Remote Sensing Methods for the detection of UXOs in Urban Environments
Image 1 - A map showing the outcomes of radar multi-temporal change detection in Syria between 2015-2020. The red dots are reported bombardments, the yellow and blue are the detected changes.
Earlier this month, Space4Good participated in the UN Mine Action Service (UNMAS) and Geneva Centre for Humanitarian Demining (GICHD)’s 8th Mine Action Technology Workshop in Geneva. Attended by mine action practitioners, researchers and policy makers, the workshop facilitated the sharing of ideas and technological developments. As part of the event, Space4Good presented our work with the HALO Trust and the Carter Centre on the use of earth observation data to aid in identifying unexploded ordnances (UXO) in urban environments. Building upon the piece we co-wrote with our friends at the Conflict and Environment Observatory (CEOBS) earlier this year on Earth Observation and Humanitarian Disarmament, we are delighted to further share our work on this important topic.
Why Earth Observation for Urban Mine Action?
Conflict has become increasingly urbanised over the past decade. Across the Middle East and North Africa, high volumes of remotely launched weapons are being fired into densely populated areas - with ordnance detonation failure rates of 10% and above still being reported (Humanity & Inclusion).
Space4Good have devised approaches combining open-source radar (Sentinel 1) and optical (Sentinel 2) data from the European Space Agency's Copernicus satellite mission. Based on this, monitoring and confirming reported bombardments for preventing the use of Explosive Weapons in Populated Areas (EWIPA) becomes easier. Data retrieval can also be achieved without entering conflict or post-conflict environments. This is particularly useful for mine action organisations who operate in insecure environments where good in-situ data is hard to come by, and budgets are often tight.
HALO Trust in Libya, 2019
In 2019, we met the HALO Trust at a community event in the Hague. From there, we explored techniques to detect damages to buildings in the town of Sirte, Libya following conflict in 2016. The first approaches offered optical change detection analyses, and radar amplitude differencing. Promising results emerged with changes detected in both cases. However, we believed we could generate more informative results. We then used Google Street Map to generate neighbourhood clusters, and observed changes of the high-resolution panchromatic data. As is shown in images 2 and 3, the outcome of this was to show changes on a scale of blue to red, where red indicated the highest level of change.
Images 2 (left) and 3 (right): These maps show changes to the panchromatic band within Google Street Map neighborhood clusters. Blue is low change, red is high.
Higher resolution panchromatic data with neighbourhood clusters provided a better view of smaller-scale conflict-affected areas. It also was an easily understandable visual tool for mine action practitioners. Unfortunately, due to a worsening security situation, HALO Trust was forced out of the area. As such, we were not able to fully discern the precise utility of each of these methods, however, we were able to gain an indication of the usefulness for remote sensing in conflict incidents.
Carter Center in Syria 2020
A year later, we met the Carter Centre at a Data for Peace and Security conference. The Carter Centre was aiming to build a database of conflict incidents in Syria using open source data. As such, they were curious about the use of earth observation to verify, assess, and monitor reported bombardments from 2015-2020.
For this, we used a large dataset of Sentinel 2 images over the 6 year period of interest to apply pixel-based change detections. By recognising patterns, seasonal differences, and noise, only the most relevant changes were represented. Additionally, we were also able to observe a reasonable level of correlation between reported bombardments and changes to land use.
Image 4: A Map showing radar multi-temporal change detection outputs, yellow dots show successful bombardments, red are unsuccessful, while the yellow areas are the areas of detected change.
We then applied radar multi-temporal change detections across the period of interest. Annual change maps were made and overlaid with the reported bombardments. Placing a 250m buffer around each reported bombardment, we could confirm an ‘effective bombardment’ if total change within this buffer exceeded 1km². As seen above, the area of Aleppo is shown between 2016 and 2017. In this period, 351 bombardments were reported, 242 of which were estimated as being harmful.
What next?
As repeatedly stated at the GICHD event, these developments are far from a ‘silver bullet’. Despite this, our work has shown the potentials of remote sensing to aid the development of effective low-cost methods to support mine action operations. Now, with closer collaborations with other experts and practitioners on the ground, we can begin to realise a safer future.
Would you like more information or are you interested in collaborating with Space4Good? Visit our website or contact us via hello@space4good.com.