Friends of the Urban Forest

Organization and scope

FUF's implementation role includes working with residents and public agencies, selecting planting sites and species, establishing trees, organizing community planting, and caring for young trees. The intervention is therefore not simply handing out a seedling. Site approval, infrastructure, labor, water, and several years of maintenance can all be necessary complements. An inexpensive tree by itself is not the complete health-producing service.

The FY2025 impact report gives the clearest newly checked annual whole-organization planting count: 1,506 new street trees. It calls this the largest annual total in a decade. Its neighborhood table includes 245 in the Mission, 204 in the Sunset, 184 in Bayview, 136 in Visitacion Valley, 86 in the Excelsior, and smaller totals elsewhere. These counts support San Francisco implementation, not measured resident-level exposure or a finding that all beneficiaries live in the city. The model's 100% Bay and 100% San Francisco health shares are nested beneficiary-geography assumptions. Commuters, visitors, migration, and regional ecological effects can make location and residence differ.

The same annual report records 2,598 care visits and about 1,200 trees watered weekly. Care visits are repeated work on a stock of trees, potentially including earlier planting cohorts. Adding them to new plantings would double count exposure and mix flows with stocks. Similarly, the workforce program's 619 trees are part of the organization's planting activity, not an extra health denominator. Eighteen workforce participants, 1,846 volunteers, and 6,806 volunteer hours are useful operational information but not automatically additional healthy life-years.

A city workforce report presented July 10, 2025 records 1,502 street trees as of June 30, 2025, the same care-visit total, and approximately the same weekly watering stock. The four-tree difference is small, but we do not force the sources to agree by inventing a reconciliation. The official annual report's explicit FY2025 total is preferred; the city total is retained as a sensitivity. This replaces the previous 1,412 grant-associated FY2024 count, which was not proven to be a full organization census.

Sidewalk gardens, yard-tree distribution, youth and workforce activity, and community engagement are all inside whole-organization expense. We do not allocate their costs away to make the tree pathway appear cheaper. We also do not add separate health effects for them without evidence. This asymmetry makes the estimate partial in outcome coverage, but not a lower bound: the quantified street-tree scalar could itself overstate benefits.

FUF's current yard-tree page describes 400 distributed cumulatively since February 2023 and funding through August 2027, including an Arbor Day Foundation Community Roots grant for the current season. Those are not another 400 annual street-tree plantings. The funding statement is relevant to baseline substitution in that program but cannot be generalized to the entire street-tree portfolio.

The annual delivery and resource counts above come from the FY2025 official impact report, compared with the July 10, 2025 city presentation. The separate funded program is described on the yard-tree page.

Finances and resources

We inspected the three latest located original Form 990 returns, not just an aggregator's summary. All use accrual accounting. Fiscal years end in June. Original return identifiers are 202440179349300544 for FY2023, 202541029349300739 for FY2024, and 202611049349302071 for FY2025. The latter was filed April 14, 2026. No later original filing was located in this review.

FUF — functional expense shares

Fiscal years ended June; original whole functional expense, not lower annual-impact-report expense.

Shares are calculated as category expense divided by the stated annual denominator. Rounding may make displayed shares differ slightly from 100%.

The functional expense rows sum exactly to whole expense in each year. Government grants are a component of revenue, not extra resources to add to total revenue. The original event, gaming, and inventory direct-expense lines do not show a separate netted expense requiring an add-back. This differs from organizations where a return's headline expense omits costs netted against fundraising revenue.

FY2025 government grants of $3,761,445 are about 73% of total revenue. This is strong evidence that public funding is a central part of the organization, not a minor outside alternative. It is not evidence that 73% of every donor's gift is displaced, nor that all government revenue is restricted to the street-tree output used here. Grant restrictions, renewal dates, allowable costs, and staff capacity determine how an ordinary gift changes the funded plan.

Reported deficits were $437,089, $666,455, and $268,035 across the three years. Falling net assets can support concern about financial sustainability, but a deficit is not a priced marginal planting opportunity. A gift might preserve staffing, meet restrictions, rebuild liquidity, reduce a planned drawdown, or expand an approved planting tranche. These uses have different counterfactual effects. The model's funding factors are explicit priors over such uncertainty, not conclusions drawn mechanically from the deficit.

FY2025 total assets were $3,342,184 and liabilities $1,015,116. Cash plus savings was about $1.055 million, while pledges receivable were $1,439,190. Receivables are not cash on hand and net assets are not synonymous with unrestricted spendable reserves. This is one reason to call the denominator reported accounting expense rather than verified cash spending. Public information here does not identify the unrestricted cash runway or the status of every pledge.

The original return's expense categories also show why whole-cost accounting matters. FY2025 officer compensation was $453,606, other salaries $2,595,830, benefits $478,549, payroll taxes $253,512, and occupancy $207,055. These are subsets of total expense, not additional costs. Skilled and administrative labor cannot be excluded merely because the visible output is a tree. Nor does a workforce program's social value make its staff and support costs disappear.

The official impact report presents a different financial summary: revenue $5,232,031.76 and expense $5,391,171.48, with program expense $3,965,388.77, administration $866,999.17, and development $558,783.54. Compared with the original filing, its expense is $18,923.52 lower and revenue $89,971.76 higher. No reconciliation was located. We use the original filing consistently, disclose the discrepancy, and do not combine whichever components produce a cheaper result. The discrepancy is too small to resolve the main health and additionality uncertainty.

Donated resources require a separate boundary. The FY2025 report supplies 6,806 volunteer hours, a useful improvement over the prior unmatched information. It does not supply a verified opportunity-cost valuation. Applying illustrative values of $25 or $50 per hour gives $170,150 or $340,300 annually. Allocating those totals by the modeled gift's expense share adds approximately $3,145 or $6,290 to a $100,000 gift and produces about $971,306 or $1,000,922 per 10 modeled Bay QALYs. These illustrations assume volunteer input scales proportionally; that has not been established.

They are also incomplete. They omit some public maintenance, water, resident time, land or infrastructure opportunity costs, and potential offsets. They should not be added to the inherited gross multiplier as though the two inventories were disjoint. The inherited multiplier already intends to stress unpriced complements. Its weighted $115,750 resource numerator produces $1,090,005 per 10 QALYs, but neither that figure nor the hours-only illustration is an audited full societal cost.

Accounting sources: original FY2023 return, FY2024 return, and FY2025 return. The differing public summary is on page 12 of the official impact report linked above.

Health evidence

The key clinical question is not whether green neighborhoods are pleasant or whether FUF plants trees. It is how much additional health is caused by the incremental surviving exposure created by an ordinary gift, relative to what would otherwise happen. The primary literature supports plausible mechanisms but does not answer this organization-specific question at the precision implied by a single cost-effectiveness figure.

Portland mortality association: correct denominator, limited causal transfer

Donovan and colleagues' 2022 Environmental International paper examines nonprofit tree planting and mortality in Portland census tracts. The outcome is an annual death rate per 100,000 tract residents. A coefficient of approximately −0.207 associated with an additional tree planted during the preceding 15 years is not 0.207 deaths prevented per tree. Using the reported mean tract population of 4,318 converts that rate association to approximately 0.00894 deaths per year per additional tract tree. Even this correctly scaled quantity remains an ecological association, not an identified causal effect of an extra FUF tree.

The paper's city-scale illustration of approximately 1.33 fewer deaths refers to a tree in each of 140 tracts, not one tree in the city. Confusing those denominators would materially exaggerate impact. Our preserved model never uses either coefficient in its arithmetic. We therefore do not claim that this V2 update repairs a previously misapplied mortality denominator. The model instead contains a separate, explicitly unvalidated QALY-per-tree-year judgment.

Age matters in the Portland results. The 1–5-year planting age band has a confidence interval including no association; older bands show larger associations. Nursery trees were already several years old at planting. Years since planting, biological age, surviving tree count, and canopy size are not interchangeable measures. An annual constant health scalar applied immediately to every surviving new planting can miss maturation and delay, unless interpreted as an already averaged effective scalar. The current base is not fitted to an age-specific health curve.

Important threats to causal interpretation remain: residual neighborhood confounding, changes in who lives in greener tracts, selection of planting locations, unmeasured exposure, and lack of matched individual health histories. Species, survival, and context also differ. These concerns justify explicit uncertainty; they do not prove that all tree health effects are zero. We preserve the old priors rather than quietly replacing them with the observational coefficient or treating uncertainty alone as a demonstrated mathematical error.

These denominator and age distinctions are checked against the Portland primary paper, rather than a news summary.

Randomized vacant-lot greening: a package, not a tree coefficient

South and colleagues' 2018 cluster-randomized Philadelphia trial enrolled 442 adults around 110 clusters comprising 541 vacant lots; 342 completed follow-up. The greening intervention included trash removal, grading, grass, a small number of trees, fencing, and continuing maintenance. The reported relative reductions in feelings of depression and worthlessness were 41.5% and 50.9%, respectively. Overall poor mental health was not significantly reduced.

Those percentages are not absolute percentage-point reductions in diagnosed depression. They cannot be multiplied by every nearby resident without the trial's outcome definitions and baseline risks. Nor can the result isolate the tree component from removing blight, maintenance, or fencing. We use the trial as evidence that a physical neighborhood intervention can change some self-reported mental-health outcomes, not as a numerical QALY conversion for San Francisco street trees.

The trial's design and reported outcomes are available in the primary article record.

Green Heart and intermediate outcomes

The University of Louisville's Green Heart project reports roughly 13–20% lower high-sensitivity C-reactive protein in the greened area after substantial tree and shrub planting. This is relevant biological evidence but an intermediate marker. Converting a biomarker change into prevented heart attacks or deaths requires a validated causal bridge; association of a marker with risk is not sufficient. No separate clinical-event credit is awarded here.

The intervention intensity, neighborhood climate, planting mix, and baseline exposure differ from FUF's. A coordinated neighborhood canopy increase is not necessarily equivalent to a few dispersed additional trees. It may be reasonable to retain a positive prior, but it is not reasonable to present the project's biomarker result as measured FUF mortality reduction.

The biomarker description is the research team's Green Heart project report, not a mortality trial result.

Survival and the difference between time and canopy

Hilbert and colleagues' 2019 urban tree mortality review reports wide variation across studies. For planting cohorts, reported quartile ranges include roughly 2.8–3.8%, 4.4–6.5%, and 7.1–9.3% annual mortality; mortality is often greatest in the first five years. These ranges support considering materially different survival assumptions, not selecting a precise local rate. Site, species, maintenance, size, climate, and vandalism matter.

Our central model applies a 75% establishment survival haircut and 6% annual mortality thereafter, with a 12-year finite horizon and 3% discounting. The resulting 5.634 discounted surviving tree-years per incremental planting is a computational quantity, not observed FUF survival. Applying an immediate establishment haircut and then annual decay is a simplified prior, not a fitted cohort model. Depending on interpretation, it can overlap early mortality already represented in published annual cohort rates. We retain it transparently rather than claiming those parameters are independently estimated.

Survival-weighted tree-years are not age-weighted canopy-years. A tree that survives its first year counts as a year in the base model even if its canopy is small. The three-year lag and ten-year maturation-ramp diagnostics make this distinction visible. They do not establish the correct local growth curve, and they should not automatically be stacked onto a scalar that a reader already interprets as a duration-averaged effective benefit.

A directly local counterpoint deserves attention. Martin, Simmons, and Ashton (2016) studied three ornamental street-tree species in San Francisco after more than 16 years and reported over 70% survival for all three, with important species-by-microclimate differences in health and vigor. The study's primary abstract is reproduced in the National Agricultural Library record distributed by FAO AGRIS; full publisher text was not accessible in this review. This is not a current FUF cohort census, and its selection and maintenance history need examination before calibration. It nevertheless cautions against treating the model's relatively high central mortality as a measured local fact. It also reinforces the distinction between surviving and healthy, growing canopy. No preferred coefficient is changed from this abstract alone. Local study abstract.

Monitoring and decision

Two questions would most directly change the giving decision. First: what dated, approved planting or maintenance tranche would an additional unrestricted $10,000–$100,000 cause, after accounting for current public grants, other expected donations, and reserves? The answer should identify site readiness, staff and watering capacity, spending period, restrictions, and what happens without the gift. A list of general organizational needs is not enough.

Second: for that tranche, how much additional surviving exposure is expected relative to replacement, delayed planting, or already funded maintenance, and what observed survival and growth cohorts support that estimate? Reporting planting dates, deaths, replacements, species, neighborhood, and follow-up canopy or size would help distinguish planted trees from durable exposure. It would still not independently identify a QALY scalar, but it would remove an important operational uncertainty.

For health calibration, a useful study would compare sufficiently similar settings with credible counterfactual exposure and health measurement, accounting for migration and pre-existing trends. A matched observational analysis could inform a prior without being treated as conclusive. A reported biomarker or subjective well-being result should retain its actual outcome definition until a defensible utility bridge is available.

Until then, the displayed estimate should be accompanied by its central case, favorable-tail dependence, whole-cost/partial-health scope, and hold status. The general donation link must not be labeled as a verified marginal offer. Updating the public funding narrative is required: later funded activity is known, even though an additional-gift opportunity remains unverified.

Sources and model version

The accompanying source ledger contains original-return URLs, official annual and city reports, and primary research. Publication dates are included only where supported; an undated current page is not assigned an invented date. All sources were reviewed September 11, 2026. The source ledger includes excerpts and links supporting the assessment; the previous model remains available for comparison.

Model version: FUF V2, source-only FY2025 expense/output update with unchanged signed priors; finite age-lag, geography, funding, exposure, and resource diagnostics. The historical executable and output are preserved. No current marginal giving offer is verified.

Limitations and qualitative judgment

This report puts a whole-organization cost in front of a partial health model. It does not claim complete organizational return. Gardens, training, neighborhood stewardship, biodiversity, aesthetics, carbon effects, and community participation are omitted from quantified benefits. Some could be valuable. Their omission does not justify inserting a generic social-benefit multiplier or calling the calculated result conservative when the main included effect remains unvalidated.

The harm scenarios represent possible health-relevant burdens, not observed FUF injury rates. Allergens, constrained ventilation, root and sidewalk damage, maintenance demands, failure risk, and displacement-related effects can vary greatly by species and site. The model does not separately monetize water or infrastructure costs as health harms and then claim an audited resource total; the categories are acknowledged as incomplete and potentially overlapping.

The Bay-first framing is strongest for the location of delivery. It is weaker for where all health beneficiaries reside. The base 100% Bay assumption is reasonable as an illustration for San Francisco street trees but remains a prior. The model never adds San Francisco impact to Bay impact: San Francisco is nested inside the Bay total. The 95% Bay / 90% San Francisco diagnostic shows the distinction without claiming a measured residence distribution.

The expected result is fragile to a shared group of assumptions: durable survival, meaningful health effect, donor additionality, and enough time for mature exposure. Improving the precision of the four-tree output discrepancy would not materially resolve that fragility. Better source accounting is useful, but the highest-value diligence is counterfactual and health calibration, not polishing the last dollar of the price.

We therefore retain an exploratory model with explicit negative findings. FUF is a verified local implementer with plausible benefits and a meaningful public role. It is not currently a verified marginal health purchase at the displayed price.

Model and results

The preferred update replaces two source inputs: $5,673,426 FY2024 expense becomes $5,410,095 FY2025 expense, and 1,412 credited FY2024 trees become 1,506 official FY2025 street-tree plantings. It changes no effect or probability coefficients. The historical model and complete historical output are preserved separately. The change was locked before the first new output was computed.

The calculation begins with 1,506 × $100,000 / $5,410,095 = about 27.84 gift-linked planting equivalents. This is retrospective allocation of whole cost, not a claim that a donor can purchase 27.84 additional trees. Each scenario applies a funding response to this base, a finite survival curve, a health scalar, and a harm deduction.

For year t from 1 to H, the survival-discount weight is s × (1 − m)^(t − 1) / (1 + d)^(t − 0.5). The sum is multiplied by incremental plantings and the effective QALYs per surviving tree-year. Harm is incremental plantings multiplied by a separate harm-per-tree prior. All harms and benefits share the same modeled geographic attribution. There is no lifetime extrapolation after H and no separate stacking of mortality, heat, mental health, or pollution outcomes.

These weights and parameters are judgments. They are not posterior probabilities estimated from a trial or calibrated frequencies of comparable organizations. The favorable world combines several favorable assumptions, so its contribution is correlated rather than a series of independently established facts.

Central incremental plantings are 11.13474. Multiplying by 5.634318 finite discounted surviving tree-years and 0.010 QALY gives 0.627367 gross QALY. Subtracting 0.022269 harm gives 0.605097 net QALY per modeled $100,000 gift. Ten times the gift divided by that net benefit is $1,652,627 per 10 QALYs. The favorable scenario gives 9.012701 net QALYs and $110,955 per 10 QALYs. The latter is not our preferred estimate.

The weighted calculation first combines signed QALYs across worlds and only then divides cost by expected benefit. It yields 1.061922 net QALYs and $941,689 per 10 modeled Bay QALYs. Averaging scenario cost-effectiveness ratios would be wrong, especially with null and harmful worlds. The positive-only calculation is also not the preferred result because it discards harm.

The historical weighted result was $1,053,267 and historical central result $1,848,441. The approximately 10.6% lower price is entirely explained by the expense/output update. No new study has improved the health bridge, and the favorable contribution remains 84.9%. Crossing a round $1 million threshold is therefore an accounting-and-volume update, not a qualitative validation of the donation.

The half-net-new diagnostic differs from half funding: it reduces benefits while retaining planting-related harms. Its purpose is to expose replacement or short-lived exposure that does not deliver the assumed health stream. It is not an empirically measured replacement rate. The lag and ramp tests similarly bracket structural interpretation; they are not source-estimated age curves.

Funding and counterfactual

The earlier report relied on a city contract that ended in August 2025. That is no longer an adequate description of the publicly visible baseline. A more recent San Francisco County Transportation Authority article describes $1.1 million in transportation sales-tax support alongside $2 million in capital improvement funding and $3 million in federal Inflation Reduction Act funding, within a program of up to 2,500 trees through June 2026. A specific project involves approximately 407 trees in Districts 9, 10, and 11. These are nested program figures, not additive counts, and not all funds are FUF revenue.

A November 2025 Public Works account separately describes a $12 million federal grant for 3,500 trees and more than $7 million in workforce planting and watering grants to local nonprofits, with 1,800 trees and maintenance of up to 3,000 young trees over the following two years. We do not assume these announcements are disjoint from other city programs or that their full balances remain available today. They nevertheless show a funded pipeline extending beyond the old contract. Saying that there is no known later public activity would be wrong.

The same account describes FUF and a neighborhood partner investigating planting sites where earlier trees failed because of vandalism or inadequate care. This makes the replacement and stewardship issue concrete. Replanting can still produce health benefits compared with leaving a site empty, but a new planting count does not by itself identify net canopy gain. If the city or another partner would have replaced the tree shortly anyway, the incremental exposure is only the delay avoided.

Public funding can complement donations as well as displace them. Private money might pay for an unfunded enabling task, maintain staff between reimbursements, or secure sites needed to use an existing grant. In such a case, the gift could be highly consequential. Alternatively, it could finance activity already planned, shift another donor's contribution, or replenish reserves. The public evidence does not choose among these cases. A statement that public funding exists is not a proof of zero additionality; a fundraising appeal is not a proof of positive marginal capacity.

The central funding response of 40% and favorable response of 75% remain inherited analyst priors. They are deliberately not replaced by the share of government revenue or by a contract's dollar total. The model asks what fraction of the historical whole-budget output equivalent becomes additional exposure because of the gift. This factor includes displacement and organizational response, but cannot capture every non-linear bottleneck. A site-specific gift plan could have a different allocation and cost structure.

The direct giving pathway is a general donation route. It is not a verified offer to produce a number of additional trees at a fixed price. Historical cost illustrations, including a roughly $2,500 tree package in an older impact report, are not current marginal quotes. Likewise, corporate sponsorship levels identify recognition categories, not necessarily incremental health-producing capacity.

The $100,000 modeled gift is a bounded illustration, not assurance of linear scaling or spendability. A smaller gift may support fungible operating costs; a larger gift may encounter site, labor, permitting, water, or grant restrictions. We do not authorize an unrestricted $100,000 recommendation on the basis of the expense ratio.

The later funding evidence is in the SFCTA project account and November 2025 Public Works newsletter. The general giving page confirms a donation route, not a marginal quote.