The Migratory Modelling Experiment (MM) lets you investigate how the potential distribution of your species might change throughout the year. The implementation of this in the BCCVL modelling wizard on EcoCommons generates a separate model for each month or season. This allows the user to select different variables for each month or season and results in different model summaries and evaluation statistics.
This kind of approach is most useful if the question being investigated involves exploration of species distribution that are thought to relate to very different variables across the year. When running an experiment each month or time period is modelled with the same algorithm. Users can start another experiment to run a migratory model using an additional algorithm, and information on SDMs can be found here with additional educational material here.
An alternative approach is to build one model for all months or time periods and predict results from that single model for each month or time period. An example of this single model migratory approach can be found in our migratory marine use case in R available here. This approach is useful when available occurrence data is not evenly distributed within each month.
Run an MM on EcoCommons in the BCCVL modelling wizard
On the top of the page click on “Analysis Hub” and then on “Modelling Wizards”. Under “Primary experiments” choose “Migratory Modelling Experiment”.
Step 1: Description tab
Enter the title of your experiment in the first box (e.g. January distribution of Greater Scooty Owl).
- (optional) You can also add a description of your experiment in the box below if you want to convey more information. Some researchers use this box to record their research questions or hypotheses for later referral.
Click “Next” at the bottom of the page.
Step 2: Occurrences tab
- Select your pre-loaded migratory occurrence dataset formatted like this, which should include the columns: lat, lon & month by clicking on "+Select an occurrence dataset".
Note: If you click this and you have no loaded species occurrence datasets you will need to visit the dataset page and upload the required data.
- In the pop-up box select the dataset you wish to use in your migratory model.
- (optional) You can visualise your occurrence data by clicking the green eye icon next to your loaded data.
- Click "Next" at the bottom of the page.
Step 3: Absences tab
You can either upload absence data or pseudo-absence data, or select pseudo-absence data below. See SDMs in R module (Step-1)for alternative ways to generate pseudo-absence data.
- Uploaded true absence or pseudo-absence data labelled ‘lat’, ‘lon’, & 'month' with the same input values as the 'month' column in your occurrence data. Complete these steps:
- Select “Yes” under the question of whether you have absence data.
- Click on “+ Select an absence dataset”.
- In the pop-up box select the pre-loaded absence dataset you wish to use.
- (optional) You can visualise your absence data by clicking the green eye icon.
- Click “Next” at the bottom of the page.
2. Generate Pseudo-absence data automatically; see here.
- Select “No” under the question of whether you have true absence data.
- You can change the pseudo-absence generation settings which will be used within each month or time period.
- Ratio absence to presence points e.g. 1 = 1:1 ratio, 2 = 2:1 ratio.
- Strategy to generate pseudo absence points; disk (with minimum and maximum distance to presence points), random within your extent, SRE (Surface Range Envelope – see here),
- Quantile is used for the ‘SRE” pseudo-absence generation strategy, which removes the most extreme values of the environmental space of your occurrence data. The default is 0.025, which captures 95% of the values in the environmental space.
- Minimum or maximum distance (m) to presence points for ‘disk’ pseudo-absence generation strategy.
Click “Next” at the bottom of the page.
Step 4: Subsets & Environmental Data tab
Click on “+ Add subset”.
Note: If you choose data layers that do not have the same resolution you can choose whether they should be scaled to the finest or coarsest resolution.
- Click on " Select datasets" on the left side of the white box.
- In the pop-up box, you can enter search terms to filter for required datasets or filter by collection, resolution and/or domain i.e, to search in the top box 'January' would return all the available data.
- Once you have found the dataset/s you are looking for select them and click “Close”.
- When back on the Climate & Environmental Data tab you can select/deselect data layers on the left side of the white box (tick each box to specify the layers you want to use).
- On the right sight, add a Title for the subset (the month or time period), and indicate which values in the 'month' column from your data should be used for this subset.
Note: The values in the 'month' column need to be numbers between 1 & 12 but can represent other time periods than months (e.g. 1 = winter, 2 = summer, or 1 = breeding, 2 = non-breeding), just as long as they match the values in your data.
- You can add more subsets (months or time periods) by clicking the Add Subset button.
- For each subset (time period) you can choose a different set of climate/environmental variables, doing this is useful to compare how different variables are related to distributions at different time periods. Choose the same kind of data for each month if you are more interested in predicting the changes in distribution over time, especially if your sampling between each time period varies. By using the same variables in each month, i.e. January 2018 NDVI, February 2018 NDVI..), it is easier to argue that changes in predicted distribution are not simply related to the different kinds of variables used in each time period.
- Once you have selected all your environmental and climate layers for each subset (month or time period) click “Next” at the bottom of the page.
Step 5: Constraints tab
In this section, you can select the extent i.e. boundary of your study area. This means that only the occurrence records from the constrained area are used, and pseudo-absence or background points are only generated in this area. It is good practice to remove parts of the geographic or environmental space where you are certain your species will not be found.
The default constraint is the convex hull (= minimum polygon) around all occurrence records indicated by a blue outline on the map. The different constraint options are:
- Use Convex Hull for each individual species
- You can add a buffer around the convex hull by nominating a distance in km. The buffer will be added to the map once you click outside the white box.
- Select constraints by pre-defined region
- Select one of the region types that are currently available in the BCCVL: i.e. Local Government Areas, National Resource Management Regions, Australian States and Territories, etc.
- Find the region of your interest in the drop-down menu. You can select multiple regions.
- You can also add a buffer around the pre-defined region constraints.
- Use Environmental Envelope
- This is the geographic extent to which all selected climate/environmental datasets overlap this environmental space is defined using SRE.
- Draw constraints on Map
- Click on “Start drawing”, then click on the map to draw a shape on the map to define the study area.
- Upload Shapefile
- Select a shapefile from your computer to use as the constraint.
Note: The model will be trained on occurrence points that fall within the selected area, and the predictions of current suitability and future suitability will be made within the area you define here.
- Once you selected the constrained area click “Next” at the bottom of the page.
Step 6: Algorithms tab
- Select the algorithm you would like to use to calibrate your model. For the MM you can choose only one algorithm to run your experiment. Don't know which one to select? You can read about each algorithm here.
Note: Each subset will be modelled with the same algorithm configuration settings
- (optional) Configuration:
- If you want to change the pseudo-absence/background selection settings for a particular algorithm, you can do that here. If you change nothing, the settings from the Absence tab will be used for all selected algorithms.
- Other configuration options are available for most algorithms. These options can be changed by changing the value or making a different selection from the drop-down menu. The configuration options are currently set to the standard default values of the R packages. More information about each configuration option can be found on the support page for that particular algorithm.
- Click “Next” at the bottom of the page.
Step 7: Run tab
- Ensure you are happy with your experiment design.
- If none of the tabs has a triangle with an exclamation mark, your experiment is ready to go.
- Click “Start Experiment”.
- If any of your tabs have a triangle with an exclamation mark, revisit them and ensure you have filled in each component correctly.
A log file will now be sent to our virtual machines where your experiment will be run.
You can view the progress of your job under “My job”. Once your job is finished you can view the results by either clicking “View all results” inside your job or clicking on the “My Results” tab under Workspace.
For now, sit back and relax, grab a coffee, or do some other work without being hampered by a slower computer that is running heavy models in the background.