Published November 1, 2026
In the Everglades, a python survey can come up empty, or remove a snake with a much greater chance of surviving and reproducing. A new University of Florida study brings machine learning into that equation, helping researchers forecast when python-removal efforts may yield captures with greater potential impact.
The model is not a snake-finding app, and it does not tell hunters where an individual python is hiding. Instead, it analyzes patterns in past surveys (among them weather, moon phase and survey conditions) to estimate when crews may be more likely to find pythons and when removals may carry higher demographic value.
That distinction matters as winter approaches. The research found that January and February can look quiet if managers judge success only by how many snakes were caught. But those months may produce a greater share of mature animals, making the removals potentially more consequential than raw totals suggest.
What the model is trying to predict
The peer-reviewed study, published in Ecological Applications, was led by researchers at the University of Florida Institute of Food and Agricultural Sciences (UF/IFAS). It combines demographic information about Burmese pythons with machine-learning methods to help prioritize removal surveys.
The researchers used records from the South Florida Water Management District’s Python Elimination Program. They tested three approaches (generalized additive models, boosted regression trees and neural networks) to predict two related outcomes: whether a survey would find at least one python, and the likely Weighted Removal Index (WRI) of its captures.
The WRI assigns different values to removals based on factors such as a snake’s size, sex, expected survival and reproductive potential. The point is to distinguish, for example, a large, mature female from a small juvenile, rather than treating both as identical units in a tally. These are relative ecological weights, not a literal measure of how many future snakes a removal prevents.
UF/IFAS photo by Alex Romer
This is where the machine learning comes in: it looks across historical survey records for combinations of conditions associated with captures and higher-value removals. The study found that the neural network performed best overall among the three models on held-out data. Its balanced accuracy for predicting survey success was 70.3%. That is a useful improvement over a random guess, not a guarantee that a survey will succeed.
The distinction between the two predictions is important. Finding any python is one outcome; finding one with higher estimated demographic value is another. A useful schedule for maximizing captures may not be identical to one aimed at maximizing WRI.
Moonlight, humidity and temperature, together
UF/IFAS reported that higher-value removal opportunities were associated with warmer, more humid conditions and certain moon phases. The study’s findings also included seasonal and operational patterns: the relationship between conditions and WRI varied, rather than reducing to a single simple rule such as “search only on warm nights.”
That means the model should not be read as a checklist hunters can apply in isolation. A particular moon phase or a humid evening does not guarantee a high-value capture. The model assesses interacting patterns in historical survey data; it is a forecasting aid for prioritizing effort, not a field instruction that overrides local conditions, access rules or safety guidance.
The findings also help explain why January and February deserve attention even when capture totals are comparatively low. The study found a greater proportion of mature males and females among captures in those months. Because the WRI gives more weight to animals with greater expected survival or reproductive contribution, the winter removals can score as more consequential than a raw count suggests.
The researchers’ point is not that smaller snakes should be ignored, or that every winter survey will outperform a summer one. It is that a simple monthly tally can hide differences in the kinds of snakes being removed.
What data went in, and what did not
A key detail: the published model used SFWMD Python Elimination Program records. UF/IFAS identifies those contractor-program data as the basis for the analysis, not public submissions from the Florida Python Challenge.
The distinction is easy to miss because both efforts involve python removals and recorded captures. The Python Challenge requires participants to submit a datasheet for each snake, but the study does not identify Challenge records as its training dataset. It would be inaccurate to say that community-science or Challenge data already feed this model.
The current findings do show how consistent survey records can be used in wildlife management: capture details and survey conditions become inputs that help researchers identify patterns. Whether other datasets, including standardized records from public events, might support future analysis is a separate question. The paper does not establish that they have been incorporated.
UF/IFAS photo by James Whelpley
What this means for hunters and managers this winter
For professional crews and wildlife managers, the model points toward a possible shift from asking only, “How many snakes did we catch?” to also asking, “What was the likely population value of those removals, and under what conditions were they found?”
That could eventually help managers schedule surveys or assess contractor performance with more than capture totals alone. But UF/IFAS describes those as potential applications; the study does not announce a new payment system or a deployed tool that directs crews in real time.
For members of the public, the timing matters too. As of November 1, the 2026 Florida Python Challenge has ended; it ran July 10–19. There is no winter edition identified in the official event information. Burmese pythons may be removed year-round in certain circumstances, but rules vary by location. The Florida Fish and Wildlife Conservation Commission’s removal guidance and the Python Challenge FAQ explain the applicable requirements and safety information. A research forecast is not a substitute for those rules or required training.
The study’s broader message is practical: timing a survey may be about more than maximizing the number of encounters. For a highly cryptic invasive species in a vast, difficult landscape, knowing which conditions have historically aligned with more valuable removals could help make limited field time count.
UF/IFAS photo by Alex Romer
The model is a research advance, not a magic answer for the Everglades. Its forecasts depend on past data, and its reported performance leaves uncertainty. But by combining environmental patterns with information about the snakes themselves, it offers managers a more nuanced way to think about the next survey, and what success should mean.
For more context on the demographic measure behind the research, read our October 22 report on the Weighted Removal Index. The original study, “Prioritizing removals of a highly cryptic invasive species using machine learning and demographic weighting,” is available through Ecological Applications, and UF/IFAS published a research summary.
