An independent browser build of the Eradication, Suppression and Surveillance planning model for the brown treesnake on Guam, rebuilt from the public-domain USGS release so the model runs without a spreadsheet. A draft for discussion; not a USGS product and carrying no endorsement by the USGS or its authors. Full provenance is in the footer.
Status. This now reproduces the published worked example exactly, including the residual density of 9.47 snakes per hectare, the large‑snake density of 1.98, the 169 snakes encountered, the 128 recommended site visits, the eradication probability, the detection probability of 99.9 per cent that the documented example records, which Excel produces where the open‑source recalculation underflows, and the $336,900 cost. The demographic engine was cross‑checked against the workbook one size class and one year at a time, and it matches the workbook across a twelve‑point density sweep and seventy further test plans. Ten points where an earlier draft of this page differed from the workbook have been brought into line; several of those are probable slips in the spreadsheet and several are points of definition the model author should confirm, and both sets are listed in the companion note. The goal verdict is judged one year after treatment, which is the basis the workbook uses; the separate long‑run panel shows the sustained tendency, which can differ. These figures are faithful reproductions of the workbook rather than an independent check of it.
You describe a control plan on the left, the model projects the snake population forward year by year, and the predicted result appears on the right. Only the two blocks on the left take input.
YOUR PLAN · INPUTS
Start from
Treatment parameters
2.32
Removal tools
Tick use to include a tool. Aerial Delivery System (ADS) needs only the number of application days; the model derives the rest from the EPA label. Cost is per application day for ADS, and per application day and placement event for the other tools.
PREDICTED RESULT · OUTPUTS
Against your goal
—
Area coverage by tool
Outcome probabilities (workbook‑exact)
Eradication probability (Pε)complete removal in the treated area—
Residual total densitysnakes/ha, one year after treatment ends—
Residual density ≥950 mm (Fe)large snakes governing recovery, one year after treatment—
Functional eradication achieved?large‑snake density below threshold—
Expected individuals encounteredtotal removed over the period—
Surveillance & detection (indicative)
Detection probabilitychance a snake would be seen if present—
Minimum recommended site visitsto support the absence target—
Estimated total costacross all tools and years—
Projected population trajectory
■ total density ■ density ≥950 mm – – recovery threshold. Treatment ends after the chosen period; later years show post‑treatment rebound.
Per‑class removal probability is averaged within each size band, per tool, as the workbook computes it. Whether a within‑band average or a running average across smaller sizes is intended is one of the questions for the model author.
Plain-language summary
—
Long-run tendency (if this annual effort were sustained · corrected engine)
— your treatment window – – continuation if effort were held – – recovery threshold
—
Corrected engine (probable spreadsheet slips removed, pending the model author's answers). This holds the current annual effort beyond the treatment window to show whether the residual is a step toward zero or a plateau, which is the question an end-of-treatment figure alone cannot answer. It assumes immigration continues at the modelled rate and is a direction of travel rather than a year-by-year prediction, since forecasts past roughly fifteen years are unreliable.
Suppress forever, or clear once (decision economics · corrected engine)
horizon yrsfence cost $/mmonitoring $/yr
— cumulative cost of suppressing forever — cumulative cost of clearing once behind a fence, then monitoring
—
Corrected engine (probable spreadsheet slips removed, pending the model author's answers). This monetises the separation choice. It assumes the suppression plan must run every year to hold its plateau, and that a fenced area is cleared once and then only monitored. The fence length is taken from the treated area as a square perimeter, and the fence cost, monitoring cost and horizon are editable assumptions of yours rather than figures from the model, so set them to your own estimates. It is a decision aid, not an accounting statement.
Sensitivity (how much the answer depends on the inputs)
Workbook‑exact. Residual density one year after treatment, recomputed as the least certain inputs are varied. The fenced‑versus‑open switch is usually the input that matters most, the same lesson the barrier and goal panels give from other angles. The response to the starting density is deliberately not smoothed: the published model switches density‑dependent recruitment off at and above about 27 snakes per hectare, so a higher starting density can leave fewer animals recruiting and a slightly lower residual, and the response is not smooth on either side of that cut‑off, moving up and down as the starting density changes near it. That step is a property of the model, and this page now reproduces the converged workbook across the whole range of starting densities, having been cross‑checked against it size class by size class. The button above downloads that run, with this page beside the converged workbook at each starting density.
What separation buys (barrier comparison · corrected engine)
— no barrier (open) — partial barrier or buffer — full snake fence
—
Corrected engine (probable spreadsheet slips removed, pending the model author's answers). The same plan is projected under three degrees of separation, modelled by reducing the immigration that reaches the treated area, with a partial barrier taken to cut most but not all of it. It shows why a bounded area can be cleared while an open one cannot, and it is a direction of travel rather than a year-by-year prediction.
Tool placement height (from behavioural data)
6.0prey index
trap height 1.0 mbait height 0.5 m
— prey-poor reference — prey-rich reference — current site, shaded band is uncertainty. Dots mark a large susceptible snake (~1000 mm) and a small snake (~550 mm).
—
Derived from Nafus et al. (2025), Behavioural plasticity in detection height of an invasive, arboreal snake, Wildlife Research, using the released height dataset. It reflects a cross-site relationship between size, prey and detection height. With the toggle off, or with the tools at their default heights (traps 1.0 m, bait tubes 0.5 m), the forecast is identical to the validated model; raising or lowering a tool away from those defaults then changes its capture in proportion to how well its height matches where snakes of each size are predicted to be. The prey index here is the behavioural-study measure, which combines lizards, birds and mammals and differs from the endotherm count the planner uses, so set it to match your site using the reference points, where a prey-depleted forest is about 1.3 and a prey-rich site about 19. Snake condition is held at the study average.
Compare two plans (A / B)
Pre-filled as an example: A is an open 60 ha patch, B a fenced 5 ha unit. Set up your own plan and press a button to overwrite either column.
Set up a plan, save it as A, change it, then save as B to compare.