Part 1
The Debated Rise and Fall of the Swedish Moose Population
A Remarkable Story
If reported moose harvest numbers reflect proportions of true population sizes, then the past 85 years of moose hunting form a story worth telling. The footprint of moose harvests in Sweden from 1939 to 2023 is nothing short of spectacular. These are not just numbers — they rather show the aftermath of human activities and the response of moose populations. The way moose densities will change is based on well-established ecological foundations, confirmed by applied research. Alterations in the forest ecosystem are mainly triggered by forestry, wildlife management, and policy makers.
This text aims to clarify how ecological processes — shaped by both biotic and abiotic factors — may have influenced moose population dynamics over time — and ultimately bring solutions to current debate about the peril of the Swedish moose population. The graph below (Figure 1) marks the starting point of our journey to understand the history of moose harvests in Sweden from 1939 to 2023.
Figure 1.1 Moose harvest in Sweden from 1939–2023, showing the early low-density period, the rapid expansion phase leading to the early 1980s peak, and the long-term decline that followed.
Historical Notes
The history of low moose densities can be traced back to King Gustav III’s 1789 decree, which granted landowners the right to hunt. This, combined with what was likely a more widespread presence of large carnivores at the time, appears to have kept moose numbers low for generations. From the late 1930s to the mid-1940s, the population remained sparse — with an average of just 0.28 moose harvested per 1,000 hectares. In practice, this meant that only one moose was culled every third or fourth year.
The Swedish Hunters’ Association, founded in 1830, initiated efforts to restore the moose population. Their efforts appear to have had an effect, as harvests increased steadily from the 1940s and leveled off around 1960. This partial trend may reflect the carrying capacity of the landscape at the time. By then, carnivore densities — primarily gray wolf and brown bear — were low on a national scale. According to official records, the wolf was declared extinct in 1966, and the brown bear population was estimated at about 300 to 400 individuals.
The mid-1960s marked a milestone in forestry, as traditional
single-tree felling was replaced by clear-cutting — a shift that
drastically altered the ambient moose habitat. This industrial approach
opened the door to rapid re-establishment of suitable forage across vast
areas of deciduous trees and shrubs. Combined with the hunting
regulations introduced by the 1938 Hunting Act, it set the stage for an
unprecedented growth in the moose population (see the logistic section
in Fig. 1.1).
Some Notes on Population Biology
The footprint of moose harvest densities can be divided into three phases. The first is the low-density phase, where populations are notably sparse and sensitive to random events. The second is the logistic growth phase, marked by consistent population increase. A typical logistic growth has two parts: the first is an accelerating, exponential growth up to a peak called the point of inflection. At this point, a population reaches its highest possible growth rate where there are no constraints posed on the individuals forming a population. The intrinsic survival and reproductive capacity of the individuals are fully supported - a condition that live-stock farmers seek to mimic. After this point of inflection, the growth rate gradually slows down due to density-dependent effects caused by intensified competition over dwindling resources. When a population reaches the limit that doesn’t support further growth is known as the carrying capacity (K), the growth will end, meaning that the net of born and deceased individuals is zero, as the forest ecosystem no longer maintains forage supply that promote further growth.
As we already have invoked some important aspects in population
ecology, it’s time to add some more, in order to explain the well
documented different stages in population dynamics. The mathematics to
calculate population growth is as simple as it gets - given you have
access to the numbers:
Figure 1.2 gives us the general picture of the distinct phases. However, note that λ depicted on the y-axis is transformed to z-scores. This transformation doesn’t alter the factual relation between the λ-values, it just transforms where values of λ = 1 now indicates zero growth - or no change.
Figure 1.2 The graph shows three distinct phases of population dynamics, each marked by double arrows indicating their duration, along with the mean λ ± standard deviation. The stochastic phase (1939–1965) shows the greatest variation in λ-values, with a coefficient of variation (CV) of 2.44 as expected when population sizes are low and sensitive to random effects. In contrast, the logistic growth phase (1966–1982) has a CV of 1.1. This means that the stochastic phase had more than twice the relative variation in growth rates compared to the logistic phase. Comparing the stochastic and post-logistic phases, the CV is again nearly twice as high. Also note the magnitude of the mean λ-values: the negative mean in the post-logistic phase clearly reflects the observed population decline.
The Capricious Nature of Sparseness
The nature of limited numbers
Figure 1.3 shows the stochastic phase, where pure effects of sparse densities in the late 1930s successively transforms to more stable populations in the early 1960s. Without doubt an effect of the moose management incentive to restore the Swedish moose population, launched by Swedish Hunter’s Association. The increase in harvest densities was about four-fold from 1939 to 1965. The stability of the λ-values confirms that moose populations were at safer ground, below prevalent carrying capacity showing smooth density footprints.
Figure 1.3 The graph shows, from top to bottom, the harvest densities and the rate of change (λ) covering the timeline from 1939 to 1965. The top panel plots harvest densities at extremely low population densities. Populations at such low crowding are extremely sensitive to random outcomes. For example, if a reproductive cow disappears it can jeopardize population recovery. However, the Swedish moose population seemed to be on safer grounds from 1945 onward. If the culling rate 1939 was 20% with a harvest density at 0.2207 the actual population size would have been ≈ 1.1 / 1,000 ha. Corresponding population density at a harvest density of 0.8304 per 1,000 ha (1961) would have been ≈ 4.2 moose per 1,000 ha. The bottom panel shows that growth rates successively became more stable. In fact, the early management initiative launched by Svenska Jägareförbundet appeared successful. Notably, the top panel suggests a logistic shape of the population harvest densities. This raises the question of whether the period from 1958 to 1965 reflects a moose population at or near its carrying capacity.
Always Stay on the Bright Side of Life
The nature of unlimited growth
Figure 1.4 illustrates the logistic growth phase lasting from 1966 to 1982 — a period when the national moose population responded to unlimited access to forage and mating opportunities. The hallmark of this phase was unbroken population growth. This explosion was directly caused by the shift in forestry practices to large-scale clear-cutting, which exposed vast areas of land. Within just 1.5 to 3 years, these areas produced ideal and easily accessible browsing conditions for moose, fueling rapid expansion. The early stages of the logistic expansion passed largely unnoticed, even though aerial surveys conducted in 1969, 1970, and 1972 indicated increasing moose numbers — despite elevated harvest levels and improved reproductive rates. Still, no regular monitoring of population size existed, unlike the routines established in today’s management. The monumental ecological response to this new forestry regime was practically unrecognized during the exponential stage of the logistic phase. Aside from anecdotal reports, no objective tools or methods were in place to track the population boom — therefore, no timely response from managers emerged, while hunters remained content with abundant culling opportunities.
Figure 1.4 The two panels summarize key aspects of the logistic growth phase of the Swedish moose population from 1966 to 1982. The top panel shows how moose harvest densities increased almost sixfold during this period, reaching 4.42 culled animals per 1,000 hectares by 1982. Estimated population densities under different harvest quotas (15–30%) are included to provide information about possible outcomes. Culling rates used to be 30% by tradition. Likely, the actual moose population density was underestimated, if assessed at all. Therefore, if management was based on a culling rate of 30%, the true culling rate may only have been somewhere between 15% and 20%. The bottom panel presents the rate of change (λ), which remained consistently above 1, indicating sustained population growth. The geometric mean of λ was 1.116, with a coefficient of variation of 7.4%, reflecting relatively stable growth. The odds ratio (0/16) shows that there were no observed years with declining population size. The bottom panel illustrates the distribution of λ values, which is skewed toward higher values. The observed geometric mean (black dashed line) exceeds the expected mean for a stable population (dashed line), further supporting the conclusion that this phase was characterized by strong and consistent growth.
If we take a closer look at the logistic phase, it seems that the national moose population never reached the carrying capacity
(see figure 1 & the left panel in figure 1.4). The population did
pass the point of inflection between 1979 and 1980 and reached the
maximum number of culled moose in 1982. By modeling the population
development, using a logistic function:
where N(t) is population or harvest density at time t, K is the carrying capacity, A is a constant based on initial culling numbers. Euler’s number e ≈ 2.71828, r is the estimated parameter for the intrinsic growth rate. This parameter can easily be transformed to λ by exponentiating Euler’s number to the power of r-parameter. Finally t marks the time steps from t to t+1 over the range of culled moose densities. The result of the function is shown in Figur 1.5
Figure 1.5 The graph illustrates the logistic growth phase of moose harvest densities in Sweden from 1966 to 1990. The black curve represents the fitted logistic model based on harvest data from 1966 to 1982, extrapolated beyond 1982 (grey area) to visualize the expected trajectory had growth continued unimpeded. The asymptotic line at K = 5.59 harvests per 1,000 hectares indicates the estimated carrying capacity. The blue symbol marks the point of inflection — where the intrinsic rate of change reaches its maximum (r = 0.465). When converted to the maximum finite rate of change, this corresponds to λ = 1.592, reflecting a very rapid population growth. The graph shows that the harvest density at this inflection point was 3.204 moose per 1,000 hectares (horizontal dashed blue line), observed during the 1979 hunting season (vertical dotted blue line). Model statistics indicate a trustworthy fit (R² = 0.997), with clearly significant estimates for the carrying capacity (K), the parameter A (describing the starting distance from K), and the intrinsic growth rate (r). The graph also highlights the abrupt deviation from the predicted growth curve after 1982. What caused the interruption of the predicted logistic growth — a turning point that pushed harvest densities back towards the point of inflection?
A valid question is why the logistic phase was broken. As historical records are hard to find, the reasoning here is based on testimonies from professionals active at the time. They claimed increased moose browsing on Scots pine seedlings replanted on clear-cut areas as the reason to reduce populations. Since then, the forestry sector has routinely monitored browsing damage. This was likely the factor that ended the euphoria of bountiful hunting — at least from the perspective of forestry.
Also note that populations were approximately kept at the inflection point, where λ had its maximum. Was this an intentional harvesting strategy, or just a lucky strike? Population biologists promoted the idea of maintaining densities where populations are most productive. The reasoning is straightforward: when a population enters densities above the inflection point, forage conditions are no longer unlimited in quality or quantity. Contested forage leads to reduced reproductive output. Thus, pushing a population back to density levels where λ is at its maximum — by culling surplus individuals — would, in theory, provide the highest yield.
From a hunting perspective, this was a favorable solution that granted a maximum trade-off according to the prevailing carrying capacity. In my opinion, however, this is too meticulous a path to tread. To succeed with such an intention, you not only need accurate information about population density, but also detailed data on age structure and sex ratios. Gathering such information is labor-intensive and costly. Even with access to population details, you’re still not in the clear. As mentioned earlier — how do you account for random effects that cause yearly variation in carrying capacity?
My take is to keep moose populations below the point of inflection — staying on the exponential side of the logistic curve. This will reduce the risk of overbrowsing by promoting recovery of the carrying capacity of forest vegetation. The field layer is especially important for calves, given their limited browsing height compared to full-grown moose. While ecologically sound, this approach may be less appealing to hunters who prefer high short-term yields. Still, a thriving moose population, improved vegetation diversity that recovers each season, and reduced losses in timber yield and quality due to browsing damage form a win-win outcome.
This leads us to the speculative question what would have happened to
a forest habitat if the moose population growth actually proceeded to
the limit set by the carrying capacity. By that question we are set to
dissect the dynamics of populations at resource equilibrium
(fig. 6).
Walk the Hobble of Equilibrium
The Nature of limited growth
Figure 6 illustrates a striking national decline in harvest densities. Between 1983 and 2023, harvest levels dropped from about 4.3 to 1.3 moose per 1,000 hectares. The odds ratio for observing λ-values < 1 compared to > 1 is 26 to 15, and the geometric mean λ equals 0.97 — indicating a slow but persistent downward trend in population growth rates.
Figure 1.6 The panels summarize the long-term dynamics of a moose population under observed culling pressure. The top panel shows annual harvest density per 1,000 hectares from 1983 to 2023, showing a gradual decline in cull intensity from 4.3 to 1.3 moose removed per 1,000 hectares for the hunting season 2023/2024. The table in the top panel shows what population density would be required for different cull rates (10 to 30 percent) to yield the observed long-term average harvest of 1.264 moose per 1,000 hectares.The bottom panel shows the corresponding annual rate of population change (λ). The green dots show growing harvest densities (above the gray line), and blue dots show declining harvest densities (below the gray line). The dashed line marks the geometric mean, indicating a general decline in the rate of change. With an upper confidence limit (95%) that is less than 1, and an odds ratio indicating a likelihood of almost 2:1 in favor of λ < 1, this strongly confirms a general national decline in the Swedish moose population since 1983.
Figure 7 amplifies this pattern. A polynomial model indicates a steep decline in harvest densities from 1983 to 1994, followed by a period of relative stability, and then another downturn beginning around 2015. The interpretation is not straightforward. In practice, management often responds to perceived population declines by reducing harvest quotas. This reaction may cause a drop in observed harvest densities that reflects hunting behavior more than actual population dynamics.
Figure 1.7 The graph shows the long-term decline in harvest density per 1,000 hectares during the equilibrium density phase and compares two alternative polynomial models fitted to the annual harvest series. A second-degree model captures the broad downward trend, while a third-degree model provides a significantly better fit, reflected in lower residual sum of squares and a clear improvement in AIC. Statistical testing confirms that the additional curvature in the third-degree model explains substantial extra variation. The point where both fitted curves converge around 2003 marks the midpoint of the declining harvest trajectory, after which harvest levels continue to fall toward the present.
Still, the oscillations around the fitted curve are too wide and
too patterned to be dismissed as noise or broken age-structures alone. A
more plausible explanation is that carrying capacity has gradually
degraded over time. Under such conditions, random effects — like annual
variation in forage quality — become more influential. This could lead
to feedback loops and delayed reactions: managers lower the harvest,
allowing a modest population recovery, which then triggers renewed
harvesting, and the cycle repeats. These lagged effects may be a key
reason why the system oscillates around a slowly declining baseline.
Altogether, these figures suggest a population walking the edge of
equilibrium — one where carrying capacity is unpredictable and
management actions are always one step behind. <
Make every bite count
The effect of poor nourishment
The quota of calves per cow gives us a fast assessment of how close a moose population is to actual carrying capacity. If a population is consuming less than what its habitat offers - then cow reproductive capability will not be limited. However, if the habitat cannot supply a sufficient amount of nourishment, the reproductive capacity will become lower. Figure 1.8 shows a general declining trend in number of calves per cow. Since 2000, the national average calf ratio has been reduced from ≈ 0.78 to ≈ 0.59 calves per cow in 2023. This decline is statistically significant. Anecdotes from the era when the moose population was at maximum growth rate (1978), state that cows having twin calves were substantially more common than cows with a single calf.
Figure 1.8 The graph shows a negative trend of the average ratio and confidence intervals of the ratio between calves and cows by year. The regression coefficient β ≈ -0.01, p < 0.001 implies that the negative trend is not a random outcome. The effect of year explains 87.1% of the variation in calves per cow ratio (F = 148.4 on 1 and 22 degrees of freedom, p < 0.001). The variation in residuals along the regression (black line) suggests variation in the quality of the vegetation season. The dashed line indicates a possible accelerating decline towards smaller ratios, however, the non-linear parameter is not statistically reliable. Time will tell.
Moose reproduction has three outcomes - either 0, 1 or 2 calves
per reproductive cow. For the sake of correctness - cows having 3 calves
are possible but a very unlikely event (< 0.03% ~ 3 in 10,000 cows)
and are therefore left out in this analysis. Figure 1.9 shows the
average proportions of cows having 0, 1, or 2 calves by year from 2000
to 2023. The figure shows that the proportion of cows without calves is
increasing (gray line and marks) while cows having two calves are
declining (dark green lines and marks). The proportion of cows having
one calf seems to be quite stable across the timeline (light green line
and marks). Note the pattern of ruggedness in the trends. When the
proportion of cows without calves increases - the proportion of cows
with two calves declines. Meanwhile, the proportions of single-calf cows
are pretty stable, at least until 2013. One possible explanation to
this phenomenon might be that a typical twin-producing cow becomes a
single-calf cow, while a single-calf cow becomes a cow without calves.
That is, if twin producing cows lose one (a negative net outcome),
single-calf cows gain one and lose one cow to the category of cows
without calves (a zero net outcome), subsequently resulting in a net
gain for the category of cows without calves.
After 2013, the single calf category trend seems to have become more volatile and shows a slight decline in its proportions. Is that the suggested accelerating decline towards smaller ratios that the non-linear parameter in Figure 1.8 tried to tell us?
The main takeaway from Figure 1.9, however, is that cows without calves dominate by making up more than 50% of the cow population since 2018. The twin-producing cow population has been linearly reduced since 2000, from 0.18 to 0.08 proportion-wise. The one-calf cow contingent has practically not changed across the timeline, but keep an eye on any significant changes that seem to be going on since 2020.
Figure 1.9 The graph shows average national proportions of reproductive status of moose cows from 2000 to 2023. The plotted categories reveal a statistically significant general increase in proportions (β0 ≈ 0.004, p < 0.001, R² = 0.845) for cows without calves (gray lines and markers) and a significant general decrease in proportions (β2 ≈ -0.004, p < 0.001, R² = 0.889) of twin-calf cows (dark green line and markers). The proportion of single-calf cows does not show any effect of year (β1 ≈ -0.000, p = 0.44, R² = 0.027). As proportions sum up to 1, any change in one proportion affects the other proportions. The correlation between the proportions of cows without calves and twin-calf cows is very strong (r = -0.971, p < 0.001) while the proportions of single-calf cows are uncorrelated (0 vs. 1: r = -0.05, p = 0.82 and 2 vs. 1: r = 0.19, p = 0.37). This result implies that the results are not just mathematical as the proportions of single-calf cows remain quite stable in comparison. Thus, any reduction in proportions of twin-calf cows causes an increase in the proportion of barren cows, directly or indirectly. This is an ecological signal that forage and nourishment conditions generally affect the reproductive capacity negatively among moose populations.
There is a prevailing belief that the poor reproductive status
is a result of impaired management, often referred to as “they’ve shot
that herd all outta shape” and claiming that foraging conditions have
nothing to do with the population decline. If these statements were
true, would you expect to observe a consistent decline in body mass
among moose calves? Probably not - there are no valid ecological reasons
for the observed stage-wise decline in female calf dressed body mass to
be an effect of population structure, nor for a general decline in
dressed body mass if foraging conditions were good (Figure 1.10).
A third-degree polynomial fits the general trend that shows three distinct phases (F = 34.43 on 3 and 20 degrees of freedom, p < 0.001, R² = 0.838). These phases show a rapid decline in dressed body mass followed by a stable but rugged stage, and finally another decline in median body mass. The median parameter targets the most common values, meaning the most probable observations. The phases of declining median body mass are statistically significant, suggesting that these are not a random outcome. In the early decline, the average loss was 1.45 kg per year and 0.87 kg per year in the later phase.
These numbers should be a reminder of what happened in the municipality of Mark where the so-called “Älvsborgssjukan” ran rampant in the late 1990s. In fact, the disease turned out to be a severe forage shortage. At that point, dressed body mass was as low as 35 kg in some cases and the reproductive output was one calf per ten cows. Obviously, body mass reflects nutritional status, which directly affects ovulation and calf survival in moose. Some management units agreed to reduce the existing moose population drastically in order to restore forage abundance. This action led to a recovery in reproductive output and body mass. Let us hope that these signals lead to proactive management measures. Do not make the same mistake twice!
Figure 1.10 The plot shows the trend in median female calf dressed body mass from 2000 to 2023. The general trend is a fall followed by a plateau and, then a decline. The general trend was fitted by a third-degree polynomial (RSE = 1.48 on 20 degrees of freedom, R² = 0.838 (83.8%), F = 34.43 on 3 and 20 degrees of freedom, p < 0.001). The partial linear regressions are testing whether the declines are random or factual outcomes. The early decline (RSE = 0.886, R² = 0.937, βe = -1.45, p < 0.001) and the late decline (RSE = 1.79, R² = 0.570, βl = -0.870, p < 0.05) are statistically reliable even though the latter decline is based on n = 7. Note the difference in the ruggedness along the general trend where the plateau shows more variation in comparison to the tails. Read more about this in Part 2.
Summary
Taken together, these results point in a consistent direction. Reproductive output has declined, body mass has declined, and the patterns are too structured to be explained by chance alone. Viewed in a broader context, the historical development of the Swedish moose population closely follows well-established principles of population dynamics, moving from low-density stochastic effects through logistic growth and into a phase constrained by resource limitation. While impaired management may contribute, it does not provide a sufficient explanation for the observed trends. Instead, the signals align with a system where forage availability is no longer adequate to sustain previous levels of reproduction and growth. If so, the current trajectory is not merely a consequence of how moose are harvested — but of what the landscape can no longer provide.