This forum is shutting down! Please post new discussions at community.naturalcapitalproject.org

Recreation Model - individual Predictor

Hi NatCap Team,

I am currently starting to use InVEST for my project and encountered some difficulties I was hoping you guys could maybe help me with :-)

My plan is to measure the recreational value of cork oak forests in three districts in Portugal. I would like to specifically look at cork oaks, so
the normal "forest" predictor might not be accurate enough. The user
guide says that "Users can optionally specify a data folder containing
additional geographic data to use as predictors (for xip values described in
How it works). The data can be in a geographic or projected
coordinate system, but it must be known and specified
in the projection file (.prj). Additionally, the geographic data can be
classified if an optional classification table (.csv) is specified (see
Categorization Tables for more
information).

I found distribution data for the cork oaks online (http://www.habeas-med.org/webgis/pt_en/) and got this map from the homepage (see attached).

So,
my question is: Is there a way I can convert this information (the
online map / the picture file) into a format which I can feed into
InVEST? From the user guide I got: "The tool
also allows users to upload their own spatial data (in any vector shapefile
format), if they have information on additional or alternative
attributes that might be correlated to people’s
decisions about where to recreate."
I tried to georeference
the "cork-oak picture"-file with a map in ArcGIS, but would that be
enough to provide the relevant information to InVEST? I feel like there
needs to be more information for the tool to process (such as an
attribute table or something). I found out that one can manually add polygons to a shapefile but I don't know if that would meet the InVEST - requirements for this model. Also, I don't know how to specify such data in a .prj-file.

Sorry for the long post and the beginner's-questions, but I'm a bit lost at the moment. If you need any more detail, just let me know.

Thank you already!

Cheers,
Marius


2324 x 1218 - 651K

Comments

  • DaveDave Member, Administrator, NatCap Staff
    edited January 2015
    HI Marius, thanks for posting here.

    You are able use your own custom predictor layers in the Recreation model. What you need is the underlying GIS data that is being displayed in the webmap at your link. That webmap is displaying either a shapefile or raster dataset of cork oak distribution, and if you can get your hands on that shapefile or raster layer, then you can feed that into InVEST.

    To run InVEST with this predictor layer, it needs to be a shapefile format with a known coordinate system. Nothing special needs to be in the attribute table.

    I would start by searching around on that webmap or related website for an option to download the GIS data appearing on the map, or at least try to find a person to email to request the GIS data for the cork oak. If the data is in a raster format, you can convert it to a vector with ArcGIS or QGIS. If you are able to find the GIS data, it will ideally come with the projection information (a shapefile is a set of files that includes a .prj file) and you won't have to add it manually. 

    Please follow up with more questions if you have any. And let us know how your project comes along, we are always eager to hear about cool research projects using our Rec model!

    cheers,
    Dave


    Post edited by Dave on
  • mvemve Member
    Hi Dave,

    Thanks for the quick reply and the explanations - already helped me a lot!

    I already contacted the people from HaBEAS project about the underlying data. Unfortunately, so far there is no possibility to extract them. But I will continue to ask people who have worked with similar topics and keep you updated about the progress!

    Cheers,
    Marius
  • DaveDave Member, Administrator, NatCap Staff
    Great, good luck! The original GIS data must exist somewhere.
  • Sooo, I contacted quite a few people who have published about the Montados (cork oak savannahs) and used GIS. i got some positive feedback regarding data and am confident that I will receive the files within the next days.

    Now I started thinking about creating the scenarios (and the input required) for the next run of model. I will be in touch with some Montado-experts and will ask them about their opinion on possible development scenarios. I might also get hold of historical data, so I could extrapolate from that.

    My current concerns are more about the actual doing part. I did the online course and read the user guide, but still I don't really know how to do it. My question is, (1) do I take my 'external' predictor (cork oak distribution in my case) and modify the numbers in the underlying data? Like in a "conservation" scenario I increase the numbers in the attribute table (?) and thereby simulate a wider spread of cork oaks? (And vice versa for a "human / agricultural" development scenario.)

    I was also thinking of using some of the predictors that are available within InVEST (e.g. agriculture, urban). (2) How do I modify these predictors? I only see the 'Data Directory' field which is explained as "Uploaded shapefiles must have identical names as those uploaded for the first run using the Initial Tool." Are there underlying shapefiles for the internal predictors as well that I need to modify accordingly?

    And, out of curiosity (I know that every case is highly specific and comparisons are difficult) (3) what have you experienced to be a reasonable and feasible number of predictors for this model?

    Thanks again and best regards from Lisbon,

    Marius
  • DaveDave Member, Administrator, NatCap Staff
    Hi Marius,

    The Rec model uses the predictor layers by counting up the number of occurrences (or the area) of a predictor in each grid cell across your area of interest. The model builds the grid itself, you can specify the size of the cells. If your predictor layers are points (e.g. locations of villages) then the model will count up how many villages are in each grid cell. If your predictor is a polygon (e.g. the distribution of oak), then the model will calculate the area (e.g. square km) of oak within each grid cell. 

    So in order to create different scenarios, you do not need to edit the data in the attribute table. Instead you need to edit the actual geometry of the shapefiles. For points, this could mean adding new points or deleting old ones. For polygons, it could mean deleting some features, adding new ones, or editing the shapes. You will probably do this step with GIS. 

    Predictors like agriculture and urban that come with InVEST, are extracted from a global landcover dataset. When you run the Rec model, check the box to "Download data". When the model completes you will have these predictor layers and you can edit them to use later in a scenario. 

    Regarding the total number of predictors, I wouldn't hesitate to include everything you think might be a useful predictor. The Rec model will run a simple linear regression with all the predictors, and all that data that goes into the regression ends up in the output grid.shp attribute table. So you can always test different regression models afterwards, outside of InVEST, with some or all of the predictors.

    cheers,
    Dave
  • Hi Dave,

    again: Thank you so much, really helpful.
    I am collaborating with my Portuguese colleagues at the moment and we re-defined the AOI. So now we have to see if it makes sense to use the recreational model on this smaller AOI - there are only very few flickr pictures for the area, so results are not significant, if there are any...
    But, people here have also started working / thinking about using InVEST, so there is quite some expertise and data available and I will keep on looking into InVEST.

    I will keep you updated and probably come up with some more questions over the next few weeks :)

    Thanks again!

    Cheers,
    Marius

    I w

  • Dear Dave, dear community of users of the recreation model,


    we are a research group on landscape management at Univ. of
    Potsdam, Germany, and work on social valuation
    of ecosystem services.



    In the course of our work in the Pentland Land Hills Regional Park
    in Scotland, we applied the scenic quality module of InVEST and
    found great coherence with the results we got from >450
    face-to-face interviews. This is remarquable, because not many photographs had been available. But topography and the network of walking trails supported the results. 

    We are very much interested in other validation efforts of the recreation model, and are open to
    share our data to pool them. In total
    they would cover a range of settings and could allow us to explain strength and weaknesses of the operational use of the model.

    Best, Ariane

     
  • ArianeWArianeW Member
    edited February 2015
    Dear Dave, dear community of users of the recreation model,

    we are a research group on landscape management at Univ. of Potsdam, Germany, and work on social valuation of ecosystem services.

    In the course of our work in the Pentland Land Hills Regional Park in Scotland, we applied the scenic quality module of InVEST and found great coherence with the results we got from >450 face-to-face interviews. This is remarquable, because not many photographs had been available. But topography and the network of walking trails supported the results. 

    We are very much interested in other validation efforts of the recreation model, and are open to share our data to pool them. In total they would cover a range of settings and could allow us to explain strength and weaknesses of the operational use of the model.

    Best, Ariane

     
    Post edited by ArianeW on
  • Hi Ariane,

    It's great to hear that you're having success applying the method using photographs to estimate visitation.  I know of a few additional studies underway or in draft form (that I can share if you'd like to contact me directly).  You might like to see the new paper by Bonnie Keeler that makes comparisons of visitation at lakes in the US with photo-user-days.  And perhaps other members of this forum can contribute other examples.

    We have a pretty significant collection of on-the-ground survey data (even beyond what is in the Wood et al paper in Scientific Reports) that we'd be happy to share if someone is interested in doing further comparisons of the method.

    Certainly let us know how your project unfolds.  We're always interested to see what people are learning and how they're applying the method.

    All the best,

    Spencer

  • Hi,
    I am incorporating the recreation model into a project where I am using different land cover scenarios. I would like to input my own land cover maps into this model, rather than downloading the LULC data that the model includes as an option.

    After reading this thread, it seems like I should convert my land cover rasters into shapefiles and input them as predictors. Is this correct? Is it ok to have all my land cover classes in the same dataset, as opposed to having a separate shapefile for urban, another for wetlands, another for agriculture, etc?

    Thank you for your help,
    Laura
  • DaveDave Member, Administrator, NatCap Staff
    HI Laura,

    You are correct, the LULC data needs to be in shapefile format. And you will need each category in its own separate shapefile.

    Best,
    Dave
  • Thank you for your response, Dave. I converted my land cover classes into separate shapefiles and used them as predictors, and got the "ValueError encountered: The upload size exceeds the maximum." I then tried dissolving the features in each of my shapefiles hoping it would make them smaller/easier to handle, but I got the same error message.

    How should I fix this?

    Thanks,
    Laura
  • DaveDave Member, Administrator, NatCap Staff
    Hi Laura,

    I wonder how far over the limit you are? The model zips up the data directory you specify and that zip file has a max size of 20MB I believe. You could zip it manually just to find the size of the zip file.

    Dissolving was a good idea. There are also "simplify geometry" tools in GIS that should reduce file sizes, at the cost of reducing precision in your data. Deleting columns in the attribute table will save some space, the Rec model ignores the attribute table anyway. Clipping all the layers to your AOI ahead of time could save space, if they cover a much larger area.

    First I would check how close to the 20MB limit you are, then decide on next steps.

    Thanks,
    Dave
Sign In or Register to comment.