Web maps are usually served by smaller tiles which are loaded separately but gives an impression of a seamless image. These are the tiled based maps. Check out the earlier blog post on Converting Satellite imagery to Tile Map Service. Usually, tiles are small in size to get loaded. But things might get tricky if we want to use the offline maps in mobile version, or the tiles are huge which would make it slow and inefficient. One way of overcoming this problem is by using MBTiles.

MBTiles is a format given by MapBox, for storing the tilesets. It’s an open specification, for storing large vector or raster data, and is based on SQLite database. The tiles are compressed and compact, and as a result, makes it easier to transfer and carry around and used in the mobile.

In this post, we will be converting the raster tif image into the mbtiles format using the open source powerful GDAL library. Check the correct installation of the GDAL using gdalinfo.

  1. Get your raster data. I downloaded the landcover maps for Laos using the SERVIR-Mekong Regional Land Cover Monitoring System for year 2018. This is a single band image with unique values for each class. The class number and the assigned color looks like:

     Class Name Class Value Class Color
     Unknown 0 6f6f6f
     Surface Water 1 aec3d4
     Snow and Ice 2 b1f9ff
     Mangroves 3 111149
     Flooded Forest 4 287463
     Forest 5 152106
     Orchard or Plantation Forest 6 c3aa69
     Evergreen Broadleaf 7 7db087
     Mixed Forest 8 387242
     Urban and Built Up 9 cc0013
     Cropland 10 8dc33b
     Rice 11 ffff00
     Mining 12 cec2a5
     Barren 13 674c06
     Wetlands 14 3bc3b2
     Grassland 15 f4a460
     Shrubland 16 800080
     Aquaculture 17 800080


  2. If you don’t have a color relief raster, you can create it. Unfortunately, the land cover map is not a color relief raster. Creating a color relief in the band is quite easy. We will be using gdaldem. Make a txt file called color and have the raster value and the required color (the color should be the name of the color as red, green, yellow or the actual RGB color like 111 111 111. The content of the color.txt file looks as:
    0           111 111 111
    1           174 195 212
    2           177 249 255
    3           17 17 73
    4           40 116 99
    5           21 33 6
    6           195 170 105
    7           125 176 135
    8           56 114 66
    9           204 0 19
    10           141 195 59
    11           255 255 0
    12           206 194 165
    13           103 76 6
    14           59 195 178
    15           244 164 96
    16           128 0 128
    17           81 118 142

  3. Next, generate the color relief map as follows:
    gdaldem color-relief landcover.lc.tif color.txt landcover-output.tif
    where landcover.lc.tif is input file, color.txt is the value and color that needs to be applied and landcover-output.tif is the output file.

  4. If you have a color table, you may want to expand your band to data with 3 band (RGB) or 4 bands (RGBA) as:
    gdal_translate -expand rgba  landcover-output.tif landcover.tif
  5. Since MBTiles works with Spherical Mercator projection, let’s reproject the raster using gdalwarp as:
    gdalwarp -t_srs EPSG:3857 -r cubicspline landcover-output.tif landcover.tif
    where r is the resampling method and landcover.tif is the output file.

  6. Create an MBTiles file from the result of step 5 using gdal_translate as:
    gdal_translate -of mbtiles landcover.tif landcover.mbtiles
  7. The last step is to create overview images at different lower zoom levels using gdaladdo.
    gdaladdo -r cubicspline landcover.mbtiles 2 4 8 16

Next, you can get your offline map tiles from above steps into the ODKCollect. For loading your data into ODKCollect, follow the steps listed which should be fairly straight forward.

Happy Geoinformatics 🙂