This forum is shutting down! Please post new discussions at

Plot sampling or run for whole landscape


I am in doubt how to use the InVEST model on a certain landscape. 

I need to compare forest certified situations in the landscape with non certified situations in the landscape (approx. 453,000 ha).

My first thought was: divide the landscape into sub landscapes which are approximately homogenous in climate and topography conditions and then do a random selection of plots (of 5 ha) in each sub landscape to run the InVEST model on each of these plots. 

Is it better to run the model on the whole landscape first and then separate the outputs into the selected plots or vice versa (run the model for each of the plots separately)?

Or is there even a better methodology for this kind of problem?

Many thanks in advance!

I am looking forward to the answer!

Kind regards



  • swolnyswolny Member, NatCap Staff
    Hi Eva -

    It would be very helpful to know which model(s) you would like to use. Each model might be different in how you approach your question. For example, for the freshwater models (water yield, NDR, SDR), you would want to run the model on the whole landscape, so you are capturing the connected hydrology of the area. 

    It would also be useful to understand the question that you are trying to answer, comparing certified vs non-certified forests. Thanks for the additional information.

    ~ Stacie

  • Hello Stacie

    Thank you for the answer.

    Here is some additional information:
    The models I would like to use are:
    - the water yield model
    - SDR and NDR models
    - seasonal water yield model
    - habitat quality and rarity model
    - habitat risk assessment
    - forest carbon edge effect model
    - carbon storage and sequestration model
    - Scenic Quality model
    - Timber production model (but as I saw  it's not available in InVEST 3.5.0??)

    I would like to evaluate the effect of forest certification on the provisioning of some ecosystem services in my study area. 
    I have a map with the spatial distribution and location of forest certified areas in the study area.
    I was thinking about comparing model outputs between forest certified areas and areas without certification, to see the effect of forest certification.
    Since the study area is approx. 453,000 ha, and has different climate zones and variation in topography, I was thinking to divide the study area (one landscape) into 3 more homogenous sub landscapes.
    And within each sub landscape, I would select 10 representative plots (mainly containing forest land use, about 5-10 ha) of forest certified areas and 10 plots of non certified areas. In this way I can compare the results of forest certified areas and non certified areas (between the plots).
    Therefore, I was doubting between two options:
    - running each model on the whole study area (consisting of certified and non certified areas), and clip the output each time to the individual plots, after which I would compare the results.
    - running each model on each of the plots individually after which I compare the results between certified and non certified plots.

    I imagine that in some cases the output is influenced by areas outside the plots.

    Therefore I was wondering which of the two options is the best when running each of the models I will use.
    Or if there is even a better and more efficient option, it will also be helpful.

    I hope you understand my question now I have given some additional information.

    Looking forward to the answer!

    Thank you in advance!

  • swolnyswolny Member, NatCap Staff
    Wow, Eva, you're ambitious with the models. I would advise looking closely at a few things, since you've got some potential duplication. Habitat Quality and Habitat Risk are a bit different, but we usually run one or the other, depending on what kind of data you have and the type of information you need. Check out this section of the HRA User Guide for a comparison.

    Similarly for Carbon Edge Effect and Carbon Storage. Carbon Storage is a very simple model that really just sums up user-supplied carbon pools for each pixel. Edge Effect also provides storage data, but takes into account differences in carbon storage that generally occurs along forest edges. 

    And again for water yield and seasonal water yield. Seasonal water yield will provide quickflow (not baseflow) information monthly, but will provide an index of where baseflow is likely being produced across the watershed. Water yield is only annual, and does not distinguish between baseflow and quickflow. Again, it's a matter of what kind of information you need, and whether you have monthly vs annual data.

    We no longer provide the timber production model, sorry about that.

    For the freshwater models, I would definitely run them on the whole area of interest, so you take into account hydrological connectivity. Water yield is the one exception, since it does not route water around the landscape.

    The habitat models also need information on the threats/stressors to the habitats, which may be located outside of the sub-landscape, so again, covering a larger area is probably important there. Scenic quality should be similar, as people could presumably be looking out over an area larger than the sub-landscape.

    The carbon storage model can be run on a smaller area, since each pixel's value does not depend on anything outside the pixel. 

    Personally, I would keep things simpler by running over the larger area. That way you have only one of each input to deal with, and later on you can extract other information from the results if needed. If you will be treating the different certified/non-certified areas differently (in terms of carbon storage, water use, habitat suitability, etc), you'll just need to create separate land use/land cover classes for each area, so you can differentiate the relevant model inputs.

    Since this is a large area, the one problem could be if you use very high-resolution input data, then it might max out the memory on your machine, or take a long time to run.

    ~ Stacie

  • Hello Stacie

    Thank you for the answer and the useful information and advice!

    If you say "Water yield is the one exception, since it does not route water around the landscape.", are you referring to both water yield models or only the (annual) water yield model?

  • swolnyswolny Member, NatCap Staff
    Only the annual water yield model. Seasonal water yield does route water, so the larger landscape is important.

    ~ Stacie

  • Ok thank you!

    So it is possible to run the Water Yield Model only for one watershed?

Sign In or Register to comment.