23/07/2026
A few weeks ago I asked myself a simple question: how bad is drought really getting in Punjab, Pakistan — and can I actually prove it with data instead of just going by news headlines?
So I pulled 43 years of climate data (1982–2024) from ERA5-Land satellite reanalysis, and calculated two drought indices — SPI and SPEI — across the whole province at multiple timescales.
The results were more interesting than I expected.
The worst drought in the entire 43-year record turned out to be 1999–2002 — a slow-building, multi-year disaster that peer-reviewed research had already documented. Good, that matched.
But then December 2024 showed something unusual: a sudden, sharp drought signal that barely showed up when I looked at rainfall alone, but became "extreme" the moment I factored in temperature. Turns out October and November 2024 were unusually hot, which quietly increased evaporation and dried things out — even in areas where rainfall wasn't dramatically low. I checked, and Pakistan's own Meteorological Department had flagged the exact same period and the exact same districts in their December 2024 drought bulletin.
That was a good feeling — building something from scratch and then watching it line up with what officials on the ground were independently reporting.
Here's what that divergence actually looks like on the ground — 2002 (the worst year on record) next to December 2024, across two different drought indices.
A few things I took away from this:
Drought isn't one thing. A 3-month view and a 12-month view of the same place can tell completely different stories — one can look fine while the other is screaming "extreme."
Temperature matters as much as rainfall. Ignore it, and you'll miss droughts that are quietly building from heat, not just from a lack of rain.
Always check your work against something real. It's easy to make a good-looking map. It's harder — and far more valuable — to make one that's actually true.
Built using Google Earth Engine, Python, and ArcGIS Pro.
