Mapping Suitable Habitat for Birds Using Google Earth Engine and Maximum Entropy Algorithm: A Comprehensive Study

GGE Image source Google

*Heri Tarmizi

This study aims to map suitable habitats for birds using GEE and the MaxEnt algorithm, providing a detailed methodology and analysis of results.

Introduction

In recent years, the intersection of remote sensing technologies and machine learning has provided unprecedented capabilities in environmental monitoring and wildlife conservation. Google Earth Engine (GEE) has emerged as a powerful platform for the analysis of geospatial data, enabling large-scale environmental monitoring with cloud computing resources. Among various machine learning algorithms, the Maximum Entropy (MaxEnt) algorithm has gained popularity for species distribution modeling (SDM) due to its robust performance with presence-only data. This study aims to map suitable habitats for birds using GEE and the MaxEnt algorithm, providing a detailed methodology and analysis of results.

Methodology

Data Collection

The first step involves the collection of relevant environmental and species occurrence data. Environmental variables such as temperature, precipitation, land cover, and elevation are essential for modeling bird habitats. These datasets can be accessed via GEE's extensive data catalog. Bird occurrence data can be sourced from global databases such as the Global Biodiversity Information Facility (GBIF).

1. Environmental Variables:

   • Temperature and Precipitation: Monthly climate data from the TerraClimate dataset.
   • Land Cover: Land cover classification from the MODIS Land Cover Type product (MCD12Q1).
   • Elevation: Digital Elevation Model (DEM) data from the Shuttle Radar Topography Mission (SRTM).

2. Bird Occurrence Data:

   • GBIF provides an extensive database of bird sightings, which can be filtered for specific species of interest.

Data Preprocessing

Preprocessing involves cleaning and preparing the data for analysis. Environmental variables need to be resampled to a common spatial resolution and extent. Occurrence data should be filtered to remove duplicates and ensure accurate georeferencing.

python
import ee
import geemap
import pandas as pd

# Initialize the Earth Engine module.
ee.Initialize()

# Load environmental data from GEE
temperature = ee.ImageCollection("IDAHO_EPSCOR/TERRACLIMATE").select('tmean')
precipitation = ee.ImageCollection("IDAHO_EPSCOR/TERRACLIMATE").select('ppt')
land_cover = ee.Image("MODIS/006/MCD12Q1/2019_01_01").select('LC_Type1')
elevation = ee.Image("CGIAR/SRTM90_V4")

# Load and filter bird occurrence data from GBIF
gbif_data = pd.read_csv('path_to_gbif_data.csv')
gbif_data = gbif_data.drop_duplicates(subset=['decimalLongitude', 'decimalLatitude'])

Model Development

The MaxEnt algorithm is implemented using the dismo package in R, which provides tools for species distribution modeling.

R
library(dismo)
library(raster)

# Load environmental variables
env_vars <- stack('temperature.tif', 'precipitation.tif', 'land_cover.tif', 'elevation.tif')

# Load occurrence data
occurrences <- read.csv('path_to_gbif_data.csv')

# Run MaxEnt model
maxent_model <- maxent(env_vars, occurrences[, c('decimalLongitude', 'decimalLatitude')])

Model Evaluation

Model performance is evaluated using metrics such as Area Under the Curve (AUC) of the Receiver Operating Characteristic (ROC) curve, True Skill Statistic (TSS), and others. Cross-validation techniques are used to assess the robustness of the model.

R
# Evaluate model performance
evaluate(maxent_model, env_vars, occurrences[, c('decimalLongitude', 'decimalLatitude')])

Mapping Suitable Habitats

The final step involves generating a suitability map using the trained MaxEnt model. The suitability scores indicate the likelihood of a given area being a suitable habitat for the target bird species.

R
# Predict suitability
suitability_map <- predict(maxent_model, env_vars)

# Plot suitability map
plot(suitability_map)

Results and Discussion

The suitability map generated using the MaxEnt algorithm provides a visual representation of potential bird habitats. Areas with high suitability scores are likely to support the target bird species, while areas with low scores are less suitable. The results should be interpreted in the context of the chosen environmental variables and the quality of the occurrence data.

Model Performance

The AUC value obtained from the ROC curve analysis indicates the discriminatory power of the model. An AUC value closer to 1 suggests excellent model performance, while values closer to 0.5 indicate a model no better than random chance. The TSS value complements the AUC by considering both sensitivity and specificity.

Implications for Conservation

The suitability map can inform conservation efforts by identifying critical habitats that need protection. It can also aid in planning habitat restoration projects and monitoring changes in habitat suitability over time.

Conclusion

The integration of GEE and the MaxEnt algorithm offers a powerful approach for mapping suitable habitats for bird species. By leveraging remote sensing data and advanced machine learning techniques, conservationists can gain valuable insights into species distributions and make informed decisions to protect biodiversity. Future work could explore the inclusion of additional environmental variables and the application of this methodology to other species and regions.

References

1. Phillips, S. J., Anderson, R. P., & Schapire, R. E. (2006). Maximum entropy modeling of species geographic distributions. Ecological Modelling, 190(3-4), 231-259.
2. Hijmans, R. J., Phillips, S., Leathwick, J., & Elith, J. (2017). dismo: Species Distribution Modeling. R package version 1.1-4. Retrieved from Link
3. Kaky, E., & Gilbert, F. (2019). Using species distribution models to assess the importance of Egypt's protected areas for the conservation of medicinal plants. Journal for Nature Conservation, 47, 111-119.
4. Global Biodiversity Information Facility (GBIF). (2021). GBIF Occurrence Download.

Bibliography

• Elith, J., Graham, C. H., Anderson, R. P., Dudík, M., Ferrier, S., Guisan, A., ... & Zimmermann, N. E. (2006). Novel methods improve prediction of species’ distributions from occurrence data. Ecography, 29(2), 129-151.
• Franklin, J. (2010). Mapping Species Distributions: Spatial Inference and Prediction. Cambridge University Press.
• Peterson, A. T., Papeş, M., & Eaton, M. (2007). Transferability and model evaluation in ecological niche modeling: a comparison of GARP and Maxent. Ecography, 30(4), 550-560.
• Yackulic, C. B., Chandler, R., Zipkin, E. F., Royle, J. A., Nichols, J. D., Campbell Grant, E. H., & Veran, S. (2013). Presence-only modelling using MAXENT: when can we trust the inferences?. Methods in Ecology and Evolution, 4(3), 236-243.