Our capacity to forecast climate impacts tends to be embarrassingly off. The best recent example is the United States in mid-2021. A high-pressure system locked some warm air over the Pacific Northwest. Some of that air then descended from the Cascades, triggering compression effects. The result? Normally cloudy, rainy, grungy locales mutated into open ovens for weeks. Portland, Oregon, repeatedly clocked temperatures above 120 degrees. I’ve seen many climate models that suggest the inevitability of hotter deserts or a hotter American South, but none have projected that Portland—freakin’ Portland—would end up being hotter than Las Vegas has ever been. The reason for such a fundamental miss is simple: we do not at present have good enough data to project climate change down to the zip code level.

We’ve obsessed somewhat narcissistically upon how climate shifts might impact cities, and our data simply isn’t good enough to generate reliable forecasts. Instead, we should focus on the far easier math of climate shifts upon continent-spanning climate bands. Not so useful for determining house insurance rates, but critical to understanding what we can grow.

Check out the below map from the agriculture chapter of The End of the World. We sometimes forget that most of the Earth’s land surface is unsuitable for growing crops, and even much of what is is not particularly great. It doesn’t take much of a climate nudge to push marginal lands into the “unsuitable” category. And nudging is certainly happening. For decades, shifts in wind currents have been tweaking regional weather patterns. Simply continuing such long-established patterns for another 30 years will have outsize impacts upon agriculture. Almost everywhere.