Calculate the annual water yield by separate different landuse types

Software Support
1

When running the annual water yield,can I separate different landuse types and calculate their water yields respectively?Is this approach reasonable?

For instance,I want to conduct a pre-experiment,but the data volume is too large.Therefore,I choose to first perform cluster analysis,calculate the proportion of a certain land type in each grid point,and then create a new grid by comparision, obtaining a grid with only the distribution of that land type and the distribution of non-this land type.Then,I use this as the new land type grid to input into the model.Finally,I multiply the output result by the proportion to obtain the annual water yield of a certain land use type.

I want to know if this is reasonable and whether it conforms to the model principle.

2

Hello @y2603186176 and welcome to the forums!

To confirm, are you finding that the data needed for the InVEST Annual Water Yield model are too large to run the model?

InVEST has for almost 10 years now, been very computationally efficient and memory efficient, so even if your model inputs are very, very large, the model should compute efficiently, even on a typical workstation, as long as you have enough disk space available to store the model outputs. Are you having issues running the model?

Thanks,

James

3

When I select a very large research area and the landuse raster has aresolution of 30m,the data will be large.In this case,the model can run,but it will take more than 12 hours for each run.Since I need to run data for over 30 years,this will consume a considerable amount of time.Therefore,to simplify this process,I want to use cluster analysis and separate differeent landuse types,and reduce the resolution to increase the running speed,so as to determine through this pre-experiment whether my research approach is correct.

However,I am not sure if this is reasonable and whether it conforms to the model principle.And whether it will cause excessive deviation in the final results.

4

Thanks for the additional information about your study area, @y2603186176 ! That does sound like a lot of computational time involved, so your study area must be pretty large to occupy a full 12 hours of compute time.

To directly answer your question, yes, you could approach the analysis differently and do the computation following the InVEST methodology on each combination of input parameters and reassemble the outputs and you should be able to get the same outputs as what the InVEST model produces. Each pixel stack is independent, except for the final aggregation at the watershed/subwatershed scale. Please keep in mind that per-pixel maps are not to be interpreted for understanding of of hydrological processes or to inform decision-making of any kind (per the docs).

If you don’t want to put in this kind of work, you may also want to try setting the n_workers parameter in the InVEST workbench settings in order to use more CPUs, which is likely to speed up your model runs somewhat if tasks can execute in parallel. Of course, this assumes that you have a low-latency, high-speed hard disk drive on your workstation.

And, if your institution has a computational cluster, you may also benefit from being able to run InVEST on the cluster.

Please let us know if you have any further questions!

James

5

Thanks for your reply,I will consider taking your suggestions!