Questions about precipitation data used in the models

Hi! I am Sharon.
I have been calculating ecosystem service for year 1990, 2000 and 2010 based on SDR, NDR and Water Yield models lately and precipitation data was involved in these models. Here I have some questions about the precipitation data.
The InVEST documentation mentioned about “Average annual precipitation” and “Ideally, the gauges will have at least 10 years of continuous data…”(in Water yield model). So does the precipitation data used in my calculation should be average annual precipitation from 1990 to 2010 (average annual precipitation of 20 years)? Or should I use the total annual precipitation of 1990, 2000 and 2010 separately? For the total annual precipitation I mean the sum of monthly precipitation in 1990, 2000 or 2010.
Looking forward to your reply! Thank you!

Hi @Sharon927,

Thanks for posting to the forums and using InVEST!

I think it depends a bit on the research question you are trying to ask. If you are looking at how Precipitation has effected ecosystem services between 1990, 2000, and 2010 then you might want to use the average Precipitation of 1985-1994, 1995-2004, and 2005-2014 for your different runs. If you are mostly interested in looking at how Land Cover / Land Use changes ecosystem services between the three years then using one large average over all the years makes sense to me.

I’m certain that someone with more research background can provide chime in here. Full disclosure, I’m a Software Engineer on the NatCap team, not a research scientist!



Hi @Sharon927 -

I agree with @dcdenu4 that if you have access to a long enough series of precipitation data, it would be great to have it coincide with the different years of your land cover map. As Doug noted, that would involve taking a 10+ year annual average over the precipitation data around your land cover dates. Often, we don’t have this kind of data though, so you can use a single, long-term precipitation average raster that covers all of your land cover dates instead.

However, we do not generally recommend using a single year of precipitation data in the models (for example, using annual precip from only the year 1990 for the 1990 land cover run) because these models represent more general averages, and it’s possible that 1990 may have been abnormally wet or dry, which would skew the results if you’re interested in longer-term changes between these decades.

~ Stacie


Hi! Doug! Thank you so much for your reply! It’s important information for me :grinning:

Thank you Stacie! I think in my case I focus more on the influence of LUCC has on ecosystem service so I will use an annual average precipitation covers my land cover dates. Thanks again :grinning: