Updated Mar 23, 2026 • ~8 min read
Chapter 5: The rainfall data
CAMILA
On day three she sent an email to the regional meteorological database and got back data that didn’t match.
This was not, technically, supposed to be part of her research scope. Her grant was for population dynamics — territorial range, density analysis, conflict behavior. Weather data was peripheral. But she’d been in the eastern sector with Dante Silva for the afternoon on day two, and the afternoon had produced a specific observation about the forest structure in the boundary zone that had sent her to the weather records that evening.
He’d taken her to the eastern boundary himself, which she’d noted. The first survey had been Marco. The eastern boundary survey was the director.
She’d looked at the forest and said: *the canopy structure in this zone is different from the southern sector.*
He’d said: *the eastern boundary has a different elevation profile. The drainage runs differently.*
She’d said: *that would affect the canopy composition but not the structure. This looks managed.*
He’d looked at her.
She’d said: *not managed by a person. Managed by the water table. Has this section experienced extended flooding in the past decade?*
He’d said: *yes. 2019.* And then: *2016.* And then: *it runs about every three years.*
She’d said: *consistent pattern.*
He’d said: *the watershed upstream is relatively predictable.*
She’d been photographing the canopy edge, noting the species distribution, running the standard survey protocol. But the conversation had planted something: a territory with consistent flooding cycles on a predictable schedule had a forest structure that would tell her things about long-term management if she could correlate it with rainfall data.
So she’d asked for the rainfall records.
The regional meteorological database had fifteen years of data for the station’s location. She’d downloaded it and run the correlation analysis that evening and gotten a result that was in the wrong direction.
The flooding cycles that Dante had described — 2019, 2016, three-year pattern — were not the flooding cycles that the rainfall data predicted. The rainfall had been above threshold for a flooding event in 2018, 2015, and 2021. It had been below threshold in the years he’d described.
One of two things was true. Either Dante’s memory of the flooding years was imprecise, or the territory’s flooding cycle was not driven primarily by rainfall.
She sent an email to the meteorological database asking for upstream gauge data.
While she was waiting, she pulled Fonseca’s 2014 hydrological study that Marco had referenced. It was a peer-reviewed paper from the Universidad Federal de Amazonas, co-authored by six researchers, documenting the upstream watershed’s distinctive characteristics in careful scientific language. She read it twice. The upstream watershed had unusual mineral content — this was real, the chemistry data was solid. The watershed had unusual flow consistency — this was also real, the flow data went back twenty years and showed a pattern of unusually steady output for an Amazonian tributary. The paper attributed this to geological factors involving the specific bedrock formation upstream.
She noted: *2014 paper is legitimate. Fonseca’s team is real. The watershed characteristics are documented and real.*
She also noted: *a watershed with unusually steady flow regardless of rainfall would produce unusually consistent flooding cycles regardless of rainfall.*
She pulled up the satellite imagery for the upstream watershed. It was what she’d expected: dense canopy, the tributary snaking through it, the standard visual of Amazon river basin terrain.
She looked at the imagery for a long time.
She zoomed in on the upstream area and looked at the tributary and looked at the forest on both banks and thought about a research station that had been operating for thirty years with relationships across multiple adjacent territories and a watershed with unusually steady flow and a flooding cycle that didn’t match rainfall data.
She thought: *there’s something upstream.*
She thought: *something is regulating the flow.*
She flagged the observation and the question in her notes and did not send a second email about it because one email about weather data was field research methodology and two emails about weather data pointed toward a hypothesis she wasn’t ready to state.
She went back to the species count data that Dante had approved her access to that morning. The size distribution was extraordinary — she’d been expecting the large-female result from her own day-two observation, but the full dataset was more extreme than she’d projected. The population was running fifteen to twenty percent above the regional maximum for healthy jaguar populations across all size categories. The demographic structure was too stable. She ran the standard population viability model with the dataset and got an output that suggested the population had been at carrying-capacity-plus for at minimum twelve years.
She ran it again to make sure she’d entered the data correctly.
Same result.
She wrote: *the population has been at this level for at minimum twelve years. Possibly longer — the dataset only goes back twelve years at this resolution. The population is not trending toward this level. It’s been here. Whatever is producing the anomaly has been producing it for a long time.*
She thought about thirty years of coordinated land management.
She thought about Dante Silva doing personal field surveys.
She looked at the rainfall data and the flooding cycle and thought about something upstream that was managing the water flow.
Her email notification came back from the meteorological database. The upstream gauge data was not available through the regional database; the upstream gauge was maintained by a private monitoring network and the data was not public record.
Private monitoring network.
She wrote that down with three underlines.
She’d met Dante for the eastern boundary survey at two in the afternoon and they’d walked for three hours through territory that he knew with the quality of someone who had walked it for a very long time. Not just familiar — fluent. He moved through the forest the way people moved through their own houses in the dark: no hesitation, no checking, the body knowing the space at a level below decision.
She’d asked him how long he’d been running the station and he’d said twenty years and she’d noted that he looked early thirties and filed it.
He’d stopped twice to read the forest — not with visible attention, but with the quality of someone taking in information she couldn’t access. Once near the eastern boundary, where a section of canopy had shown wind disturbance in the camera array data and she’d wanted to photograph it. Once at a water access point that was too well-maintained to be natural.
She’d said, at the water access point: *this has been cleared recently.*
He’d said: *yes.* No follow-on.
She’d said: *regularly?*
He’d said: *the water access points are maintained as part of the station’s habitat management protocol.*
She’d said: *by staff.*
He’d paused — the same controlled pause from the afternoon session, the one that was deciding rather than retrieving. He’d said: *by various means.*
She’d photographed the water access point and written: *maintained. Various means. Not elaborated.*
She put down the rainfall data and sat back and thought about what she had.
She had: a population anomaly that was real and old. A territory that showed sustained management at a precision that exceeded the station’s visible capacity. A watershed with flow characteristics that were attributed to geology but that she was starting to think might have another explanation. A flood cycle that didn’t match rainfall. A private monitoring network upstream that maintained gauge data that wasn’t public record. A director who knew the territory better than any survey team, who had been running it for twenty years, who approved her access because the population dynamics research was manageable — she’d understood that from day one — but who was showing her more than manageable.
She thought: *why.*
She thought: *he approved my access, he’s personally involved in the surveys, he offered me the local dataset, he took me to the eastern boundary himself. He is showing me more than he has to.*
She thought: *what is he showing me toward.*
She picked up her pen.
She wrote: *Day 3. Rainfall data doesn’t match flooding cycle. Upstream gauge is private monitoring network — not public. Fonseca 2014 is legitimate. Watershed characteristics could explain flow consistency if something upstream is managing the watershed. Population at anomalous level for minimum 12 years. He knows where this is going. He’s deciding how fast to let me get there.*
She looked at the last line.
She wrote: *question: does he want me to find it?*
She sat with the Amazon rain and the river and thought about a man who moved through the forest like it was his house in the dark, and thought she was going to need to go upstream.
Reader Reactions