How Do Human-Modified Landscapes Shape Bat Observations in California?

Environmental Data Science Capstone Project
Michele Perry | Yale School of the Environment | 2025–2026

---

Overview

This project examines how human-modified landscapes shape bat observation patterns across California. Using 455 publicly available occurrence records from GBIF for three focal bat species (Aeorestes cinereus, Myotis yumanensis, Myotis californicus), observations are mapped onto a 10 km equal-area grid and analyzed with logistic regression to assess whether species respond differently to an urban intensity gradient.

Key findings

1. Population density alone predicts bat presence nearly as well as multi-variable models (cross-validated AUC ~78%). At 10 km resolution, nighttime light and percent developed land are highly correlated with population density and add less than 1% predictive improvement.

2. All three species show positive associations with increasing urban intensity, consistent with observer bias in opportunistic data.

3. Species differ in response strength. Myotis californicus shows a notably weaker association with the urban gradient than the other two species, suggesting greater sensitivity to human-modified landscapes.

A note on observer bias

GBIF data are opportunistic, meaning they reflect where people looked, not just where bats are. Population density captures both ecological signal and reporting effort, so results are interpreted as patterns in observed presence rather than confirmed ecological preference. The differences between species are more informative than the absolute trends.

---

Repository structure

├── capstone_analysis.Rmd           # Main narrative notebook (source)
├── capstone_analysis.nb.html       # Pre-rendered HTML notebook (download to view)
├── capstone_analysis.md            # GitHub-rendered version (view on GitHub)
├── capstone_analysis_files/        # Figures for the .md rendering
├── preprocessing_scripts/          # Data processing pipeline (scripts 00–06)
│   ├── 00_setup.R
│   ├── 01_get_clean_bat_points.R
│   ├── 02_get_ca_boundary.R
│   ├── 03_build_grid_and_domain.R
│   ├── 04_build_grid_covariates.R
│   ├── 05_build_grid_presence.R
│   ├── 06_build_grid_model_dataset.R
│   ├── run_all.R
│   └── README.md
├── data/
│   ├── raw/                        # (gitignored) Downloaded source data
│   ├── processed/                  # Key processed outputs (see data/README.md)
│   ├── README.md                   # Data acquisition instructions
│   └── data_dictionary.md          # Variable descriptions for final dataset
├── communication/                  # Presentation materials
│   └── lightning_presentation_slide.png
├── renv.lock                       # R package dependencies
├── .gitignore
└── LICENSE

---

Reproducing this project

Quick start (just want to read the analysis?)

Browse on GitHub: Open capstone_analysis.md to view the rendered markdown with all figures directly in your browser.

Full interactive version: Download capstone_analysis.nb.html and open it in any browser for the HTML notebook with collapsible code sections.

Re-run the analysis locally

The repository includes the final processed dataset and key spatial files, so you can re-knit the notebook without running the full pipeline:

1. Clone the repository
2. Restore the R environment: renv::restore()
3. Open capstone_analysis.Rmd in RStudio and knit

Full pipeline (from raw data)

To reproduce everything from scratch, including downloading raw data:

1. Clone the repository:

   git clone https://github.com/Michele-75/california-bat-urbanization.git
   cd california-bat-urbanization

2. Restore the R environment: renv::restore()

3. Set GBIF credentials in .Renviron:

   GBIF_USER=your_username
   GBIF_PWD=your_password
   GBIF_EMAIL=your_email
   

4. Download raw covariate rasters (see data/README.md for sources and file placement)

5. Run the pipeline: source(here::here("preprocessing_scripts", "run_all.R"))

6. Knit capstone_analysis.Rmd

---

Data sources

Dataset	Source	Resolution
Bat occurrences	GBIF	Point records, 2012–2024
Nighttime light	VIIRS VNL v2.1/v2.2	~500 m annual composites
Land cover	NLCD 2019	30 m
Protected areas	PAD-US 4.1	Vector polygons
Population density	GPWv4 rev. 11 (SEDAC/CIESIN)	~1 km, 2020
State boundary	US Census TIGER/Line	via tigris R package

---

Focal species

Species	Common name	Ecology
Aeorestes cinereus	Hoary bat	Long-distance migratory, tree-roosting
Myotis yumanensis	Yuma myotis	Water-associated, often found near bridges and buildings
Myotis californicus	California myotis	Crevice-roosting, widespread across the state

---

Related work

An interactive R Shiny application for exploring California bat occurrence patterns is available here: California Bat Occurrence Explorer.

Acknowledgements

This project was completed as part of the Yale Environmental Data Science Certificate Program, 2025–2026 cohort. I am very grateful to the program's leaders, professors, and mentors for their guidance and support. Capstone projects from the full cohort can be viewed at the Cohort 1 Capstones repository. Bat occurrence data were provided by the Global Biodiversity Information Facility (GBIF).

License

This project was created for educational purposes as part of a graduate certificate program. All code and original text are provided as-is for learning and reference. Data used in this analysis are publicly available from the sources cited above and are subject to their respective terms of use. GBIF-mediated occurrence data should be cited according to GBIF citation guidelines.