Fire season outlook, Bogota and the Amazon
Fire season outlook for the Bogota Cerros Orientales and the Amazon
Issued 26 July 2026, using soil moisture through week 29 and the Pacific ocean index through May 2026. Every figure below is out of sample.
In plain language
Forests catch fire when they dry out, and they dry out on a schedule that is partly written months in advance by the ocean and by how much water the ground is already carrying. We measured how dry each place is every week since 1950, compared this year against thirty years of its own normal, and then asked a simple question: knowing only what we know today, how likely is the coming dry season to be one of the bad ones. For Bogota the answer is 68 per cent, against a normal year rate of 33 per cent, and the model that counts individual dry weeks is even more emphatic at 79 per cent. For the Amazon it is 63 per cent. In both places the odds have roughly doubled against an ordinary year. This is not a prediction that the mountains will burn. It is a statement that the deck has been stacked, and that the cheapest moment to respond to a stacked deck is now, while responding still means staging equipment and briefing brigades rather than fighting a fire that has already started.
Two honest limits. Four months out we can tilt the odds but we cannot predict the number, so this is a reason to prepare and not yet a reason to spend heavily. And the picture sharpens a great deal in early September, when the next ocean reading is published and the Amazon model roughly quadruples in reliability.
The four forecasts issued today
The bar is the probability the coming season is critical. The circle is how often that happens in an average year, so the gap between them is the whole content of the forecast. A critical season is one in the driest third of the record, and those are the seasons that burn: in Bogota they carry 2.9 times the fire detections, in the Amazon 2.1 times the burned area.
Show the numbers
| Forecast | Probability | Average year | Central value | Critical when |
|---|
How dry every site is right now
Each bar is the average of the last eight weeks, measured against that site's own normal for the time of year. Zero is normal, minus one is the level at which fire activity in Bogota doubles. Bogota itself is near normal today, in the middle of its wet season, which is exactly why a forecast is needed rather than a thermometer.
Show the numbers
| Site | Index, last 8 weeks | State |
|---|
Does waiting make the forecast better
Yes, and by a lot in the Amazon. Discrimination is the chance the model correctly ranks a bad season above a quiet one, where 0.5 is a coin toss. Waiting from 1 August to 1 September lifts the Amazon from 0.69 to 0.91, because by then August itself is on the record and August is most of the answer. For Bogota the equivalent jump comes on 1 December, the day the season opens.
Show the numbers
| Model and issue date | Discrimination | Skill | Correlation |
|---|
When would each biome reach the critical level
The critical level here is the weekly action line, a moisture index at or below minus 1.0 held for two consecutive weeks so that a single dry week does not count. The curves combine the crossing dates of the past 76 seasons with the probability each basin ends up in a critical season, so they answer when rather than whether. Read them as the chance the crossing has already happened by that date.
Show the milestone dates
| Milestone | Amazon | Bogota |
|---|
The one variable to watch in each window
Each window has a single number that carries most of the information available at that moment, with the level that should trigger action. Hit rate is the share of critical seasons the trigger catches, false alarm rate the share of quiet seasons it wrongly flags.
What this supports doing
Now, on the August issue
Low regret preparation only: brigade readiness, staging equipment, checking water access on the eastern slope, and starting the public communication cycle. A false alarm at this stage costs a few weeks of standby. Missing a real one costs a burned slope, decades of restoration, and a smoke event over a city of eight million.
Early September, when the ocean index updates
Reissue both forecasts. This single update takes the Amazon from a watch to a usable warning and adds the September Bogota model. The Bogota decision point for anything expensive or disruptive is the 1 December issue, which gets three quarters of bad seasons right while crying wolf on fewer than one quiet season in ten.
Note: Only a member of this blog may post a comment.