might be useful and interesting to do so in the future. Figure 3 shows an example of the settlements.
Figure 3: Example Settlement Delineation
As noted, the results revealed: 1) possible mining areas that were there in 2011 and 2005, 2) those that were present in 2011 but not in 2005, 3) 2011 areas that we can’t check for 2005 and, 4) Present in 2005 but not 2011. Figure 4 is a screenshot indicating those categories. Below are the statistics for these cases:
These numbers are tentative until accepted by our
Figure 4: Overview with Mining Results
Results & impacts
Chimanimani National Reserve is one of the most pristine Protected Areas in
The results from this study are important to support the existing data the Reserve has on the main areas affected, and to allow the Reserve Management to draft an appropriate strategy to address and try to bring down the level of this threat.
The support made by GISCorps to Chimanimani National Reserve has been of great value. From the work of the recruitment team in finding the best volunteer for this task, their extra efforts in getting additional information and always keeping the project on the flow, to the dedication, expertise, and patience of Kerry Brooks. We are very grateful and it has been pleasant working with such efficient team and we look forward for future collaborations.