Blog

Exploring Climate with GIS, and Data Science

Browsing for something specific? The archive lists every post by year.

Two runs are written to be read in order rather than dipped into. Bias Correction is 30 posts over two and a half years, from why Indonesia is the hardest place on earth to measure rain through to the published paper and the thesis behind it. El Niño 2026 is 10 posts and counting on the forecast event as it develops, and on why drought is what arrives minus what leaves.

Twenty-six hours to three: a global drought index at four kilometres

Data Science

Climate

Seventy-six years of monthly water balance for every land pixel on Earth at four kilometre resolution used to take twenty-six hours to compute. It now takes under three.…

Categories

Twenty-six hours to three: a global drought index at four kilometres

Into early 2027, and how little forty-four events support

Where El Niño drought risk is real, and where it is not

What the five closest El Niños did to the land

Where the world is dry right now, in rainfall and in thirst

Ten minutes over the Sunda Strait

Rain and thirst move together under El Niño, which is worse news

El Niño starts dry spells, it does not make them longer

Drought has two halves, and most maps measure one

The same thesis, in a different typeface

Forty-four ENSO events, and only eight that can answer anything

Which El Niño are you measuring? The index changed this year

A very strong El Niño is forecast, and nothing in the record matches it

Two years, a full-time job, and a master’s

The thesis, in LaTeX and in git

The defence that reframed the thesis

What the reviewers actually said

Published: the paper is out in Remote Sensing

Making research citable

NetCDF to H3: getting climate data into a join

Measuring whether your own work reproduces

The dashboard nobody asked for

Migrating a Squarespace site to Quarto

precip-index: fit once, not every month

Documenting a research project so someone else can run it

Sentinel-1 flood mapping you can argue with

csv2gpkg: 184 million buildings, 250,000 at a time

Scoping a thesis: the questions I answer with no

30 Day Map Challenge, 2025

The twenty-three hour day: satellite and gauge calendars

Global EDDI in Earth Engine, by precomputing the expensive half

Why bias correction cannot fix your correlation

The bugs that never raised an exception

Where the heavy-rain skill curve stops

Admin boundaries: drawing each border only once

Dry days and detection: every improvement costs something

Forty gigabytes that cannot live in git

The CNN that refuses to help where gauges are thin

Earth Engine has no rename

SPI in xclim: fifty-five minutes to twenty-four seconds

Word Clock

TensorFlow will not install, and it is not your fault

EDDI, a drought index that never looks at rainfall

Two convolutions and fifty million parameters

Linear scaling: start with the dumbest thing that works

Thirty-one metrics into one number

NetCDF that other tools can actually read

Finishing the HarvardX data science certificate

The tail that quantile mapping flattens

Two ways to cut night lights into six classes

How to read a Taylor diagram

Correct the parameters, not the data

Rainfall to flood probability: two numbers decide everything

xkcd style for Country map

xkcd style for LSEQM illustration

Skip PEARSON fitting on climate-indices python package

Utilizing CUDA

Three datasets, three different truths

L-moments, and why ordinary moments let you down

Ten days: why the dekad is the right time step

Why Indonesia is the hardest place on earth to measure rain

Two kinds of zonal statistics in Earth Engine

Maximizing Thinkpad T14 Gen 2 AMD

Drought Propagation

Firmware upgrade on Thuraya SatSleeve for iPhone

Hourly Humidity Data

Annual maximum daily rainfall, and the day it fell

A certified GISP

Monthly mosaic of modified Radar Vegetation Index

Parsing BMKG’s daily climate data

SPI-based drought characteristics

Fourier regression model to generate monthly to daily temperature data

Regression analysis with dummy variables

Sentinel-1 modified Radar Vegetation Index

Second-order Markov chain model to generate time series of occurrence and rainfall

Impact of climate change in cities

Fuzzy Inference System (FIS) for Flood Risk Assessment

Visualising the WRF output

Monitoring dry and wet season in South Sudan