The Dove Ledger: 77 Years of Dove Data

Explore it yourself. Every series in this article — national, management-unit, all 41 hunt states, and the 1948–2023 California ledger — is interactive in the companion explorer.

Open the Dove Data Explorer →

Jump to a section

Charts

01 The population 02 National harvest 03 State leaderboard 04 48 years of call counts 05 California, 1948–2023 06 Two official answers 07 Recruitment 08 Effort & efficiency

Part IThe TL;DR — eight numbers that carry the whole story

Doves in eight numbers

337M

Doves in the U.S. last September

The federal integrated population model’s 2024 estimate. No other North American game bird comes close.

14.6M

Doves harvested, 2024 season

By ~947,000 hunters over 2.85 million hunter-days. In most seasons doves out-tally all duck species combined.

1 in 23

Share of the fall population taken

About 4.3%. Doves rebuild it annually: pairs raise multiple broods, and most birds don’t live two years either way.

−36%

National harvest since 2002

22.7M then, 14.6M now — driven more by fewer hunters (1.15M → 0.95M) and fewer birds per hunter (19.7 → 15.4) than by scarcity.

5.1M → 0.8M

California’s bag, 1967 vs. 2023

An 84% collapse from the peak. Dove hunters fell from 278,000 to 47,000 over the same span.

−65%

California’s call-count index, 1966–2013

Doves heard per survey route fell from 25.8 to 9.1 before the survey was retired. The steepest drop of any big dove state.

77 yrs

Of California dove data, assembled here

1948–2023, much of it OCR’d from scanned typewriter reports and cross-checked against each report’s printed year-over-year changes.

Published models that predict a dove hunter’s day

Population science is deep. Daily, hunter-scale prediction — the thing I built for ducks — doesn’t exist for doves. Yet.

If you read the duck series here (Part 1: the white paper, Part 2: the production engine), you know how I work: gather every observable number first, establish honest baselines, then — and only then — try to predict something. This article is that first step for doves. There is no model here. There is a complete, sourced, replication-grade dataset, and a map of where a model could go.

Part IIThe plain-English tour — what the numbers actually say

The most abundant game bird you barely hear about

The mourning dove is a paradox. It is the single most-harvested game bird in North America — in the 2023 season, hunters took an estimated 16.8 million doves against 14.8 million ducks of all species combined — and yet it gets a fraction of the scientific and press attention that waterfowl do. There is no dove equivalent of the duck stamp, no dove refuge system, no breathless annual coverage of the “dove flight forecast.”

Quietly, though, the U.S. Fish and Wildlife Service runs a full monitoring apparatus for this bird, and it produces an estimate that deserves more attention than it gets: roughly 337 million mourning doves were alive in the United States at the start of last September’s season. That number comes from an integrated population model that fuses band recoveries, harvest estimates, and wing-based age ratios — the same class of machinery used for the most heavily managed waterfowl stocks.

FIG 01#

A third of a billion birds

Absolute abundance at the start of September, by management unit, 2007–2024 · shaded band = ±1 SD · millions of doves

The Central unit carries the population. Of 337 million doves in September 2024, about 200 million were in the Central Management Unit, 84 million in the Eastern, and 54 million in the Western. The national total has oscillated between roughly 250 and 420 million with no sustained trend across the 18 years the integrated model covers — abundant, stable, and (as we’ll see) declining only in specific places.

How many get shot — and why that’s fine

Against that population, the harvest is small. The 2024 take of 14.6 million birds is about 4.3% of the September population — one bird in twenty-three. Decades of banding studies underpin the federal harvest strategy’s conclusion that harvest at these rates is largely compensatory: doves are short-lived, fast-breeding birds (a pair can raise three to six broods a season in warm states), and annual mortality is high whether or not anyone is hunting. This is why dove limits have stayed liberal — 15 birds a day in most states — even in units where breeding indices declined for decades.

The more interesting story is not whether hunting hurts doves. It’s what the harvest series says about hunters.

FIG 02#

The national harvest, 2002–2024

HIP-based estimates of dove harvest (bars, millions) and active dove hunters (line, thousands) · United States

A one-third decline, mostly about people. Harvest fell from 22.7 million (2002) to 14.6 million (2024). Hunters fell from 1.15 million to 0.95 million, and each hunter’s season got leaner: 19.7 doves per hunter in 2002, 15.4 in 2024. Handle 2022 with care — the survey moved online before the 2023 season and the Service itself flags 2022–24 estimates as preliminary and possibly non-comparable; the 2022 collapse-and-rebound (8.3M → 16.8M) is at least partly a survey artifact, not a bird artifact.

FIG 03#

Where doves get shot

2024 season harvest by state, top 15 · label = doves per hunter that season

Texas is its own management unit in all but name. 5.4 million doves — 37% of the national harvest — came out of Texas in 2024, more than the next four states combined. California still ranks third at about 731,000 — but that’s a seventh of what the state’s own surveys recorded in the 1960s. Per-hunter efficiency is remarkably flat across big states: most fall between 13 and 19 doves per hunter per season, roughly one limit-and-a-half spread over an average of three hunting days.

The longest yardstick: 48 years of standing on roadsides, listening

Before harvest surveys got rigorous, the federal government measured doves by ear. From 1966 to 2013, observers drove fixed 20-mile routes at dawn each spring, stopping every mile to count every dove they heard cooing. The Call-Count Survey is the longest consistent dove yardstick that exists, and its story is uncomfortable for the West.

FIG 04#

Doves heard per route, 1966–2013

Call-Count Survey annual index by management unit, with California highlighted · birds heard per route

The East held; the West halved; California did worse. Over 48 years the Eastern unit’s index slipped 13%, the Central 33%, the Western 51% — and California specifically fell 65%, from 25.8 doves heard per route to 9.1. The Breeding Bird Survey, which replaced the call count as the official trend source, tells the same story in slow motion: California’s long-term trend is −0.8% per year, compounding since 1966. The birds went somewhere the survey can still hear them — dove numbers grew across the southern Great Plains — but the West Coast breeding population has thinned for two generations.

Part IIIThe California ledger — 77 years, rebuilt from the archives

What California knew before the feds did

Here is the part of this project I care most about, because nobody had put it in one place. California has surveyed its hunters about dove take since 1948 — a half-century before the national Harvest Information Program existed. The Game Take Hunter Survey started as a mail card stapled to license renewals, and its results were published in annual reports that today survive only as scanned typewriter pages.

I pulled all of them — every Game Take Hunter Survey report from 1950 through 2010, plus the modern online-survey reports — ran the scans through OCR, and rebuilt the statewide dove series year by year. Where a scan garbled a number, I recovered it from the next year’s report, which always reprinted the prior year for comparison; where reports disagreed, I kept the revised figure and noted both. Every value in the series carries its source report and a provenance note in the replication data.

FIG 05#

California’s dove bag, 1948–2023

CDFW Game Take Hunter Survey statewide estimates (line) with federal HIP estimates overlaid from 2002 (dashed) · millions of doves · gaps = years the survey didn’t run or didn’t ask about doves

A five-million-dove state, once. California hunters bagged 5.13 million doves in 1967, the all-time peak, with 278,000 people hunting them. By 2023 the bag was under 800,000 and the hunters numbered 47,000 — one-sixth the participation, one-sixth the harvest, and per-hunter take that has stayed surprisingly stable (about 17–18 doves per hunter per season across seven decades). The decline is real birds (Fig 04) and vanishing hunters, in roughly equal measure. Note the gaps: no dove questions in 1949–50, a budget-cut hole in 2009, and a long dark stretch from 2011–2021 when the survey was redesigned and doves were left off the form entirely.

FIG 06#

Two official answers to one question

California dove harvest: state survey (CDFW) vs. federal survey (HIP), years where both exist · millions

They disagreed by up to 100%, then converged. In 2005 the state said 2.03 million doves; the feds said 1.01 million — double, for the same birds in the same fields. Part of it is species scope (the state counted all doves including Eurasian collared-doves; HIP counts mourning and white-winged only), part is methodology (mail cards vs. HIP sampling frames). By 2023, with both surveys modernized and on the same mourning-dove scope, the gap was down to about 9%. If you take one lesson from this figure into any harvest statistic you ever read: these are survey estimates, not counts, and reasonable instruments can disagree by a factor of two.

Part IVThe nerd deep dive — the machinery behind every number

How the federal government counts an uncountable bird

You cannot census 337 million doves. Every number in this article is the output of an estimation machine, and knowing how each machine works is the difference between reading these charts and misreading them. There are five machines.

1 · The Call-Count Survey (1966–2013) — the retired index

Roughly a thousand randomized roadside routes, surveyed once each spring in the last week of May, observers counting doves heard at 20 stops. It measured a relative index, not abundance: 25 doves-heard doesn’t mean 25 doves, it means “more than 12, fewer than 40, on a consistent yardstick.” Its strength was consistency across 48 years; its weakness was that calling intensity varies with breeding phenology, weather, and observer hearing. It was retired after 2013 when the Breeding Bird Survey — more routes, more analysis rigor, hierarchical-model trend estimation — was shown to track it closely.

2 · The Breeding Bird Survey (1966–present) — the current trend source

The BBS is the general-purpose North American bird survey: ~3,700 routes, 50 stops, all species counted by sight and sound. For doves it yields state-level trend estimates with credible intervals from a hierarchical model. The current numbers: Eastern unit +0.2%/yr, Central −0.3%/yr, Western −0.9%/yr long-term (1966–2024), with California at −0.8%/yr. Compounded over 58 years, California’s trend implies the breeding population is roughly 38% smaller than in 1966 — consistent with what the retired call counts said more bluntly.

3 · The Harvest Information Program (1999–present) — the take

Every migratory-bird hunter in every state must register annually (that’s the HIP certification on your license). From that registry the Service samples hunters for detailed diary surveys, and from the diaries estimates harvest, active hunters, and days afield, state by state. It is a two-stage estimate with real sampling error — state-level dove harvests carry confidence intervals of roughly ±10–30% — and it replaced a duck-stamp-based survey that had run since 1952. Doves entered the modern series in 2002, which is why my national charts start there.

4 · Banding (2003–present) — survival and harvest rates

The piece that turns indices into biology. Since 2003, states have banded doves preseason every summer — 788,000 bandings and 39,000 recoveries through 2024, with California contributing about 53,000 birds. Band recoveries give direct estimates of survival (typically ~40–55% adult, lower for juveniles) and harvest rate (mostly 1–9% depending on unit). If you shoot a banded dove and report it, this is the dataset you’re feeding.

5 · The Parts Collection Survey & the integrated model (2007–present) — the synthesis

Selected hunters mail in a wing from each dove they shoot; feather molt tells the ager whether it was hatched that year. The resulting age ratios (juveniles per adult — nationally between 1.2 and 1.9, higher in good breeding years) measure recruitment. The integrated population model then solves for the abundance that makes all the machines agree: if harvest estimates say X birds were shot, band recoveries say that was Y% of the population, and wings say Z% were juveniles, there’s only a narrow range of population sizes consistent with all three. That’s where 337 million comes from — a Lincoln-style estimate wrapped in a state-space model, not a count.

FIG 07#

Recruitment: juveniles per adult in the harvest

Wing-based age ratios by management unit, 2007–2024 (corrected for differential vulnerability) · from the Parts Collection Survey

The engine that absorbs the harvest. Every fall, harvested doves run roughly 1.1 to 2.2 juveniles per adult — the population more than replaces itself annually in most units and years. This is the biological reason a 4% harvest rate is sustainable, and the statistical reason dove populations swing: recruitment is weather-sensitive, so a cold wet spring shows up as a low age ratio that fall and a smaller population the next.

FIG 08#

Effort and efficiency, 2002–2024

Doves per hunter per season, by management unit · HIP estimates, hunter-weighted

The season, not the shooter, sets the ceiling. A Central-unit hunter averages 17–22 doves a season; a Western hunter 12–16; an Eastern hunter 12–15. Those gaps are structural — season timing, bird density, field access — and they’ve persisted for two decades. Within a unit, though, year-to-year swings are modest: the national average has stayed inside 15–20 doves per hunter for the entire HIP era. Hunter efficiency is a remarkably stable quantity, which is exactly what makes it a promising target for the kind of anomaly prediction I do with ducks.

Reading the caveats like an adult

What these numbers can and cannot support

The 2022–24 harvest estimates are flagged preliminary. The federal survey moved online before the 2023 season; the Service explicitly warns the recent estimates “may not be comparable to previous years.” The 2022 dip in Fig 02 is at least partly instrument, not birds.
Species scope shifts under your feet. The California series counts “all doves” through 2010 (including Eurasian collared-doves after their mid-2000s invasion) but mourning doves only from 2022. HIP counts mourning + white-winged. I flag the scope on every row of the dataset.
Indices are not abundances. Call counts and BBS measure trend on a consistent yardstick. Only the integrated model estimates actual birds, and only since 2007.
Survey redesigns leave scars. California’s 1985 report prints no self-estimate; 2011–2021 has no dove data at all; the 1981 values had to be computed from printed percent-changes because the scanned cells were illegible. The provenance notes exist so you can decide which years to trust.

Part VThe gap nobody fills — and why I gathered all this

Deep population science, zero hunter-day science

Survey the dove literature and a pattern jumps out. The population scale is thoroughly studied: survival analysis from bandings, harvest strategies with formal decision frameworks, recruitment modeling from wings, even a national harvest strategy document that sets thresholds for tightening regulations. The hunter scale has a thinner but real literature: managed-field studies (sunflower and wheat plantings measurably concentrate doves), shooting-hours experiments from the 1960s–70s, wounding-loss and nontoxic-shot studies.

But nobody — as far as I can find, anywhere — has published a model that predicts how a dove hunter’s day will go.

For ducks, I showed that California’s public-refuge daily harvest data supports real out-of-sample prediction: a defensible academic model beats history-based guessing by 6.3% at day scale, an exploratory variant doubles that, and a production engine reaches 27% daily and 43% at season scale. The raw material for that work was CDFW’s daily per-refuge hunt results — a public ledger of hunters and birds, day by day, place by place.

The dove equivalent of that ledger mostly doesn’t exist — dove hunting happens on private fields and open BLM ground, not through refuge check stations. But it doesn’t entirely not-exist either. CDFW operates managed dove fields and publishes locations annually; some wildlife areas log dove take in their upland hunt results; and the state series above proves per-hunter seasonal efficiency is stable enough to anchor a model. The honest statement of the opportunity:

What a dove model would need, in order of blockage

A daily or per-trip outcome ledger. This is the hard one. Candidates: CDFW managed-field check data (partial, scattered), eBird-style effort-corrected observations as a proxy for bird presence, or prospectively collected hunt logs.
Covariates — already solved. The weather stack from the duck work (ERA5 reanalysis, day-of-season structure, opener effects) ports directly. Dove-specific additions are known from the literature: September cold fronts push migrants out, grain harvest timing concentrates birds, and the first two days of the season carry an outsized share of the annual take.
A climatology baseline — buildable today. Seventy-seven years of state data and 23 of federal data pin the seasonal averages this article documents. Any future model must beat these numbers out-of-sample, or it doesn’t exist. That’s the standard the duck work set, and publishing the baseline first is the point of this article.

Appendix AData source glossary — everything used, everything checked

Every source, and what it contributed

All raw reports, OCR output, parsing code, and tidy CSVs ship in the public replication package: gitlab.com/duckchugger/dove-data. Columns below: whether the source contributed data to the charts, its span, and where to get it yourself.

Source	Used?	Span	What it provides
USFWS Mourning Dove Population Status reports	YES	1999–2025	The backbone: call-count indices, BBS trends, abundance estimates, banding tables, age ratios. I parsed 20 annual PDFs.
USFWS Migratory Bird Hunting Activity & Harvest reports	YES	2002–2024	HIP harvest, hunters, and days afield for all 41 dove-hunt states. 905 state-season rows extracted.
CDFW Game Take Hunter Survey reports	YES	1948–2023	The California ledger. 44 scanned reports OCR’d with Apple Vision; values cross-checked against each report’s printed year-over-year comparisons.
North American Breeding Bird Survey (USGS)	YES	1966–2024	Long-term and 10-year dove trend estimates by state and unit, as republished in the status reports.
USGS Bird Banding Laboratory	YES	2003–2024	Preseason dove bandings by state (via status-report tables). Recovery-level microdata available on request; not needed for this article.
FWS Parts Collection Survey (wing survey)	YES	2007–2024	Age ratios (recruitment) by state and unit, corrected for differential vulnerability.
Mourning Dove National Harvest Strategy (FWS, 2014)	CONTEXT	—	The decision framework that consumes these data; informs the sustainability discussion. No numbers charted from it.
eBird Status & Trends	NOT YET	2007–	Weekly relative-abundance surfaces — the most promising covariate source for any future daily model. Out of scope for a trends article.
CDFW managed dove field lists	NOT YET	annual	Locations of planted public dove fields. The seed of a future outcome ledger; no harvest results published with them.

Appendix BTerminology

Glossary

Management unit (EMU / CMU / WMU): The three blocks of states — Eastern, Central, Western — that dove regulations and population estimates are organized around. Doves mostly breed, migrate, and get harvested within their unit, so each is managed as its own population. California is in the Western unit.
Harvest Information Program (HIP): The federal-state registry of migratory bird hunters, fully national since 1999. Sampling from this registry produces the harvest, hunter, and days-afield estimates in Figs 02, 03, and 08.
Call-Count Survey (CCS): The 1966–2013 spring roadside survey: fixed routes, 20 stops, count every dove heard cooing. An index of relative abundance, retired in favor of the Breeding Bird Survey.
Breeding Bird Survey (BBS): The continent-wide survey of all breeding birds, running since 1966. Supplies the official dove trend estimates via a hierarchical model with credible intervals.
Age ratio: Juvenile doves per adult in the harvest, measured from mailed-in wings. A direct read on recruitment — how well the population reproduced that year.
Integrated population model (IPM): A statistical model that combines harvest estimates, band-recovery rates, and age ratios into a single estimate of absolute abundance. Source of the 337-million figure.
Game Take Hunter Survey (GTHS): California’s own hunter survey, run since 1948 — mail cards through 2010, an online panel since 2022. The source of the 77-year state series.
Climatology: The history-based guess: what you’d predict knowing only past averages. In the duck articles, every model had to beat climatology to count. The dove numbers in this article are the raw material for building exactly that baseline.
Compensatory mortality: Deaths that substitute for other deaths rather than adding to them. Doves die at high rates regardless of hunting; moderate harvest largely replaces natural mortality, which is why a 4% take doesn’t dent the trend.

Replication

Everything is rebuildable from scratch: raw PDFs → OCR/text extraction → parsers → tidy CSVs → this page. The pipeline is Python (pdfplumber, pypdfium2, Apple Vision OCR) plus D3 for rendering; no numbers were typed into this article by hand except where a provenance note says so. The dataset ships as seven tidy CSVs with full source annotations.