Under the direction of Dr Harry Gibson, and previously that of Dr Dan Weiss, the ROAD-MAP Team has accumulated a large estate of covariate datasets, which are used in the statistical models developed by the MAP group to support the generation of malaria maps.
These covariates are held in the form of raster data (images), which divide a map of the world into a regular grid pattern and allocate a value to each “pixel” or cell. In MAP we use a fixed set of resolutions for this, and a fixed representation of the world’s coastline, ensuring that all our different covariate datasets align precisely and that the maps we generate using the covariates are consistent. (For GIS experts – we work in EPSG:4326 coordinates, at spatial resolutions of 30 arcseconds (~1km); 2.5 arcminutes (~5km), or 5 arcminutes (~10km)).
Some covariates represent variables that vary over time, such as temperature. We hold these for time-steps as frequent as 8 days (although generally we produce our models at monthly or annual time-steps). We refer to these as dynamic covariates. Others are “fixed” in time (at least for the purposes of our work), such as elevation; we refer to these as static or synoptic covariates.
Many covariate datasets are acquired from third-party sources such as NASA and the WorldPop project; others are developed in-house by MAP’s scientists. Datasets include land surface temperature, vegetation health indices, precipitation, land cover types, population maps, and night-time lights.
For datasets that are prone to having gaps (for example, remotely-sensed datasets where cloud cover prevents acquisition of data) we fill the gaps using a variety of different algorithms so that the data we store, and thus the models we generate, cover the complete land area represented by our standard coastline template. The most notable of these are the MODIS MOD11A2 and MCD43B4 products (see table below) which are gap-filled using an algorithm developed by Dr Dan Weiss and implemented globally by Dr Harry Gibson
|Daytime LST (Land Surface Temperature)||MODIS MOD11A2||Gapfilling, alignment||30″, 2.5′, 5′||2000-present, 8-daily, monthly, annual, synoptic|
|Nighttime LST||MODIS MOD11A2||Gapfilling, alignment||30″, 2.5′, 5′||2000-present, 8-daily, monthly, annual, synoptic|
|EVI (Enhanced Vegetation Index)||MODIS MCD43B4||Gapfilling, alignment||30″, 2.5′, 5′||2000-present, 8-daily, monthly, annual, synoptic|
|TCB (Tasselled Cap Brightness)||MODIS MCD43B4||Gapfilling, alignment||30″, 2.5′, 5′||2000-present, 8-daily, monthly, annual, synoptic|
|TCW (Tasselled Cap Wetness)||MODIS MCD43B4||Gapfilling, alignment||30″, 2.5′, 5′||2000-present, 8-daily, monthly, annual, synoptic|
|IGBP-classification Landcover||MODIS MCD12Q1||Alignment, reclassification||15″, 2.5′, 5′||2000-2013, annual|
|Night-time lights||VIIRS||Alignment||15″, 2.5′||2014-present, monthly, annual|
|Population||WorldPop and GPWv4||Reallocation to template, normalising to alternate totals||30″, 2.5′, 5′||1980-present, annual|
|Accessibility to cities||MAP||Developed by MAP||30″||2015, nominal|
|Precipitation||GPM (IMERG product)||Developed by MAP||30″||2014-present, monthly, annual|
This table shows some of the main covariate datasets we use in MAP
We can provide our covariate data on request – although due to the volumes of data involved the full resolution dynamic datasets are not distributed online (a single global 30″ image can be several GB in size). Please contact us if you would like to obtain any of our covariate data, with details of what you require. Data that have not been gapfilled or generated by MAP can equally be obtained from the original suppliers, all of which are linked in the table above.