Malaria Atlas Project spatiotemporal covariate cubes


Raster digital data


These datasets were produced by Harry Gibson and Daniel Weiss, of the Malaria Atlas Project (Big Data Institute, University of Oxford, United Kingdom, http://www.map.ox.ac.uk), to provide independent variables for spatiotemporal disease modeling. The datasets currently available include daytime Land Surface Temperature (Day_LST), night-time Land Surface Temperature (Night_LST), delta Land Surface Temperature (i.e., the difference between daytime and night time temperature) (Delta_LST), Enhanced Vegetation Index (EVI), Tasseled Cap Wetness (TCW), and Tasseled Cap Brightness (TCB). The data are derived from MODIS products, as described below, but have been processed to fill in data gaps caused by cloud cover.


March 2000 to (currently) December 2014 (LST); February 2000 to (currently) December 2014 (EVI / TCW / TCB)


Weiss, D.J., P.M. Atkinson, S. Bhatt, B. Mappin, S.I. Hay & P.W. Gething (2014) An effective approach for gap-filling continental scale remotely sensed time-series. ISPRS Journal of Photogrammetry and Remote Sensing, 98, 106-118.


Daniel Weiss or Harry Gibson

ADDRESS: Malaria Atlas Project
Big Data Institute, University of Oxford
Roosevelt Drive, Oxford, OX3 7LF, UK


Initially updates will be every 3 to 6 months, with a goal of eventually automating this process for more frequent and rapid updates following the release of the source MODIS data.


This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.

Users assume all risks in using these data.


The underlying datasets for the available products are MODIS LST data (MOD11A2) and the MODIS BRDF-corrected product (MCD43B4), the latter of which was used to derive the EVI and tasseled cap indices. The MODIS data have a native spatial resolution of ~1 km. The LST dataset consists of 8-day composites, and thus has a nominal 8-day temporal resolution. The BRDF dataset, in contrast, consists of 16-day composites, but an 8-day temporal resolution is achievable if products from the MODIS sensors on both the Aqua and Terra satellites are utilized (i.e., the 16-day composites are temporally offset such that one is centered temporally each 8 days). The MODIS data were acquired from LP-DAAC (Land Processed Distributed Active Archive Center – https://lpdaac.usgs.gov/products/modis_products_table) in HDF tile format. These tiles were stitched together and reprojected to form seamless global 30 arc-second resolution mosaics in WGS-84 for the requisite spectral bands. In the case of the BRDF imagery the single band mosaics were then used to derive the EVI and tasseled cap indices according to the equations published in Huete et al. (1999) and Lobser and Cohen (2007), respectively. The mosaics were then gap-filled using the approach outlined in Weiss et al. (2014) to eliminate gaps in the dataset caused by factors such as cloud cover. The gap-filled 8-day products were then aggregated to produce monthly ~1 km products before finally being spatially aggregated to produce ~5 km resolution outputs. Quality control data are available for each output layer in the form of a second raster layer that defines the percentage of each resulting ~5 km pixel that was original data (as opposed to gap-filled estimates).


The model used for producing these products differs slightly from the model published in Weiss et al. (2014). Changes include:

  1. The EVI, TCB, and TCW datasets were derived using reflected wavelengths of energy from the sun, and therefore can’t be reliably calculated in high latitudes during polar winters. As such, gap-filled EVI, TCB, and TCW data were cropped in months when sunlight was not universally available. The cutoff latitudes were as follows: January 60oN, February 68oN, March through September 80oN, October 68oN, November 62oN, December 60oN. The southern limit was clipped to 60oS in all months.
  2. The EVI, TCB, and TCW results were also impacted by topographic shadowing during winter months in very rugged areas (e.g., north-facing slopes in a few, very steep sided, east-west oriented canyons found in central Asia). This initially caused the annual mean values to be skewed in affected pixels, which in turn impacted the quality check and gap-filling algorithms. To account for this issue we identified affected pixels that would be in shadow at local noon during winter months using a digital elevation model, and then further tested seemingly reasonable raw values from these pixels for validity. In cases where the calculated mean EVI values in the winter months were higher than those from spring or autumn we discounted the winter values from the calculation of the overall temporal mean.
  3. Increases in processing efficiency enabled the A1 model described in the Weiss et al. (2014) paper to be run to a distance of 30 km and thus be used to gap-fill a higher proportion of cells than was presented in the manuscript. The A2 algorithm was still required, however, for filling cells far from their nearest usable neighbor.


  1. The LST data should be considered “clear sky” land surface temperature (i.e., the mean temperature present on occasions without clouds).
  2. The EVI, TCB, and TCW products have small percentage of error cells during winter months, at latitudes in excess of 30 degrees north or below -30 degrees south, and in areas of high topographic relief that are caused by topographic shadows. The majority of these artifacts were removed from the EVI data by the process described above; however some may remain.
  3. The TCW and TCB data may be spurious in areas where the land was covered by snow or ice.


  1. Day_LST, Night_LST, and Delta_LST are all in degrees Celsius.
  2. EVI, TCB, and TCW are unitless.


(of data as gapfilled – source MODIS data are in a sinusoidal projection and Google Earth Engine data are in Google Maps projection):

  • Resolution: 0.0418 x 0.0418 degrees
  • Horizontal_Datum_Name: World Geodetic System of 1984
  • Ellipsoid_Name: World Geodetic System of 1984
  • Semi-major_Axis: 6378137
  • Denominator_of_Flattening_Ratio: 298.25722210088




  • Huete, A., Justice, C. & Van Leeuwen, W. (1999) MODIS vegetation index (MOD13). Algorithm theoretical basis document.
  • Lobser, S.E. & Cohen, W.B. (2007) MODIS tasselled cap: land cover characteristics expressed through transformed MODIS data. International Journal of Remote Sensing, 28, 5079-5101.
  • Weiss, D.J., Atkinson, P.M., Bhatt, S., Mappin, B., Hay, S.I. & Gething, P.W. (2014) An effective approach for gap-filling continental scale remotely sensed time-series. ISPRS Journal of Photogrammetry and Remote Sensing, 98, 106-118