Findings

Important findings from 40 years of research in Área de Conservación Guanacaste, and an evidence map of what is missing.

lang = "en"
root = "./"

Research in ACG spans more than 60 years and, as of October 2026, 917 papers in the Atlas. A few results stand out across topics: a degraded dry forest can grow back (often within 50 years), the area holds far more species than anyone expected, and long-term monitoring is now registering losses linked to a drier, hotter climate.

Important findings

1. Protection can bring back a tropical dry forest. ACG was set up to buy land, stop fires and let forest regrow with neighboring communities as partners (Janzen 1986; Janzen 1988; Janzen 2000). Forest, soils, birds and frogs recover over decades without planting (Powers 2015; Owen 2020; Thompson 2023). Succession, Conservation, Ecosystems

2. Far more species than expected, many of them hidden. ACG taxonomists (trained residents who collect and rear specimens) and DNA barcoding have documented over 11,000 of an estimated 15,000 moth and butterfly species (Janzen 2004; Janzen 2016). Barcodes show that many look-alike butterflies, wasps and flies are separate species, and that “generalist” parasitoids are often several specialists (Hajibabaei 2006; Smith 2006; Smith 2008; Burns 2008). Barcoded insects suggest the global count of insect species is well above six million (Colwell 2026). Taxonomy, Interactions

3. One of the largest records of who eats who. Over about 40 years, the caterpillar inventory has logged more than 430,000 rearing records linking caterpillars, food plants and parasitoids (Quicke 2024; Quicke 2025). Interactions, Natural history

4. Animals are part of how the forest works. Monkeys and bats disperse the seeds of many trees (Fleming 1981; Chapman 1989), lianas reduce tree reproduction and growth (Stevens 1987; Becknell 2022), and jaguars hunt nesting sea turtles on Santa Rosa beaches (Escobar-Lasso 2017). Interactions, Populations

5. Climate change is showing up. Long-time observers report falling insect numbers and diversity (Janzen 2021). El Niño droughts stress the dry forest (Zou 2021), and drier, hotter conditions raise fire risk (Campos-Vargas 2021). Repeat surveys follow cloud-forest ants, orchid bees and the boundary between dry forest and rain forest over a decade or more (Warne 2020; Méndez 2022; Walter 2024). Climate, Populations

6. Fire is still the main management problem. Fires remain frequent in the dry season (Campos 2021; Román 2018), and maps of where fires are likely can guide prevention (Vargas-Sanabria 2018). Conservation, Climate

Shared gaps

• Data are rarely open. 917 papers, 235 open datasets. The gap is widest for land and ecosystems (189 papers, 36 datasets) and for geology and people (222 papers, 14 datasets).
• One sector dominates. Santa Rosa accounts for most (about three-quarters) of the papers tagged to a sector in every group except geology and people, where Rincón de la Vieja leads (96 of 148). Rain and cloud forest, Sector Marino and the other sectors are not well represented in this dataset.
• A few groups of organisms dominate. Mammals, moths and butterflies, and wasps lead; microbes, fungi and marine life are scarce.
• Few true time series. Most work compares places at one moment. Exceptions include the caterpillar inventory, the primate project begun in 1983 (Melin 2020) and repeat LiDAR flights (Liu 2023).
• Output peaked in 2011–2020, though 2021 onward is only a partial decade. Grey literature is common: in Succession & restoration, 23 of 77 papers have no DOI.

By theme group

What lives here (Taxonomy, Natural history, Evolution; 481 papers, 128 datasets)

• Barcoding plus community taxonomists is a model for inventorying a tropical country (Janzen 2011).
• The marine inventory, BioMar-ACG, began in 2015 and is adding species fast (Cortés 2020).
• The Santa Rosa primate project, begun in 1983, is one of the longest-running primate studies anywhere (Melin 2020).

How species interact (Species interactions, Populations & communities; 388 papers, 107 datasets)

• Caterpillars, plants and parasitoids form a food web that is more specialised than it looks: many apparent generalists are several specialist species (Janzen 2009; Smith 2007).
• Olive ridley arribadas (mass nesting) at Nancite beach have shrunk by about 90% since the 1970s (Fonseca 2023).

How the land works and recovers (Ecosystem processes, Succession, Climate; 189 papers, 36 datasets)

• Work on how dry-forest trees cope with drought links the closing of leaf pores to the risk of failure in their water-transport tissue (Brodribb 2003).
• Forest age and recovery can be mapped from the air (Arroyo-Mora 2005; Duan 2023).
• Drought kills young trees, and lianas are spreading (Gerhardt 1993; Becknell 2022).

Geology and people (Geology, Conservation & people; 222 papers, 14 datasets)

• Rincón de la Vieja’s hot acid crater lake often produces steam-driven eruptions, yet the volcano is poorly monitored (Battaglia 2019). Newer work tests whether ground swelling gives warning (Mick 2026).
• Santa Elena Peninsula preserves Cretaceous ocean-floor rocks that help reconstruct Caribbean plate history (Baumgartner 2006; Boschman 2019).
• Parks need enough size and buffers to hold off outside pressure (Janzen 1983). Land expropriated for conservation displaced rural families (Lenti 2023).

Evidence and gap map

import { strings } from "/assets/strings.js"
import { split } from "/assets/catalog.js"
t = strings[lang]
papers = d3.json(root + "data/papers.json")
catalog = d3.json(root + "data/catalog.json")
themes = Object.keys(t.themeNames)
decade = y => y ? Math.floor(y / 10) * 10 : null

// Count every (a, b) pair where a and b are values of two (possibly
// multi-valued, ";"-separated) fields.
pairs = (rows, fa, fb) => {
  const m = d3.rollup(rows.flatMap(d => split(fa(d)).flatMap(a => split(fb(d)).map(b => [a, b]))),
                      v => v.length, p => p[0], p => p[1]);
  return [...m].flatMap(([a, inner]) => [...inner].map(([b, n]) => ({ a, b, n })));
}
heat = (data, xDomain, yDomain, yLabel) => Plot.plot({
  width: 1000, marginLeft: 240, marginBottom: 70, height: 28 * yDomain.length + 90,
  x: { domain: xDomain, label: null, tickRotate: -35 },
  y: { domain: yDomain, label: null, tickFormat: yLabel },
  color: { scheme: "greens", type: "sqrt" },
  marks: [
    Plot.cell(data, { x: "b", y: "a", fill: "n", inset: 0.5, tip: true }),
    Plot.text(data, { x: "b", y: "a", text: "n", fill: d => d.n > 25 ? "white" : "#1d2420", fontSize: 11 })
  ]
})
htl.html`<p class="lead-sm">${t.findingsIntro}</p>`

gap = themes.map(th => {
  const p = papers.filter(d => split(d.themes).includes(th)).length;
  const ds = catalog.filter(d => split(d.themes).includes(th)).length;
  return { theme: t.themeNames[th], papers: p, datasets: ds, ratio: p ? ds / p * 10 : 0 };
}).sort((a, b) => a.ratio - b.ratio)
htl.html`<h2>${t.evGapTitle}</h2><p>${t.evGapText}</p>`

Plot.plot({
  width: 1000, marginLeft: 240, height: 340,
  x: { label: t.evGapAxis, grid: true },
  y: { domain: gap.map(d => d.theme), label: null },
  marks: [
    Plot.barX(gap, { y: "theme", x: "ratio", fill: "#b5793e", tip: true,
                     title: d => `${d.theme}\n${d.papers} ${t.paperResults} · ${d.datasets} ${t.results}` }),
    Plot.text(gap, { y: "theme", x: "ratio", text: d => `${d.papers} · ${d.datasets}`, dx: 6, textAnchor: "start", fontSize: 11 }),
    Plot.ruleX([0])
  ]
})

withSector = papers.filter(d => split(d.sectors).length)
sectorData = pairs(withSector, d => d.themes, d => d.sectors)
sectorOrder = d3.groupSort(sectorData, v => -d3.sum(v, d => d.n), d => d.b)
htl.html`<h2>${t.evSectorTitle}</h2><p>${t.evSectorText.replace("{n}", withSector.length).replace("{total}", papers.length)}</p>`

heat(sectorData, sectorOrder, themes, d => t.themeNames[d])

decadeData = pairs(papers, d => d.themes, d => d.year ? String(decade(d.year)) + "s" : "")
decades = [...new Set(papers.map(d => decade(d.year)).filter(Boolean))].sort().map(d => d + "s")
htl.html`<h2>${t.evDecadeTitle}</h2><p>${t.evDecadeText}</p>`

heat(decadeData, decades, themes, d => t.themeNames[d])

taxaData = pairs(papers, d => d.taxa, d => d.themes)
taxaOrder = d3.groupSort(taxaData, v => -d3.sum(v, d => d.n), d => d.a).slice(0, 16)
htl.html`<h2>${t.evTaxaTitle}</h2><p>${t.evTaxaText}</p>`

heat(taxaData.filter(d => taxaOrder.includes(d.a)).map(d => ({ a: d.a, b: t.themeNames[d.b], n: d.n })),
     themes.map(th => t.themeNames[th]), taxaOrder, d => d)

htl.html`<p class="photo-note">${t.evNote}</p>`