Droughts are an increasing global concern due to increased agricultural land-use pressures, rapidly increasing populations and the effects of climate change. One key effect of droughts is the reduction in water available to plants putting them under water stress. This is difficult and time-consuming to map at ground level, and it is therefore difficult to get a synoptic overview of water stress conditions in plants across a state, region, country or continent.
Satellite imagery can be used to produce a product called the Normalised Difference Water Index, (NDWI) which uses a simple calculation to provide a measure of the water content in vegetation. This can then be combined with land cover maps (also available from satellite imagery) to predict the level of water content expected in the vegetation at that time of year, and then map anomalies, where negative anomalies will be evidence of water stress.
At present, no operational NDWI product is produced from any satellite, and no products are available with NDWI taken through to water stress, so this would provide a new resource to a number of communities of users (including academics), as well as achieving developmental goals.
The system would involve:
- An automated system which downloads MODIS data, processes it to NDWI and stores it in a suitable format.
- An 'intelligent' processing algorithm which combines the NDWI data with Land Cover information, compares to data in a lookup table and produces maps of water stress.
- A number of interfaces to both the NDWI and Water Stress data including some of the following:
- Web-based interface with good usability (draggable maps etc) - showing overlays on top of Google Maps/Open Street Map/Other Mapping Sources
- FTP download of individual processed images
- Subsetting interface to allow data for a specific region/country/continent to be easily downloaded
- API to allow other interfaces to be built (eg. SMS-based mobile services)
Example 1:
An aid agency is concerned about issues with agriculture in sub-Saharan Africa due to forecasts for drought. Using this system they can monitor, in near-real-time, the water content and water stress of plants across their region of interest, which can then inform their decision-making.
Example 2:
A farmer is concerned about the water stress in plants over his (large) farm. He can use this system to get quantitative information about the water stress, and then use this to inform his usage of irrigation - thus saving water which can then be used for other purposes.
To produce an operational system, a free source of satellite imagery would have to be used. A number of satellites provide free imagery - including the MODIS sensor on Terra/Aqua (run by NASA), and the Landsat 5 and 7 satellites. Landsat data has a higher resolution, and thus would provide more detailed data, but the continuity of Landsat data is under question (the current satellites are failing and a replacement has yet to be launched), and using Landsat data would require mosaicing many images together. MODIS, on the other hand, provides lower resolution data - ie. with less detail - but has far larger images, is easy to mosaic, and is likely to be operating for many more years.
As I see it, the project splits into a number of sections, listed below in decreasing order of importance (ie. number 1 should be done first, and is most important)
- Development of operational NDWI product (involves acquiring data, processing to NDWI, mosaicing, storing somehow)
- Development of methods to access this product - eg. draggable maps, FTP download, subsetting interface, interface to GIS
- Development of algorithm to produce water stress index from NDWI and ancilliary data - along with (very importantly) a document describing the algorithm!
- Development of more mature methods to access the product
Any progress down this list would be very useful - so anything beyond number 1 could be considered 'extra credit'.
A wide range of data products are available operationally from satellites - see for example http://modis.gsfc.nasa.gov/data/dataprod/index.php for MODIS. Adding NDWI to these products would be a valuable exercise.
A number of useful resources are:
- The MODIS Web Service (http://daac.ornl.gov/MODIS/MODIS-menu/modis_webservice.html) for downloading images
- The MODIS ftp site for the same - but in a different way
- A number of packages for various programming languages may be useful including SpectralPython and the 'raster' package for R
- Some sort of GeoDatabase may be useful to store the data in (eg. PostGIS?)
Information on the science of it all - NDWI and various other products is given in the following articles. Acces should be available through the university, but I will bring print outs with me:
- http://www.sciencedirect.com/science/article/pii/S0034425796000673 - defines NDWI
- http://www.preventionweb.net/files/1852_VL102119.pdf - applies NDWI and NDVI to assessment of drought, and produces new index of NDDI
- http://res.mesonet.org/~jbasara/UT/Gu_et_al_2008.pdf - similar to the above
- http://www.sciencedirect.com/science/article/pii/S0034425703001895