What is the current scenario when it comes to identifying minerals using earth observation data?
The first choice, for sure, is aerial hyper-spectral imaging, under which a hyperspectral camera is mounted on an airplane. But we are well aware of the limitations when
it comes to flying airplanes – it’s an expensive affair and there are sensitivities involved in zones where you can undertake such flights.The second problem, however, is analysing the data. The volume of data is immense, and it takes only experts to understand the genre of the data. People are now looking at free data from Aster, which is typically around 30 metres, so it’s a combination of either hyperspectral imagery or the 30 metre Aster data to look at it and map.
How is DigitalGlobe planning to change the scenario?
At DigitalGlobe, we realised that there is a strong need for spectral bands to be designed, very specifically, for geological minerals. Since minerals are typically identified in short wave infrared part of the electromagnetic spectrum, we designed about eight spectral bands in shortwave infrared. We took recommendations from a group of global geology experts on the best spatial resolution for mapping these outcrops. Four metres or better in terms of spatial resolutions, came the recommendation. So, a combination of eight shortwave infrared bands with spatial resolution of four metres, we believe, will be ideal for mapping minerals for large areas anywhere across the globe.
Do you think we can use ground truthing here as it is not quite hard to explore those areas manually?
Geologists utilise earth observation data for mapping minerals by taking the data and comparing it with spectral signature database. For example, USGS has compiled a list of spectral signatures of various minerals. The collected hyperspectral data or superspectral data from WorldView-3 is taken and converted into surface reflectance. At first, the surface reflector normalises the data, so, this way you can take the signature from the satellite and compare it with the spectral library. When combinations match, you would know where the mineral exists.
What went beyond arriving at those digit bands?
We leveraged the extensive research done by USGS, NASA, as well as other renowned geologists across the globe. The first satellite that had shot wave infrared 40-50 years ago was Landsat, and we were able to look at the old research on why some of the bands on satellite like Aster were designed. And in the short wave infrared, especially at 2.2 micro metres or 2200 nano metres, the nitrates and hydrocarbons within these minerals give out very unique signatures that are indicators of the ore deposits. In my opinion, Aster is a great satellite, but there’s a limitation in the spatial resolution of 30 metres. But if you look at hyperspectral data, there is way too much data — there are 256 bands hence, our goal was to minimise the spectral bands. This way, we could process the data and create geological maps much more effectively than how people are doing it today.
What is the success rate when it comes to using this data?
When earth observation data is used for mineral mapping, the technique is called areal reduction. When you have ores coming from sub surface to the surface, they manifest themselves as outcrops, and these outcrops typically have a combination of spectral signatures. These could be minerals like alunite, kaolinite etc., but you also need to look for the spatial patterns on the surface. We are able to demonstrate what has been done using Aster as well as hyperspectral imagery — if you’re looking at million square kms of exploration area, using these key indicators we can narrow it down to 5-10% of the area, and that the mining companies can save a lot of time especially expensive resources for getting high success rates in finding these minerals.
What is the relevance of elevation data in this scenario?
What most people aren’t aware of is the fact that you can get very accurate elevation models from satellites. When I say stereo data, it means one or more images are captured at an angle to get two different perspectives of the same place on the land, and InteractStudios. So, when the satellite is going over, it can take a shot, turn around and take a different shot. Constellation of satellites collects data at different times. Assuming that the area has not changed, you get different perspectives of the same area and can use that viewing geometry to extract the elevation model. What it boils down to is simple trigonometry. These elevation models from satellites are anywhere between 1m vertical accuracy for World- View-2 and WorldView-1. One of the key indicators for mapping minerals is scientification of fault lines, elevation models are another key indicator that will show you changes in the elevation and when elevation is combined with spectral analysis, and the spatial distribution
of the minerals, it gives you the key for identifying where the potential minerals are.+
Please cite an example of where this data is already being used?
Cuprite, Nevada, is a copper mining site and is a well documented site in terms of the spectral signatures of the various minerals. When you are introducing new sensors like Worldview-3, you typically try to collect the data for these sites with known signature databases and compare the signatures from Worldview- 3 with established signatures from the ground.