This repository contains the source code for the quantitative analysis of data center water consumption and drought risk in Texas. It serves as the computational companion to the research paper "The Material Cost of the Cloud."
- Geographic Focus: 70+ Texas counties with active or projected digital infrastructure.
- Temporal Horizon: Baseline analysis for 2025, with efficiency scenarios projected through 2030.
- Key Variables: IT Load Capacity (MW), Water Usage Effectiveness (WUE), Municipal Water Stress, and Drought Intensity.
The primary dataset (data/texas_dc_refined_v2.csv) is a synthesized index constructed by merging three distinct data streams:
- Infrastructure Load: Estimated IT capacity (MW) derived from facility size and power density factors.
- Environmental Stress: Drought data integrated from the U.S. Drought Monitor (USDM) and local aquifer reporting.
- Water Consumption: Calculated using industry-standard Water Usage Effectiveness (WUE) ratios (L/kWh) under three efficiency scenarios (Low, Medium, High).
The analysis pipeline consists of three stages:
- Risk Intersection Model: Triangulates IT load against municipal supply to calculate the "Water Share" metric.
- Unsupervised Learning: Uses K-Means Clustering (
src/spatial_analysis.py) to categorize counties into 4 risk archetypes based on drought intensity and infrastructure density. - Econometric Analysis: Uses OLS Regression (
src/econometrics.py) to quantify the marginal impact of facility agglomeration and cooling efficiency on local water resources.
data/: Contains the refined dataset.src/: Modular Python scripts for data processing, spatial clustering, and econometrics.output/: Contains the 9 generated figures and statistical regression tables.
To reproduce the findings and generate all figures:
- Install dependencies:
pip install -r requirements.txt