Dear Ivan,

the easiest way to get a precipitation field from station data is to use e.g. an inverse distance weighted interpolation. Therefore, you define a search radius around each pixel and compute a weighted average where the weights are based on the distance between the respective pixel center and the gauges within the search radius. If you say that topography does not play a big role, this might already give you a quite good first guess.

However, depending on where you are, you might have to consider topographic effects as well. Then, you can use methods like external drift kriging. While these methods are still quite straightforward and there are tools in most languages available, they require a lot more understanding and knowledge about the data and the processes that need to be taken into account.

Depending on your study area and the spatial and temporal resolution that you need, it might be also worth to have a look at publicly available gridded datasets like GPCC or EOBS.

Best regards,

Christof

Dear Christof, thanks for the answer, however I'm interested in discharge and I've heard there could be specific methods for it besides inverse distance weighting and kriging ...

I need spatial resolution between 1 and 4 km. Study area is Saxony.

Following.

Dear Ivan,

sorry for the misunderstanding (I've somehow missed the word "discharge" in your title :) ). That makes the whole thing a lot more complex as you have to include further parameters and variables in your interpolation. In fact, the topography, soil type, etc. should to be taken into account as these parameters determine infiltration, flow direction and speed of discharge, etc. And this does not even include any kind of "human"-made impacts on your discharge (e.g. irrigation, reservoirs, etc.). I guess that the most straightforward approach would be to use a hydrological or land surface model (e.g. mHM).

Depending on the desired spatial and temporal resolution of your runoff field, you can probably think of other, more data-driven methods like e.g. a Kalman Filter. Such approaches would also allow for including e.g. water balance constraints in the estimation. However, as you need a quite high spatial resolution of up to 1 km and you "only" have 100 gauges, you have to deal with a highly underdetermined system (i.e. more unknowns than observations). There are of course several "workarounds" for these issues (e.g. localization). But this again requires a lot of knowledge about your data, the study domain and, of course, the technique itself.

You can have a look at a publication from us, where we've used a purely data-driven Ensemble Kalman Filter for predicting monthly basin-scale runoffArticle Basin-scale runoff prediction: An Ensemble Kalman Filter fra...

I hope that I could help you and give you some ideas with which you can proceed.

Cheers!