Dear Mohammadtaghi Avand,

I think, rather down-scaling the large scale model data (which can introduce error), it is better to up-scale the observational data to the model grid (e.g. by averaging the each observation point available within a model grid). Then you can try for correlation, which could possibly give your better results..

Best Regards

thanks for your response.

But I didn't understand what you meant.

Suppose, you have global data of 2 degree resolution, and observations with 0.5 deg resolution. Then, spatial average all the observation data point within the 2 degree grid. Then try to correlate this low resolution observational data (which now also have 2 degree resolution) with the global data.

Best

Just yo give some clarityon the example from Bijay. Saying observations with 0.5 degree resolution is a bit miss leading, since there is most likely only gauge stations at irregular distances. So, what you need to do is some kind of spatial interpolation at 2 degrees resolution. You havet o be aware though, that by doimg so, you are assuming ipmlicitely that the gauge data is in fact representative of the spatial average.

The resolution of my large scale data is 2 degrees. In what method and software can I increase the spatial resolution of observational data?

Estrictly speaking you cannot imcrease the resolution of gauged data,.never. Either it was measured or not. This is the only "true value" that you can have.

What you could maybe do however, is to make an estimate at higher resolutions...if and only if there are other information sources that allow you to do this. Either ground Radar or Satellite are the usual and in many cases the only alternatives. Another possibility would be so called reanalysis data, but this is again not a true value.

What I am suggesting is, just to do spatial averaging (simply using MATLAB for example). Provided, you have multiple observational data available within a 2x2 deg grid.