I have population data (inhabitants per city and year) of 31 cities of the DR Congo for the years 1960, 1970, 1980 and 1990. I want to interpolate the population of all cities for the year 1990, to estimate the growth of them in order to locate the sites that will minimize the competition for resources.

I have used an IDW and a layer of existing lakes as a barrier. However, it seems to me that the resulting layer does not have sense, and so I would like to ask if IDW is the best interpolation tool for my data or it will be better to use other, such as Kriging.



I have to create a protected area in the DR Congo, so I want to estimate the possible growth of the cities in order to find the more suitable place to avoid competition for space and resources.

enter image description here

  • If you already have the population for 1990, why do you need to interpolate? Are you saying you have a population for 31 x,y locations, but you want to model the population density in between the cities for the year 1990? Please provide the sample IDW result so we can see what it looks like.
    – Taylor H.
    May 7, 2014 at 21:55
  • @TaylorH. Sorry, my mistake. I have update the question. Hope it will be clear now.
    – Krahsin
    May 7, 2014 at 22:16
  • No worries, I'm out right now but will give you an answer later if no one has responded.
    – Taylor H.
    May 7, 2014 at 23:03
  • You should look at the WorldPop website and have a look at their methodology.
    – Hornbydd
    May 8, 2014 at 11:31
  • @Hornbydd. Thanks for the website. I have found very interesting information and data.
    – Krahsin
    May 8, 2014 at 22:02

2 Answers 2


The short answer is: no, but it might be the best you can do without additional layers. The problem is that humans do not settle evenly. If all you have is populations of cities then IDW might be your best approximation, because population density does decay the further you go from the center of a city (as IDW models). However, population does not decay evenly in all directions. Settlements will stick very close to major transportation routes. So if you were able to get a transportation layer from 1990 for your study area then you could limit your IDW output to a buffer around the roads.

In my opinion, you will need other layers to draw any meaningful conclusions from the IDW.

  • 2
    That is an intelligent and well written answer, I totally agree, one should always be cautious about using interpolated data and if it's not 100% certain it should never be used in models or mapping. If you do interpolate warn the user that the data is not based on fact but supposition so it doesn't come back to bite. May 8, 2014 at 2:20
  • 1
    I would go further : you should not use any interpolation in this case. The rationale for interpolation is that knowing a value close to you will help you guess tha value at your loacation. In your case this is not true.
    – radouxju
    May 8, 2014 at 5:49

In your case (pressure on the resources), I recommend that you use some neighbourhood analysis instead of interpolation. for instance, you can estimate the distance travelled in one day by the people of the city (a simple buffer to start with, or more advanced cost-distance analysis). Then you sum the population that can reach each location. If you use the same distance, a quick method would consist in using focal statistics. If you want a better model, you need to loop on every city, perform cost distance, reclassify, compute the zonal statistics and sum each part.

  • Thanks for your comments. I will look for extra data to try your solution.
    – Krahsin
    May 8, 2014 at 22:00

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

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