Predicting Pollinator Range Shifts With MaxEnt for Climate

ISEF Category: Animal Sciences

Subcategory: Ecology and Agriculture  ·  Difficulty: Advanced  ·  Setup: Home Setup  ·  Time: 1 to 2 Months

The Hook

A pollinator can vanish from one region even while it still thrives somewhere nearby. That shift can change crop yields, wild plant reproduction, and local ecosystems. With public data and MaxEnt, you can test where the species might live in 2050. Your project can turn scattered sightings into a real climate forecast.

What Is It?

A species-distribution model asks a simple question: based on where a species has been found, what places should also be suitable? MaxEnt is one common tool for that job. It looks for patterns in occurrence records and environmental variables, then estimates where the species is most likely to occur.

Think of it like matching a puzzle piece to a map. If a pollinator shows up in cool, moist, flower-rich areas today, the model checks where those same conditions exist now and in future climate scenarios. Your job is to feed it clean data, pick good variables, and test whether the predicted range makes sense.

The climate part matters because habitats do not stay fixed. Warming can push a species uphill, northward, or into smaller pockets of suitable habitat. That makes this topic a strong fit for ecology, conservation, and agriculture.

Why This Is a Good Topic

This is a strong science fair topic because you can test a real ecological question with public data and clear outputs. You are not guessing where the pollinator lives, you are building and checking a model against known records. The project connects to food security, habitat planning, and climate change, and it teaches you data cleaning, mapping, and model evaluation.

Research Questions

• How does using iNaturalist records versus GBIF records change the predicted current range of the pollinator?
• What is the effect of adding climate variables on model accuracy compared with location data alone?
• Does spatially thinning occurrence points change the final MaxEnt map?
• To what extent do different 2050 climate scenarios shift the predicted habitat centroid?
• Which climate variable contributes most to the model's predicted range for the pollinator?
• How does changing the background region affect the size of suitable habitat predicted by MaxEnt?

Basic Materials

• Laptop with internet access.
• iNaturalist account for downloading observation records.
• GBIF account or portal access for occurrence downloads.
• QGIS installed for map viewing and raster checks.
• MaxEnt software or a compatible species-distribution modeling package.
• Spreadsheet software for cleaning CSV files.
• Regional climate raster files for current and 2050 scenarios.
• Local species checklist or field guide for confirming the focal pollinator.

Advanced Materials

• University computer with enough memory to handle raster layers.
• R with species-distribution packages such as dismo or biomod2.
• Verified occurrence dataset with expert-reviewed records.
• High-resolution climate and land-cover rasters.
• GPS-backed field survey records for independent validation.
• External storage for large GIS files.
• Access to a university map printer or plotter for final figures.

Software & Tools

• QGIS: Maps occurrence points, clips raster layers, and compares predicted habitat zones.
• R: Cleans data, runs repeat model tests, and calculates evaluation metrics.
• MaxEnt: Fits the species-distribution model from occurrence and climate layers.
• OpenRefine: Cleans messy records, standardizes fields, and removes duplicate entries.
• Google Sheets: Gives you a quick way to sort records and flag obvious data problems.

Experiment Steps

1. Define one focal pollinator, one region, and one clear climate question.
2. Choose your occurrence sources and set rules for filtering bad or duplicate records.
3. Select the climate variables that match the biology of your species and drop redundant ones.
4. Plan how you will split data for training and testing so your results are not inflated.
5. Decide which model settings, thresholds, and comparison metrics you will use before you run the map.
6. Set up one current-climate run and one or more 2050 scenario runs so you can compare change.

Common Pitfalls

• Using duplicated iNaturalist and GBIF records, which overweights common sightings and creates fake hotspots.
• Leaving in records with weak location precision, which can put the pollinator in the wrong habitat.
• Mixing climate layers with different resolutions, which makes the MaxEnt output hard to trust.
• Training and testing on the same geographic cluster, which makes model accuracy look better than it really is.
• Ignoring sampling bias near roads and cities, which can tilt the predicted range toward well-sampled places.

What Makes This Competitive

A stronger project does more than make a pretty map. It compares multiple data sources, tests different background regions, and shows how much the answer changes when you tweak key modeling choices. You can raise the level again by reporting uncertainty, not just one final prediction. Good spatial validation and a clear argument for your variable choice will matter a lot.

Project Variations

• Compare a bee, butterfly, and hoverfly to see whether climate limits them in the same way.
• Add land cover or flower abundance data to test whether habitat matters more than climate alone.
• Rebuild the model with museum specimen records and compare the result with citizen science records.

Learn More

• GBIF: Search the portal for occurrence records, filters, and download tools for species data.
• iNaturalist Help: Read the site guides on research-grade observations and exporting sightings.
• MaxEnt software page: Find the model download, user guide, and example runs from the official source.
• QGIS Training Manual: Learn the free GIS workflow for clipping rasters, mapping points, and checking layers.
• NOAA Climate Data Online: Look up climate datasets and background material for current and future conditions.
• Search PubMed: Find review articles on species distribution models, range shifts, and climate change.