Tag Archives: Remote sensing

Developments in remotely sensed data for geology

Over several decades remote sensing – the interpretation and analysis of image data – has become a central part of many geologists’ ‘toolkit’. It continues a ‘tradition’ founded in the interpretation of panchromatic (black and white), stereoscopic aerial photographs that began after World War 2. But after 1972 and the launch of the first Landsat platform, it has been served by more synoptic views from space using a variety of systems that produce data in many wavelengths of EM radiation, thereby providing opportunities to study spectral properties of the Earth’s surface. This imagery also possesses the analytical flexibility afforded by being recorded in digital form. Since the 1986 launch of the first SPOT platform digital stereoscopic potential from space entered the options for geological interpretation. The Landsat Thematic Mapper (TM) launched in 1982 expanded the spectral range of data. Previously that had been restricted to the visible and near infrared (VNIR) affected mainly by living vegetation and the iron oxy-hydroxides that are the main colorants of rock and soil and TM added a shortwave infrared (SWIR) band. Natural reflectance spectra in that region are affected by Al-OH, Mg-OH and C-O bonds in various hydroxylated silicates and carbonate minerals. The data from TM and its successor the Enhanced Thematic Mapper (ETM) resulted in an explosion of effort into lithological mapping and structural analysis. The last depended on a step-change in resolution to 15 m in the panchromatic band of the ETM system since 1993, together with 10 m stereoscopic resolution from the SPOT family, that enable confident mapping at around 1:100 000 to 1:50 000 scales.

The ETM, its successor on Landsat-8 in 2013 – the Operational Land Imager (OLI) – and the somewhat similar ESA Sentinel-2 system (2015) suffer from one major frustration. Their single broad SWIR band is unable to discriminate –OH and C-O spectral features and hence the lithologically useful range of hydroxylated silicates and carbonate mineral spectra. Also missing from the spectral ‘toolkit’ was any data relating to the major rock-forming silicates. Both drawbacks were remedied to some extent by the launch in 1999 of the Japan/US Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER). As well as the VNIR in three bands, including a stereo-image pair, this covered the mineralogically useful SWIR with 6 narrow  wavelength range bands imaging and 5 bands in the thermally emitted infrared (TIR) where common silicates show substantial spectral differences. ASTER produced primarily geoscientific data that have been found to be of enormous use in geological and mineralogical mapping at the 1:100 000 scale.

Nowadays all the data types mentioned so far, except SPOT, are available for download free of cost from the Earth Explorer site operated by the US Geological Survey (use the Data Sets tab at the EE home page): a superb resource that would suit most geological applications. Yet none of these data have spatial resolution better than 10 m. The commercial Earth observation sector has mainly focussed on increasingly finer spatial resolution, mainly panchromatic and the VNIR range of wavelengths that yield information on vegetation and surface topographic and cultural detail, for which there are many profitable markets. Apart from the follow-on to SPOT – the Pléiades system with resolution as fine as 0.5 m – data from a whole constellation of once independent hi-res systems (WorldView, Quickbird, GeoEye, IKONOS and OrbView) are now administered by one vendor Digital Globe. The finest resolution currently available publically is that of WorldView-3 (0.3 m), beyond which is the classified purview of the US intelligence community. The figure illustrates just how much more detailed geological information there is in the finest resolution data than in the same kind of image reduced to 15 m resolution, the best offered by ASTER. That detail needs to be tempered by a few facts: by comparison with the high-res image ASTER shows a regional context, i.e. large-scale geological structures; it covers more spectral bands and is therefore more revealing lithologically; the highest resolution data (WorldView-3 archived) are priced at US$14 to 19 per km2  for each of 6 different band-bundles with a minimum order of 25 km2. Note: for some areas Google Earth has coverage at high-resolution captured at several dates, though some remain at 15 m resolution (based on Landsat-7 ETM).

30cm v 15m

An area in Utah, USA, with almost 100% exposure and very low vegetation cover shown by simulated natural colour images at ~0.3 m with a scale of ~1:1225 (top) and ~15 m at ~1:61275. Credit: Google Earth

The geologist’s dream data would, I suppose, consist of many bands that divide the VNIR, SWIR and TIR into narrow wavebands so that rock and soil spectra can be accurately reproduced, thereby allowing considerable discrimination between different rock types and their main constituent minerals. Oh yes, and it would have decent resolution – better than 15 m. There is indeed such a hyperspectral instrument called CRISM and data from it can be downloaded freely but, before there is a stampede to get access, note that the acronym stands for Compact Reconnaissance Imaging Spectrometer for Mars! For the Earth most hyperspectral data are captured from airborne missions, except for one orbital mission that occasionally functioned over a tiny fraction of the Earth from 2001 to 2017 – NASA’s EO-1 Hyperion system that produced 7.5 km swaths at 30 m resolution with 220 spectral bands covering the VNIR and SWIR regions. Apart from one aimed at oceanic and atmospheric issues, that will say little about rocks, NASA and ESA have no plans in this niche. One commercial developer, Satellogic of Argentina, has hyperspectral plans but only where an income stream is guaranteed, which seems to be just for crops and vegetation spanning the VNIR range. Other outfits have wish lists but few concrete plans in the geoscientific spectral range.

With pending budget cuts to NASA’s Earth science programme (9%), NOAA (22%) and the USGS (14%) demanded by the Trump administration, progress with US contributions to Earth observation can’t be anticipated with much hope. Commercial interests have to pay the shareholders and their dominant focus is on government intelligence agencies, the media, private weather forecasters and agribusiness. So do not expect another or better CRISM in Earth orbit. But it is possible to get by quite nicely at the reconnaissance, small-scale level of mapping, lithological discrimination and some mineral identification with the moderate resolution 14 spectral bands captured by ASTER. If you have the cash, then WorldView-3 offers similar panchromatic, VNIR and SWIR data options at 0.3, 1.2 and 3.7 m resolution, respectively, that should enable very intricate geological mapping.

You may learn more about geological remote sensing here.

‘Big data’ on water resources

 

Two petabytes (2×1015) is a colossal number which happens to approximate how much data has been collected in geocoded form by the Landsat Thematic Mapper and its successors since it was first launched in 1984. In tangible form these would occupy about half a million DVDs, weighing in at about 8 metric tonnes; ‘daunting’ comes nowhere near describing the effort needed to visually interpret this unique set of multi-date imagery. Using the Google Earth Engine, the free cloud-computing platform for big sets of image data which hosts all Landsat data and much else (but not yet the equally daunting ASTER data – roughly a million 136 Mb scenes) the 32 years-worth has been analysed for its content of hydrological information by the European Commission’s Joint Research Centre in Italy, with assistance from Google Switzerland. Using the various spectral characteristics of water in the visible and infrared region, the team has been able to assess the position on the continents of surface water bodies larger than 900 m2, both permanent and ephemeral, and how the various categories have changed in the last 32 years (Pekel, J.-F. et al. 2016. High-resolution mapping of global surface water and its long-term changes. Nature, v. 540, p. 418-422; doi:10.1038/nature20584). The results are conveniently and freely available in their entirety at the Global Surface Water Explorer, an unparalleled and easy-to-use opportunity for water resource managers, wetland ecologists and geographers in general.

Among the revelations are sites and areas that have been subject to gains and losses in water availability, the extents of new and vanished permanent and seasonal water bodies and the conversion of one to the other. A global summary gives a net disappearance of 90 thousand km2 of permanent water bodies, about the area of Lake Superior, but exceeded by new permanent bodies totalling 184 thousand km2. There has been a net increase in permanent water on all continents except Oceania with a loss one percent (note that Antarctica and land north of the Arctic Circle were not analysed). More than 70 % of the losses are in the semi-arid Middle East and Central Asia (Iran, Iraq, Uzbekistan, Kazakhstan and Afghanistan), due mainly to overuse of irrigation, dam construction and long-term drought. Much of the increase in water occurrence stems from reservoir construction, but climate change may have played a part through increased precipitation and melting of high-altitude snow and ice, as in Tibet.

The Aral Sea in Uzbekistan and Kazakhstan has suffered dramatic loss of standing and seasonal water cover due to overuse of water for irrigation from the two main rivers, the Amu (Oxus) and Syr, that flow into it. Note the key to the colours that represent different categories of changes in surface water. (Credit: Global Surface Water Explorer)

The Aral Sea in Uzbekistan and Kazakhstan has suffered dramatic loss of standing and seasonal water cover due to overuse of water for irrigation from the two main rivers, the Amu (Oxus) and Syr, that flow into it. Note the key to the colours that represent different categories of changes in surface water. (Credit: Global Surface Water Explorer)

Many of the lakes in the northern Tibetan Plateau have grown in size during the last 32 years, mainly due to increased precipitation and snow melt. (Credit: Global Surface Water Explorer)

Many of the lakes in the northern Tibetan Plateau have grown in size during the last 32 years, mainly due to increased precipitation and snow melt. (Credit: Global Surface Water Explorer)

There are limitation to the accuracy of the various categories of change, one being the persistence of cloud cover in humid climates, another being the sometimes haphazard scheduling of Landsat Data capture (in some case that has depended on US Government interest in different areas of the world).

More detail on using remote sensing in exploration for and evaluation of water resources can be found here.

Free course on remote sensing for water exploration

250 million people who live in the drylands of Africa and Asia face a shortage of water for their entire lives. Hundreds of millions more in less drought-prone regions of the ‘Third World’ have to cope repeatedly with reduced supplies. A rapid and effective assessment of how to alleviate the shortfall of safe water is therefore vital. In arid and semi-arid areas surface water storage is subject to a greater rate of evaporation than precipitation, so groundwater, hidden beneath the land surface, provides a better alternative. Rainwater is also lost by flowing away far more quickly than in areas with substantial vegetation. Harvesting that otherwise lost resource and diverting it to storage secure from evaporation – ideally by using it to recharge groundwater – is an equally important but less-used strategy. Securing a sustainable water supply for all peoples is the most important objective that geoscientists can address.

In practice, to assure good quality water supplies to a community in the form of productive wells, surface water harvesting schemes or planning the recharge of exploited aquifers requires skill, a great deal of work and considerable financial resources. Yet in many parts of sub-Saharan Africa and arid areas of Asia knowing where to focus effort and increase the chances of it being fruitful is one the biggest hurdles to overcome. Such reconnaissance – highlighting the most probable localities on geological and hydrological grounds, and screening out those least likely to yield water for drinking and hygiene – depends on details of the geology and topography of the terrain in which needy communities are situated. For most of the Afro-Asian dryland belt adequate geological and topographic maps are in as short supply as potable water itself.  Remote sensing combined with an understanding of groundwater storage and surface-water harvesting is a powerful tool for bridging that knowledge gap, and is routinely used successfully in areas blessed with abundances of experienced geoscientists, money and engineering infrastructure. Again, most of the Afro-Asian dryland belt is poorly endowed in these respects.

dvd-sleeve-front

Having long ago written a textbook on general remote sensing for geoscientists, now out of print (Image Interpretation in Geology (3rd edition): 2001. Nelson Thorne/Blackwell Science), I decided to re-issue revised parts of it framed in the specific context of water exploration in arid and semi-arid terrains, and to add practical case studies and exercises based on a free version of professional image processing and desktop mapping software. Some of the most geologically revealing remotely sensed image data – those from the Landsat series of satellites and the joint US-Japan ASTER system carried by Terra, one of NASA’a Earth Observing System satellites – are now easily and freely available for the whole of the Earth’s land surface. Given basic familiarity with theory and practicalities, a computer and appropriate software together with a moderately fast internet connection there is nothing to stop any geoscientist, university geology student or engineer working in the water, sanitation and hygiene (WASH) sector from becoming a proficient, self-contained practitioner in water reconnaissance. Water Exploration: Remote Sensing Approaches has that aim. Online access to the theoretical parts is free, and a DVD that combines theory, software, exemplary data and several exercises that teach the use of image processing/desktop mapping software is available at cost of reproduction and postage.

If you visit the website, find what you see potentially useful and wish to know more, contact me through the Comments form at the H2Oexplore homepage.

Earthquakes in Nepal

The magnitude 7.8 Gorkha earthquake hit much of the Himalayan state of Nepal on 25 April 2015, to be followed by one of magnitude 7.3 150 km to the east 18 days later. As would have happened in any high-relief area both events triggered a huge number of landslides as well as toppling buildings, killing almost 9000 people and leaving 22 000 injured in the capital Kathmandu and about 30 rural administrative districts. Relief and reconstruction remain hindered 9 months on in many of the smaller villages because they are accessible only by footpaths. Nepal had remained free of devastating earthquakes for almost 6 centuries, highlighting the perils of long quiescence in active plate-boundary areas.

Damage in Kathmandu, Nepal, after the Gorkha earthquake in May 2015 (Credit: CNN)

Damage in Kathmandu, Nepal, after the Gorkha earthquake in May 2015 (Credit: CNN)

The International Charter: Space and Major Disasters consortium of many national space agencies was activated, resulting in one of the largest ever volumes of satellite images ranging from 30 to 1 m resolution to be captured and made freely available for relief direction, analysis and documentation. This allowed more than 7500 volunteers to engage in ‘crowd mapping’ coordinated by the Humanitarian OpenStreetMap Team (HOT) to provide logistic support to the Nepal government, UN Agencies and other international organizations who were swiftly responding with humanitarian relief. Most important was the location of damaged areas using ‘before-after’ analysis and assessing possible routes to remote areas. The US NASA and British Geological Survey with Durham University coordinated a multinational effort by geoscientists to document the geological, geophysical and geomorphological factors behind the mass movement of debris in landslides etc that was triggered by the earthquakes, results from which have just appeared (Kargel, J.S. and 63 others 2016. Geomorphic and geological controls of geohazards induced by Nepal’s 2015 Gorkha earthquake. Science, v. 351, p. 140 – full text purchase).

The large team mapped 4312 new landslides and inspected almost 500 glacial lakes for damage, only 9 had visible damage but none of them showing signs of outbursts. As any civil engineer might have predicted, landslides were concentrated in areas with slopes exceeding 30° coincided with high ground acceleration due to the shaking effect of earthquakes. Ground acceleration can only be assessed from the actual seismogram records of the earthquakes, though slope angle is easily mapped using existing digital elevation data (e.g. SRTM). It should be possible to model landslide susceptibility to some extent over large areas by simulation of ground shaking based on various combinations of seismic magnitude and epicenter depth modulated by maps of bedrock and colluvium on valley sides as well as from after-the-event surveys. The main control over distribution of landslides seems to have been the actual fault mechanism involved in the earthquake, assessed from satellite radar interferometry, with the greatest number and density being on the downthrow side (up to 0.82 m surface drop): the uplifted area (up to 1.13 m) had barely any debris movements. Damage lies above deep zones where brittle deformation probably takes place leading to sudden discrete faults, but is less widespread above deep zones of plastic deformation.

The geoscientific information gleaned from the Gorkha earthquake’s effects will no doubt help in assessing risky areas elsewhere in the Himalayan region. Yet so too will steady lithological and structural mapping of this still poorly understood and largely remote area. As regards the number of lives saved, one has to bear in mind that few people buried by landslides and collapsed buildings survive longer than a few days. It seems that rapid response by geospatial data analysts to the logistics of relief and escape has more chance of positive humanitarian outcomes.

In the same issue of Science appears another article on Nepalese seismicity, but events of the 12th to 14th centuries CE (Schwanghart, W. and 10 others 2016. Repeated catastrophic valley infill following medieval earthquakes in the Nepal Himalaya. Science, v. 351, p. 147-150). As the title suggests, this relates to recent geology beneath a valley floor in which Nepal’s second city Pokhara is located. It lies immediately to the south of the 8000 m Annapurna massif, about 50 km west of the Gorkha epicentre. Sections through the upper valley sediments reveal successive debris accumulations on scales that dwarf those moved in the 2015 landslides. Dating (14C) of interlayered organic materials match three recorded earthquakes in 1100, 1255 and 1344 CE, each estimated to have been of magnitude 8 or above. The debris is dominated by carbonate rocks that probably came from the Annapurna massif some 60 km distant. They contain evidence of extreme pulverisation and occur in a series of interbeds some fine others dominated by clasts. The likelihood is that these are evidence of mass movement of a more extreme category than landslides and rockfalls: catastrophic debris flows or rock-ice avalanches involving, in total, 4 to 5 km3 of material.

Remote sensing for fossils

With the growing diversity of data from those parts of the electromagnetic spectrum that pass freely though Earth’s atmosphere, mainly acquired from orbit, an increasing number of attributes of the surface can be mapped remotely. The initial impetus to launch remote sensing satellites in the 1960’s and early 70’s had two strands: to monitor weather conditions and assess vegetation cover with the early metsats, such as TIROS-1, and the first Landsat platform that exploited green plants’ propensity for absorbing visible and largely reflecting near-infrared (NIR) radiation. With the incorporation in the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) instruments of wavelength regions in which minerals show spectral diversity, in the reflected short-wave infrared (SWIR) and emitted thermal infrared (TIR), remote sensing became a viable and useful tool for geologists. It figures strongly in lithological mapping and also in the detection of minerals related to various kinds of alteration associated with metal mineralisation and the migration of hydrocarbon-related fluids. The more wavebands with narrower coverage of radiation wavelengths, the more likely are the subtle differences in mineral spectra able to be detected and mapped. Yet, apart from one experimental system (Hyperion aboard NASA’s EO-1 orbital platform) our home planet is not as well served by such hyperspectral systems as is Mars, blessed by two which have fuelled the on-going search for past habitable zones on the Red Planet.

The May 2014 issue of Scientific American includes an article on remote sensing that follows what to many might seem an odd direction: how to increase the chance of finding rich fossil deposits (Anemone, R.L. & Emerson, C.W. 2014. Fossil GPS. Scientific American, v. 310(5), p. 34-39). Apart from targeting a particular stratigraphic unit on a geological map, palaeontological collection has generally been a hit-or-miss affair depending on persistence and a keen eye, with quite a lot of luck. Once a productive locality turns up, such as the Cambrian Burgess shale, the dinosaur-rich Cretaceous sandstone of the Red Deer River badlands of southern Alberta in Canada and the hominin sites of Ethiopia’s Afar Depression, palaeontologists often look no further until its potential is exhausted. Robert Anemone and Charles Emerson felt, as may palaeobiologists do, that one fossil ‘hotspot’ is simply not enough, yet balked at the physical effort, time and frustration needed to find more by trekking through their area of interest, the vast Tertiary sedimentary basins of Wyoming, USA. They decided to try an easier tack: using the few known fossil localities as digital ‘training areas’ for a software interrogation of Landsat Enhanced Thematic Mapper data in the hope that fossiliferous spots might be subtly different in their optical properties from those that were barren.

Satellite image of the Wyoming Basin, Wyoming,...

Satellite image of the Wyoming Basin, USA. credit: Wikipedia)

The teeth and bones of early Eocene mammals that had drawn them to Wyoming turn up in sandstone beds of the basins. They are pretty distinctive elements of landscape, forming ridges of outcrop because of their relative resistance to erosion, yet for that very reason present a huge selection of possibilities. Being simple mineralogically they also presented a seemingly daunting uniformity. Anemone and Emerson decided on a purely statistical approach using the six visible, NIR and SWIR bands sensed by Landsat ETM, rather than a spectrally oriented strategy using more sophisticated ASTER data with 14 spectral bands. Their chosen algorithm was that based on an artificial neural network that the fossil rich sandstones would train to recognise patterns present in ETM data recorded over them. This purely empirical approach seems to have worked. Of 31 sites suggested by the algorithm 25 yielded abundant vertebrate fossils. Applied to another of Wyoming’s Tertiary basins it also ‘found’ the three most productive known mammal sites there. So, what is it about the fossil-rich sandstones that sets them apart from those that are more likely to be barren? The authors do not offer an explanation. Perhaps it has something to do with reducing conditions that would help preserve organic material better than would sandstones deposited in an oxidising environment. Iron minerals and thereby colour might be a key factor, oxidised sandstones are generally stained red to orange by Fe-3 oxides and hydroxides, whereas reduced sandstone facies may be grey because of iron in the form of sulfides

Enhanced by Zemanta