1

I am trying to match rainfall data from USGS/FEWS to coordinate locations of survey households in Ethiopia, so I will need to be able to project the rainfall data. So far, I downloaded the dekadal RFE 2.0 files for East Africa (bulk by year), and for each dekadal, there are five files: .bil, .blw, .clr, .hdr, and .stx. I first added the data (the .bil file) to QGIS, using Africa Albers Equal Area Conic as Projected Coordinate Systems. And then, I opened CSV data of village latitude and longitude and saved it to Africa Albers Equal Area Conic.They looked fine and I could extract point values using point sampling tool.

However, there was wide dispersion from the precipitation data provided by survey. The survey uses the same USGS/FEWS rainfall data as I did, but uses household location rather than village location information. Since dispersion is so wide, I suspect that dispersion came from CRS I chose, rather than household/village level difference. From the USGS/FEWS website, I know following projection information:

 Coordinates for corners:

 Lower left lat : -42.243 deg
 Lower left lon : -23.490 deg
 Upper left lat : 43.711 deg
 Upper left lon : -24.600 deg
 Lower right lat : -42.242 deg
 Lower right lon : 63.414 deg
 Upper right lat : 43.712 deg
 Upper right lon : 64.523 deg

 Image size : 1152 rows x 1152 cols
 Center lat,lon : 1.000000, 20.000000
 Pixel size h x w : 8.000000 km x 8.000000 km
 Origin of latitudes : 1.000000 deg
 Central meridian : 20.000000 deg
 First std parallel : -19.000000 deg
 Second std parallel : 21.000000 deg
 Projection = ALBERS Conical Equal-area projection uses the clarke 1866 spheroid

Using this information, I am trying to custom projection as follows:

+proj=aea +lat_1=-19 +lat_2=21 +lat_0=1 +lon_0=20 +x_0=0 +y_0=0 +datum=NAD83 +units=m +no_defs +ellps=clrk66 

Can anybody confirm whether this projection is correct? I am beginner of QGIS user and only know the following basic parameters.

  • lat_1 : First std parallel
  • lat_2 : 2nd std paralell
  • lat_0 : Origin of latitudes
  • lon_0 : Central meridian

But I am not sure about +datum and +units.

  • It might help if you include a download link to the source data. – AndreJ Feb 11 '17 at 8:04
  • Thank you for your suggestion. I downloaded the East Africa rainfall estimate from USGS/FEWS – Satoko Feb 11 '17 at 8:28
1

I downloaded some East Africa rainfall data and set that projection for it in QGIS. It was definitely in East Africa but seemed to reach further west than in the red box outlining the data area on the web page. So I downloaded a rainfall file for Africa as a whole, and set the same projection for it. This lines up exactly with the East Africa raster, and seems to cover Africa correctly. So I think its correct.

This image shows the Africa rainfall (blue-green) and the East Africa rainfall raster (grey) over an OpenStreetMap world. The extent of the Africa raster looks correct, and I can toggle the East Africa on and off and it matches the underlying continental rain exactly. So if Africa rain is in the right place, then East Africa rain is in the right place.

enter image description here

The corners of my Africa raster don't match the ones given above, but then this raster is 994x1089 and not 1152x1152, so I suspect that refers to a larger raster.

I would suggest you proceed with caution, and if possible keep looking to try and find a definitive description of what the file's metadata are. Also, my QGIS map looks almost exactly like the rainfall map in the "Product" section (except it is cropped to the land), which further makes me think you've got this right.

enter image description here

So the African rain is in the right place, so therefore the East African rain raster is in the right place. Just apply that projection to the raster and you should be good to go.

0

@Spacedman, thank you very much for exploring actual data. I hope what I did is fine.

I re-did all the process using projection I made above:

  1. Add dekadal rainfall data as raster layer using projection I customed.
  2. Add Ethiopia administrative data as vector layer. I downloaded it from the website of Global Administrative Areas. This layer appeared as a small point on QGIS screen because Ethiopia administrative data is provided as WGS84 projection.
  3. Save Ethiopia administrative data to the projection I customed.
  4. Add village longitude and latitude csv data as delimited text layer using WGS84.
  5. Save village longitude and latitude data as ESRI shapefile to the projection I customed.
  6. Extract rainfall data using point sampling tool. Save the newly created layer as csv.

And I got exactly the same attribute table I got with Africa Albers Equal Area Conic projection. So, I do not know why the rainfall data I extracted do not match the survey rainfall data. Is my entire procedure correct?

Below is the table I made with survey data and extracted rainfall data; h2011_tot (annual total rainfall of 2011 from the survey) and total2011 (sum of all the 36 dekadal data I extracted as above). The survey data contain 2011's annual total rainfall as well as village level longitude and latitude information which I used above. I wanted to produce 2012's annual total rainfall but I first made 2011's annual total rainfall to confirm it matches with survey rainfall data. And I found they do not match each other.

As we can see below, there is wide dispersion between h2011_tot and total2011, although their means and standard deviations seem very close to each other. Do you have any idea why h2011_tot and total2011 look so different? survey rainfall vs. extracted rainfall

statistics of two

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.