Skip to content

Quick start

This walkthrough takes you from a freshly installed plugin to a finished habitat-suitability map and a list of field survey sites for Bradypus variegatus, the brown-throated three-toed sloth. It is the same species the original Maxent paper used (Phillips, Anderson & Schapire 2006).

Every step shows the real QGIS screen. The red boxes and numbered circles on each picture match the numbered instructions below it, so you can follow along click by click.

Before you start

Complete Installation and Dependencies first. The QMaxent environment ready status must be shown in green before you continue.

How long does it take?

About 10 minutes in total. Most of that is the model fit in Step 9, which runs in the background while QGIS stays usable.

Part 1. Get the example data

Step 1. Open the example dataset downloader

Plugins menu with QMaxent and Download Example Dataset highlighted

  1. Open the Plugins menu in the QGIS menu bar.
  2. Point to QMaxent.
  3. Click Download Example Dataset...

Step 2. Choose Bradypus and download

Download Example Dataset dialog with Bradypus selected

  1. Select Bradypus variegatus.
  2. Check the folder in Save to. The default is a qmaxent_examples folder in your home folder. Click ... to choose another place.
  3. Click Download.

The download takes a few seconds. When the message below appears, click OK.

Message saying the example dataset is ready

Step 3. Check the layers on the map

QGIS main window with the Bradypus layers loaded

  1. The Layers panel now lists the bradypus presence layer and the nine environmental rasters (bio1 to bio17 and biome). The blue points on the map are the 116 places where the sloth was recorded.

Part 2. Prepare the data (① Data tab)

Step 4. Open the Analysis dock

Choose Plugins → QMaxent → QMaxent Analysis. The Analysis dock opens on the right side of QGIS.

Analysis dock with the five tabs and the Run Maxent button highlighted

  1. ① Data: choose the presence points and the environmental rasters.
  2. ② Parameters: set how the model is fitted and evaluated.
  3. ③ Training: follow the model fit and read the log.
  4. ④ Results: look at response curves and variable importance, and make the suitability map.
  5. ⑤ Survey Site Selection: pick field survey sites from the map.
  6. ▶ Run Maxent starts the model fit. You use it once the first two tabs are ready.

Work through the tabs from left to right.

Step 5. Choose the presence points

Data tab with the presence layer selector highlighted

  1. In Presence Points Layer, choose bradypus. The line below it reads 116 presence points loaded.
  2. Open point file... loads a point file (GeoPackage, shapefile and others) that is not yet in your project.
  3. Load CSV (X/Y columns)... loads a table with coordinate columns. You do not need either button for this example.

Step 6. Add the environmental rasters

Data tab after Add from project, with the raster list filled

  1. Click Add from project.
  2. Every raster in the project appears in the list. The button on the right of each row shows whether QMaxent treats the raster as [continuous] or [categorical]. Click it to change the type. Scroll down the list and check that biome is marked [categorical], because its values are class codes, not amounts.

Step 7. Check that the rasters line up

Check Raster Consistency button and the green status line

  1. Click Check Raster Consistency.
  2. The status line turns green: ✓ All 9 rasters share grid (CRS: EPSG:4326, resolution: 0.5 × 0.5). If it turns red instead, see Fixing rasters that do not line up.

Part 3. Fit the model (② Parameters and ③ Training)

Step 8. Review the parameters

Click the ② Parameters tab. For this first run, keep the defaults.

Parameters tab with the Feature Types, Regularization and Spatial evaluation groups highlighted

  1. Feature Types: Auto (maxnet default rule by sample size) chooses the feature classes from the number of presences. With 116 records it uses linear, quadratic, product and hinge (LQPH).
  2. Regularization: a multiplier of 1.00 is the usual starting value.
  3. Spatial evaluation: Geographic K-Fold with 5 folds checks the model on areas it did not see during fitting. Fix random seed: 42 makes the run repeatable.

These defaults are a sensible starting point (Phillips & Dudík 2008; Phillips et al. 2017), not a guarantee of the best model for your own data.

Step 9. Run Maxent

Click ▶ Run Maxent at the bottom of the dock. QMaxent switches to the ③ Training tab.

Training tab while the model is being fitted

  1. The progress bar shows how far the run is.
  2. The line under it names the current step.
  3. The log records every step as it happens.

When the run ends, the tab looks like this:

Training tab after the run, with the final summary

  1. The status line reports the result.
  2. The log ends with a summary of the model and the evaluation.
  3. The bar at the bottom of the dock keeps the key numbers in view, for example presence=116 background=10,108 | train AUC=0.9135 | CV AUC=0.7524.

The test AUC from spatial cross-validation is lower than the training AUC. That gap is normal and honest: it shows how well the model does in places it was not fitted to (Roberts et al. 2017).

Part 4. Read the results and make the map (④ Results)

Step 10. Look at the response curves

Click the ④ Results tab.

Response Curves sub-tab with the variable selector and the curve highlighted

  1. The Response Curves sub-tab is open.
  2. Choose a variable in Variable.
  3. The curve shows how predicted suitability changes along that variable. The shaded band is the range of values at the training points.

Step 11. See which variables matter

Jackknife Importance sub-tab with ROC curves and jackknife bars

  1. Click Jackknife Importance.
  2. The left panel shows the ROC curves. The right panel compares the model fitted with only one variable and without that variable.

Permutation Importance sub-tab with the bar chart and the save buttons

  1. Permutation Importance ranks the variables by how much the model gets worse when each one is shuffled.
  2. Save PNG... saves the chart.
  3. Save CSV... saves the numbers.

Step 12. Make the suitability map

Spatial Projection sub-tab with the output settings and the run button

  1. Click Spatial Projection.
  2. Keep cloglog in Output transform. It gives values from 0 to 1 that read as relative suitability (Phillips et al. 2017).
  3. Output raster is where the map is saved. Click ... to change it.
  4. Click ▶ Run Spatial Projection.

Status line reporting the saved prediction raster

  1. When the status line reads Done: followed by the file path, the map is saved and added to the project.

QGIS main window with the prediction layer on the map

  1. The prediction layer appears in the Layers panel. Darker colours mean higher predicted suitability.

Part 5. Pick survey sites (⑤ Survey Site Selection)

Step 13. Set up the survey

Click the ⑤ Survey Site Selection tab.

Survey Site Selection tab with the purpose, search settings, sampling strategy and run button highlighted

  1. Survey purpose: choose High-suitability search to look for new populations in places that are suitable but not yet surveyed.
  2. High-suitability search settings: Minimum suitability is the lowest map value a site may have.
  3. Sampling strategy: set how many sites you want and how far apart they must be from known records and from each other (in metres).
  4. Click ▶ Select Survey Sites.

Step 14. Check the survey sites

Status line reporting how many survey sites were selected

  1. The status line reports the result, for example ✓ 20 survey sites selected (MinSuit = 0.9000), and where the GeoPackage was saved.

QGIS main window with the survey sites on the suitability map

  1. The Survey sites layer is added on top of the map. Open its attribute table to see the coordinates, the suitability value and, if reverse geocoding was on, the address of each site.

What you have done

In these steps you ran a complete species distribution modelling workflow inside QGIS. You loaded presence points and environmental rasters, checked that the rasters line up, fitted a regularized Maxent model, evaluated it with spatial cross-validation, read the response curves and variable importance, mapped habitat suitability and turned the map into a list of field survey sites.

Next steps

  • Read the User guide chapters to learn what each control does and when to change its default.
  • Work through the Bradypus worked example for a closer reading of every plot.
  • Practise fixing rasters that do not line up with the Ariolimax example.
  • Reproduce a published study with the Pitta nympha example.