Hello, everyone!
I am running the Seasonal Water Yield model for a very large area in Brazil that contains several basins.
I reprojected the 1000x1000-meter evapotranspiration and precipitation rasters from Wordclim to 30x30 meters in order to run SWY, and I received some criticism for this because I am “losing” microclimatic information. Even though my intention is to conduct a large-scale analysis, this concerns me.
The reprojection was very simple, using the following code:
raster_evap_reproj ← terra::project(raster_evap_cutted, lulc,
method = “bilinear”)
I would like to know if you know of any way to reduce the scale of WorldClim or interpolate the data so that I can obtain more accurate precipitation and evapotranspiration rasters. I have already researched some articles on this, but I am having difficulties because my study area does not have many weather stations. Thank you for reading this far!
I assume that for your climate data from WorldClim you’re using multiyear means. Over how many years are these rasters averaged?
If your DEM is already 30x30m, I do not recommend reprojecting the climate inputs since the model will do this for you, as noted in the User Guide,
Raster inputs may have different cell sizes, and they will be resampled to match the cell size of the DEM. Therefore, all model results will have the same cell size as the DEM.
Another consideration to be mindful of is that running this model over a very large area with fine resolution pixels will require considerable computing power and/or could take quite a long time to complete. The more pixels, the longer it will take to run. Additionally, outputs will be large in data volume. So if you really want to run it with 30m pixels over a very large area, ensure you have enough data storage available to absorb the results and that your computer can handle the task in an amount of time that’s tolerable for you.
Also, to model a hydrologically complete area, be careful not to cut off the headwaters of your basins. For Brazil, this may require that you extend your area of interest to the crest of the Andes in neighboring countries to the west, such as in Peru. If you have not included the full upslope extents of your basins already, you may have to obtain more data.
Hi @jesseG !
I averaged the rasters between 1970 and 2000. My colleagues and I decided to reproject to standardize the data, but we used that excerpt from the manual you mentioned as justification.
We were able to run the model, fortunately, my computer is good and if necessary, my university has supercomputers. I have this question because one of the reviewers of my paper questioned this downscaling situation, and I intend to produce another paper using SWY and would prefer not to encounter these issues.
Regarding the watersheds, I am using a large number of smaller watersheds ottocoded by the Brazilian National Water Agency. My area of study is located at the headwaters of the eastern mountain basins of Brazil.The area is large, but its rivers are not influenced by the Andes (something yes, due to the flying rivers, but not in relation to the headwaters).
