Sensitivity analysis for InVEST models

Are there best practices for implementing sensitivity analyses for InVEST models, and are they dependent on the specific module? I am implementing the Pollination, Carbon, GLOBIO, and SDR modules for reference. In some cases I have information about parameter distributions (e.g. 25th-mean-75th percentile) but other parameters have unclear constraints based on the literature.

Best,
Katherine

Hi @kjones6, and welcome to the forum!

It’s great that you have some parameter distribution information to base a sensitivity analysis off of. If others have unclear constraints, then we often try adjusting them by +/- 50% or +/-100%, in whatever increments seem appropriate. Depending on which input you’re considering, you can also try a different dataset. For example, the erosivity input to SDR is usually based on annual rainfall, so you could create an alternate erosivity layer using a different source for rainfall and see how it changes the result.

I asked around NatCap, and several of our science team highly recommended a paper by Hamel and Bryant called " Uncertainty assessment in ecosystem services analyses: Seven challenges and practical responses" - https://doi.org/10.1016/j.ecoser.2016.12.008. Perhaps there are additional ideas in there that can help.

If anyone else has experience to share, please do!

~ Stacie

2 Likes