Conserving Biodiversity EfficientlyWhat to Do, Where, and When

Imagine you're trying to save as many species as possible, but you can't buy the whole planet or even all the land you'd like. You have a real budget, time is moving, and the threats aren't the same in every place. Hotspot maps and crisis lists are good at showing where life is rich or where habitat is vanishing. They don't tell you what to do next Tuesday with the dollars you actually have. As Hugh Possingham used to quip, conservation is an optimization problem disguised as ecology. Wilson, Underwood, Morrison and colleagues took that seriously and asked a sharper question: if we treat conservation as a sequence of targeted actions, not just land purchase, where and when do we spend to protect the most species per dollar? Their test bed is bold and messy: 17 of the world's 39 Mediterranean-type ecoregions, from California to South Africa to Australia. These are places bursting with endemics and hammered by multiple, persistent threats. The team named every ecoregion-action pairing an "ecoaction," and they didn't limit actions to buying land. Invasive predator control, management of Phytophthora for a soil-borne pathogen, revegetation, and off-reserve agreements—each with a cost specific to that place and a benefit linked to the threat it abates. The goal was clear and quantitative: maximize the total number of plant and vertebrate species conserved, given a fixed annual budget of one hundred million United States dollars. To make money comparable across very different actions, they built costs from regional data assembled with expert input and geographic information analyses. Some actions happen once—like buying land or restoring a site. Others tick like a metronome—predator control you have to do every year. So they converted recurring costs into a one-time equivalent over a twenty-year horizon, using a net present value approach with 3.2 percent inflation and a 6.04 percent discount rate, pegged to a ten-year United States Treasury proxy. They also included ongoing management costs where needed, such as for newly protected areas or private-land agreements. Benefits were tied to risk: how many species in that ecoregion are threatened by a given pressure, and, if you reduce that pressure across some area, how many are predicted to persist? For invasive predator control in Australia, they counted vertebrates only, because that's who the foxes and cats are eating. Now, here's the crux. Returns diminish. The first tract you fence or the first weeds you pull give you a big jump; the next ones a little less. To wire that intuition into the model, they used a species-area relationship, a workhorse in ecology that says the number of species grows as area to a power. In their version, the number of species protected, S, depends on a scaling constant, a, multiplied by the area effectively conserved, A, raised to a small exponent, z. That exponent controls how quickly the curve flattens. They set z to 0.2 as a baseline—typical for terrestrial, non-island systems—and calculate the constant a by dividing total species in the ecoregion by the original habitat area raised to z. Then they swap out area for what really matters to a funder: the cost of protecting that equivalent area. Same curve, but now it's species per dollar. Because there's uncertainty in how steep that curve should be for actions beyond land protection, they reran the schedules thirty times drawing z between 0.1 and 0.4. If the headline survives that wobble, you can bank on it. Choosing among dozens of ecoactions every year for two decades is a combinatorial nightmare, so they didn't brute-force it. They went with a disciplined heuristic that's intuitive to anyone who's ever stretched a grant: maximize short-term gain. Each year, put money into the ecoactions that deliver the biggest immediate bump in species protected per dollar, then update and repeat. They also ran a land-acquisition-only comparison and, to make the budgets commensurate, rescaled the costs there—over the first five years, the acquisition-only case worked out to one hundred forty-eight million dollars to reflect higher total costs. Finally, they checked whether their ecoaction priorities just tracked where vertebrates are rich anyway. Spoiler: not really. Let's get concrete. The Swan Coastal Plain in southwestern Australia is a great case study. Its original habitat is about fifteen thousand two hundred ten square kilometers, and the threat ledger is crowded: five hundred sixty-five species at risk from fragmentation, two hundred fifty-six from the pathogen Phytophthora cinnamomi, and one hundred forty-three from invasive predators. Three actions were on the table there: manage Phytophthora, control invasive predators, and revegetate. The price tags per square kilometer were wildly different—about seven thousand one hundred twenty-five dollars for predator control, three hundred one thousand one hundred eighteen dollars for revegetation, and five hundred fourteen thousand six hundred twenty-six dollars for Phytophthora management. That's not a typo. Predator control is cheap per unit area; pathogen control is expensive. But cost alone doesn't tell you where the payoff sits on the species-investment curve. In a two hundred square kilometer increment, Phytophthora management sat on the steep part of the curve and could protect roughly one hundred eight species. The same area under predator control would add about three vertebrate species. Revegetation would deliver about four. If you flip to a cost lens, an early two-million-dollar outlay in predator control—because it's so cheap per kilometer—could protect about four species. Spend that same two million on revegetation and the marginal gain is negligible; on Phytophthora, the return is still high because you're buying into a threat that hits many species hard. The exact numbers differ by whether you slice by area or spend, but the logic stands: the mix that saves the most species isn't the intuitive "cheap first"; it's the "most species per dollar right now." Across the full set of 17 ecoregions and 51 ecoactions, that logic scaled. Over five years, the ecoaction framework protected about two thousand seven hundred eighty species. Land acquisition alone protected about seven hundred three. That's not a rounding error; it's an average ratio of about three point four nine to one, and that held steady across the thirty sensitivity runs as z wandered from 0.1 to 0.4—the ratio ranged only from three point four six to three point five one. Stretch to twenty years and the gap narrows as diminishing returns bite, but the ecoaction approach still protected more than twice as many species as buying land alone. In other words, targeting threats and sequencing actions pays off, and it keeps paying even after the early, easy wins. What did the early spending actually look like? Concentrated. In the first five years, only twenty-four of the fifty-one ecoactions received money. A big share of funds went to land protection, especially in South Africa—roughly two-thirds of the budget flowed into three South African ecoregions early on. One line item alone, protecting land to slow agricultural conversion in the Montane Fynbos and Renosterveld, drew about twenty-one percent of the entire five-year budget. That's not bias; it's math. The species-investment curves in those places were steep, the threats acute, and the returns high. As years passed, returns flattened in the early winners. The schedule diversified. By year twenty, thirty ecoactions were funded and every one of the seventeen ecoregions received some investment. Here's an important nuance that undercuts a lot of simplistic mapping. When Wilson and colleagues compared their ecoaction rankings with a map of vertebrate richness per unit area, the match was weak. The Spearman rank correlation was about 0.39, with a p-value of 0.12 based on one hundred thousand random pairings to handle non-independence. Translation: chasing raw richness would have sent money to places that weren't the best buys once costs and threat-specific benefits were on the table. Richness is a description. Cost-effectiveness is a decision rule. Let's circle back to the mechanics for a moment because the equation isn't the whole story. The framework assumes the biodiversity impact of one ecoaction is independent of the others. That's a simplifying assumption, and the authors are upfront about it. Real landscapes have interactions: threats co-vary, species depend on multiple processes, and an action might be only partially effective. In their baseline, each action is treated as fully effective at abating its specific threat in the treated area. That can overestimate gains if, say, a plant species needs both pathogen control and habitat restoration to persist. But the structure is transparent, and it's designed to be sharpened as better data arrive—on success probabilities, on how threats overlap, and on how quickly species are lost if nothing is done. What I like about this work is that it turns an abstract budget into a living schedule. It says start where the curve is steep, keep watching as it flattens, and don't be shy about pivoting. In the Swan Coastal Plain, that meant prioritizing Phytophthora management at the outset, because the marginal return was enormous despite the per-square-kilometer price. Predator control, though cheap, bought only a few species per early increment. Revegetation, even more expensive than pathogen management, returned little at small scales. As the model invests and updates, the best buys change. And across regions, that same adaptive pattern emerges: a few high-return actions dominate early; later, the portfolio spreads out. The acquisition-only comparison is worth one more beat, because many conservation plans quietly assume "buy it and biodiversity will come." Wilson and colleagues didn't just pit that against their framework; they tried to make the comparison fair by adjusting for cost structure, which is why the acquisition case carried a higher dollar tag in the first five years. Even with that allowance, acquisition alone underperformed badly in the short term and stayed behind over two decades. That's not an argument against protected areas—land protection was a major, early line item in the ecoaction schedules. It's an argument for treating protection as one tool among many, chosen because of the species it saves per dollar at that time and place. Every model has edges. This one is no different. It leans on the species-area exponent, though the key results were robust to wide swings there. It simplifies interactions among threats and assumes action success. It uses a heuristic, not an exhaustive search, because the latter would take forever and a supercomputer. And it works with the data we have: Red List assessments to tie species to threats, cost estimates that are always improving, and expert judgment to fill gaps. The team even ran a smaller-budget case—ten million dollars per year—to probe sensitivity, though the paper's core story doesn't hinge on the outcome of that variant. So what's the takeaway if you're a funder or a government agency staring at a fixed line in your spreadsheet? Don't chase the biggest blocks of color on a hotspot map. Start with high-return ecoactions tied to real threats, invest until the marginal gain per dollar slips, then pivot. Accept that some places will absorb a lot of money early because the biology and economics line up there. Make room for others as returns level out. And as you go, feed in better data on how actions actually perform, update the curves, and let the schedule move. Wilson, Underwood, Morrison and their collaborators don't claim this is the last word. It's a template you can pick up and run—transparent, scalable, and ready to absorb better knowledge about costs, threats, and success. If future work brings in dynamic budgets, more nuanced threat interactions, and tighter tracking of action outcomes, the framework will get sharper. But the core insight won't change: conservation decisions should be about which actions, where, and when, not just where biodiversity is. In a world of limited money and moving threats, that shift turns a map into a plan.