rainfall-trend-analysis

This repo contains GEE script to find the trend analysis for the last 25 years in the Bihar state of India.

Bihar Annual Rainfall Trend Analysis (2000–2025)

Spatial trend analysis of annual rainfall across Bihar, India, using CHIRPS precipitation data in Google Earth Engine, with trend magnitude and significance mapped separately and compared.

What this does

• Aggregates CHIRPS pentad/daily precipitation into annual totals per pixel (2000–2025)
• Computes Sen's Slope (Theil-Sen estimator) to quantify trend magnitude (mm/year)
• Computes Kendall's tau to assess trend consistency/significance
• Visualizes both as spatial rasters, styled and compared in QGIS

Data

• Rainfall: CHIRPS Pentad/Daily, UCSB Climate Hazards Group
• Boundary: Bihar state, dissolved from district-level administrative boundaries

Methodology notes

• Sen's Slope was used over ordinary least-squares regression for magnitude estimation, since it's more robust to outlier years (e.g., a single extreme flood/drought year skewing an OLS fit).
• Kendall's tau (via ee.Reducer.kendallsCorrelation()) indicates trend consistency, not a formal p-value — treat as a relative significance indicator, not a hypothesis-test result.
• Color stretch for both rasters uses a percentile-based (rather than raw min/max) range, since a small number of outlier pixels otherwise dominated the visual scale. Ranges are symmetric around zero to keep the diverging color ramp meaningfully centered.
• Observed ranges: Sen's Slope roughly -22 to +25 mm/year; Kendall's tau roughly -0.40 to +0.33 — the moderate tau ceiling suggests spatial trends, while present, are not strongly monotonic anywhere in the state.

Output

Left: Sen's Slope (mm/year). Right: Kendall's tau. Rendered in QGIS with matched extents and independent legends/color ramps.

Files

• rainfall_trend_analysis.js — full GEE script (data prep, Sen's Slope, Kendall's tau, visualization)
• images/bihar_slope_tau_comparison.png — final QGIS comparison figure

Caveats

• 5566m analysis scale (CHIRPS native resolution) — not suited to sub-district-level claims.

• Trend estimates are annual-total-based; no seasonal (e.g., monsoon-only) breakdown in this version.

• Kendall's tau here approximates trend consistency; a formal Mann-Kendall significance test (with variance-corrected p-values) would strengthen any claim of statistical significance.