Skip to main content

EEE330 Final exam — study guide

Population Ecology · Autumn 2026

The final exam of EEE330 Population Ecology: what to know and be able to do, lecture by lecture, the formula sheet and a sample exam.

The exam tests whether you can think with the core of the course, not whether you remember everything we showed. This page lists that core, lecture by lecture.

How the exam works

  • When and where. Friday 18 December, 14:00–17:00, room 13J96.
  • What it looks like. Two case studies of species new to you (40 points each), and five short claims, of which you judge four (20 points). Each case has five parts of 10 points: the first, the life cycle and matrix, is compulsory, and you choose three of the other four. Only the first three you answer are marked.
  • Each part stands on its own. It gives you the numbers it needs, so a mistake in one part does not cost you the next. The questions are spread across all lectures.
  • Numbers are given. λ, sensitivities and simulation results are printed; you read, check and use them. Calculations are limited to what we did by hand in class. Bring a calculator.
  • Formula sheet (below) printed in the exam: know when to use each formula, not the formula itself.
  • What earns marks: reasoning. A life cycle that follows the biology, a prediction with its reason, the rate and mechanism behind a result, an assumption that could fail, a decision you can defend.

Not examined: the Further reading papers, R code, and the numbers of the course examples (know what the cranes or owls showed, not the values). The slides are enough; the practicals are practice.

Lecture by lecture

Each lecture ends with a synthesis slide with its five key points. Know lists them here in full. Be able to lists what you should be able to do with them.

L1 — What population ecology asks

Know: L1 sets out the questions the rest of the course answers; it has no synthesis slide of its own, and nothing from it is examined beyond what the later lectures cover.

Be able to: say through which rates a given threat or intervention would act.

L2 — Life histories and population structure

Know (the key points of L2):

  • A life history is a schedule of survival and reproduction across age.
  • Fitness depends on when offspring are produced, not just how many.
  • Limited resources force trade-offs among survival, growth and reproduction.
  • Trade-offs make individuals of the same age demographically different.
  • Life history determines which vital rates most influence population growth.
  • Population composition reflects past vital rates and shapes future dynamics.
  • The best state structure is the simplest one that keeps the differences that matter and can be estimated.

Be able to: compare two life schedules (which grows faster, and what else you would need to know); place a species on the fast–slow continuum from its rates; say which states a model of a described species needs.

L3 — From life cycle to matrix

Know (the key points of L3):

  • A life-cycle graph shows, for one time step, how each individual in each state contributes to each state at the next census.
  • When the census happens decides what each state means and what each reproduction arrow contains.
  • Every arrow must span exactly one time step.
  • States should be the variables that best predict fate, which need not be age.
  • The projection matrix is the life-cycle graph written as a table: each column is “from”, each row is “to”.
  • Matrix entries are combinations of lower-level vital rates such as survival, growth and fecundity.
  • States can be ages, sizes or breeding status; the graph should follow the biology that decides fate.

Be able to: draw the life-cycle graph of a described species and write its matrix with entries as products of vital rates, for a pre- or post-breeding census; say what each entry means.

L4 — Projection, growth and structure

Know (the key points of L4):

  • λ, r and the doubling time describe the same constant growth.
  • A matrix in place of λ and a vector in place of N show what happens inside the population.
  • With constant rates, growth and composition settle: λ and the stable structure, the dominant eigenvalue and its eigenvector.
  • Populations with the same size and rates can follow different paths before their structure settles: transient dynamics.
  • Reproductive value measures each state’s contribution to the future population; long-run size follows the total reproductive value at the start.
  • A count gives the population’s growth rate only when the rates are constant and the structure is stable.

Be able to: project one step by hand; convert between λ, r and doubling time; read λ, the stable structure and reproductive values from given output and explain them; explain why a population with λ = 1 can shrink for a while.

L5 — Which rates matter?

Know (the key points of L5):

  • Sensitivity: how much λ changes per small change in an arrow; large when many individuals pass through it (w) and it leads to a valuable state (v).
  • Elasticity: the proportional change in λ for a proportional change in an arrow; elasticities add up to 1, and arrows in the same loop share them.
  • Managers change vital rates, not matrix entries: the sensitivity of λ to a vital rate adds up its effects through every arrow it enters.
  • High leverage does not make a good target on its own: what managers can change, and by how much, decides too.
  • Sensitivity and elasticity ask which change would move λ; an LTRE asks which changes did: contribution = difference × sensitivity.
  • A rate with a low elasticity can still explain most of an observed change: it changed a lot, and λ is sensitive to it.

Be able to: interpret given sensitivity and elasticity matrices; compute a sensitivity by changing one rate, and with the chain rule for a rate in two entries; compute one LTRE contribution and check that the contributions add up; recommend a management target and defend it.

L6 — Variable environments and small populations

Know (the key points of L6):

  • Growth varies from year to year, though part of that is counting error.
  • Where we know the driver, growth follows the environment.
  • The long-run rate is the geometric mean: variation itself lowers growth.
  • The average future overstates what a typical population does.
  • Variation with no known driver we treat as environmental stochasticity.
  • The same variation costs most in the rates λ is most sensitive to.
  • Runs of bad years raise the risk, even with the same mean and variance.
  • A trend in the environment shifts the mean rates.
  • Environmental stochasticity hits all together; demographic stochasticity matters when N is small.
  • Risk depends on threshold, horizon and model.

Be able to: compute arithmetic and geometric means of a few yearly λ and explain the difference; say which rate’s variation would cost more, given sensitivities and variation; sort examples into environmental or demographic stochasticity; read a quasi-extinction curve.

L7 — Density dependence and regulation

Know (the key points of L7):

  • Density dependence: growth changes with the population’s own numbers, but a count series alone is weak evidence.
  • Regulation needs a feedback; limitation is whatever sets the level.
  • With density dependence the matrix changes every year; \(K\) is where its λ reaches 1.
  • \(K\) comes from the rates and the habitat. Below a threshold there is no equilibrium at all.
  • “Density” can be several numbers, each acting on different rates, and differently by season.
  • Strong or delayed feedback alone can produce cycles, and at low numbers the feedback can change sign: Allee effects.
  • Another species makes a feedback only if it responds to the focal population; its effect still enters through vital rates.

Be able to: say which rate of a described species is likely to depend on density and why; read growth against numbers (\(K\), \(r_{max}\)); explain what changes \(K\); explain why a negative slope of growth on counts is not yet proof of density dependence.

L8 — Harvest

Know (the key points of L8):

  • Harvest is a loss like any other; what matters is which rates it touches.
  • Harvest deaths can replace natural deaths or add to them, and compensation stops at the deaths that would have happened anyway.
  • Compensation is density dependence at work, through survival, breeding or immigrants.
  • A fixed quota can collapse a population; a fixed rate corrects itself.
  • Fast species compensate; in slow species adult harvest is mostly additive, and heavy losses can even slow their life down.
  • Who is removed matters as much as how many.
  • Males and social structure are not free.
  • A hunting bag needs effort to say anything about the population, and management can learn as it goes.

Be able to: predict whether harvest of a described species is likely to be compensatory or additive; find the equilibria of a quota and a rate on a recruitment curve and say which are stable; choose which class to harvest from elasticities; criticise a harvest plan.

L9 — Space and dispersal

Know (the key points of L9):

  • Space matters when one total hides differences in rates, movement, risk, or where to act.
  • Location is part of the state; dispersal is leaving, moving and settling, each a rate.
  • A sink persists only through immigrants; local growth cannot tell a source from a sink, the contribution \(C = R + E\) can.
  • A steady count or harvest can hide a drain on a source; a pseudo-sink only looks like a sink.
  • Sensitivity and LTRE transfer to space: a dispersal rate is a lower-level vital rate.
  • Size, quality, distance and shared weather shape connectivity; rescue and synchrony pull in opposite directions.
  • Choose the resolution the question and data support: occupancy may be enough.

Be able to: draw the life cycle and matrix of a species living in two places with dispersal; interpret given local and regional λ, elasticities and LTRE contributions for spatial rates; calculate a contribution \(C = R + E\) from given rates and say whether a place is a source or a sink; explain why a stable count or a steady harvest can hide a sink, and how a pseudo-sink differs from a true sink; find the entries a dispersal rate enters for the chain rule; say how a patch feature (size, quality, distance, a barrier) enters a rate; reason about rescue and synchrony; defend a choice between a stage-structured spatial model and patch occupancy. Not examined: dispersal kernels, covariance calculations, occupancy estimation, and the numbers of the examples.

L10 — Where vital rates come from

Added after the lecture.

L11 — Population viability and management

Added after the lecture.

The formula sheet

This sheet is printed in the exam.

  • Unstructured growth: \(N_{t+1} = \lambda N_t\); \(r = \ln\lambda\); doubling time \(= \ln 2 / r\).
  • Projection: \(\mathbf{n}_{t+1} = \mathbf{A}\,\mathbf{n}_t\); columns “from”, rows “to”.
  • Sensitivity of λ to entry \(a_{ij}\): \(\dfrac{\partial\lambda}{\partial a_{ij}} = \dfrac{v_i\,w_j}{\mathbf{v}\cdot\mathbf{w}}\), with \(\mathbf{w}\) the stable structure and \(\mathbf{v}\) the reproductive values.
  • Elasticity of λ to entry \(a_{ij}\): \(\dfrac{a_{ij}}{\lambda}\,\dfrac{\partial\lambda}{\partial a_{ij}}\); the elasticities of all entries add up to 1.
  • Chain rule for a vital rate \(\theta\): \(\dfrac{\partial\lambda}{\partial\theta} = \sum_{ij} \dfrac{\partial\lambda}{\partial a_{ij}}\,\dfrac{\partial a_{ij}}{\partial\theta}\).
  • Small change: \(\Delta\lambda \approx \dfrac{\partial\lambda}{\partial\theta}\,\Delta\theta\).
  • LTRE: contribution of \(\theta\) \(\approx (\theta^{(2)} - \theta^{(1)})\,\dfrac{\partial\lambda}{\partial\theta}\), with the sensitivity midway between the two matrices.
  • Geometric mean of \(\lambda_1, \dots, \lambda_T\): \(\left(\prod_t \lambda_t\right)^{1/T}\), or \(\exp(\text{mean of } \ln\lambda_t)\).
  • Cost of variation in a rate \(\theta\) with standard deviation \(\sigma\) among years: \(\lambda - \lambda_s \approx \dfrac{\sigma^2}{2\lambda}\left(\dfrac{\partial\lambda}{\partial\theta}\right)^2\).
  • Logistic growth: \(r(N) = r_{max}(1 - N/K)\); added per year \(R(N) = r N (1 - N/K)\), largest (\(rK/4\)) at \(N = K/2\).
  • Demographic stochasticity: the year-to-year swing in growth from individual luck shrinks roughly as \(1/\sqrt{N}\).

Practising

The sample exam below shows the style; its answers are folded under each question. The in-class questions of each lecture, and their answers at the end of each deck, are the best practice: the exam asks the same kinds of question about new species.

Sample exam

This sample shows the style of the final exam. The real exam has different species and different questions in the same style. The species are real; the numbers are invented for the exam. Each case has five parts of 10 points: answer the first and three of the other four; only the first three you answer are marked. Answer concisely, show your reasoning, and define any symbol you introduce. The formula sheet of the study guide is printed with the exam.

Short answers that make the same points earn full marks. The answers are folded under each question. ## Case 1 — Rock ptarmigan in the Alps (40 points)

The rock ptarmigan is a grouse of the Alpine zone above the tree line. Females breed from their first summer. Chicks hatch in July; many die in their first weeks and over their first winter. Adults have the same annual survival whatever their age. A long-term study counts females each spring, before breeding. Yearling females (one year old) raise 2 female chicks each, older females 3. Of these chicks, 20% survive to the next spring count. Yearlings and adults each survive the year with probability 0.5.

Answer 1a and three of 1b–1e (10 points each).

1a (10 points, compulsory). Draw the life-cycle graph with two states, yearling and adult females, and write the projection matrix, each entry first as a product of vital rates (define your symbols), then as a number. Which census does your model assume, and what does that mean for the reproduction arrows?

Yearling Adult sY = 0.5 fA s0 = 0.6 sA = 0.5 fY s0 = 0.4

Rates: \(f_Y\) = 2, \(f_A\) = 3, chick survival \(s_0\) = 0.2, \(s_Y = s_A\) = 0.5. Arrows: Y → Y \(f_Y s_0\) = 0.4; A → Y \(f_A s_0\) = 0.6; Y → A \(s_Y\) = 0.5; A → A \(s_A\) = 0.5.

\[\mathbf{A} = \begin{bmatrix} 0.4 & 0.6 \\ 0.5 & 0.5 \end{bmatrix}\]

Pre-breeding census: the reproduction arrows include the chicks’ survival to their first count, \(s_0\).

1b (10 points). In one spring the study area holds 200 yearling and 50 adult females. Project one year ahead. The dominant eigenvalue is λ = 1.0 and the stable structure is 50% yearlings, 50% adults. Why does your projection still show a decline?

Yearlings 0.4 × 200 + 0.6 × 50 = 110; adults 0.5 × 200 + 0.5 × 50 = 125; total 235 instead of 250. λ = 1 holds only at the stable structure. This population has mostly yearlings, which produce fewer young than adults, so it first shrinks while its structure settles (transient dynamics).

1c (10 points). The sensitivities of λ to the four entries (left) and their elasticities (right) are:

\[\begin{bmatrix} 0.45 & 0.45 \\ 0.55 & 0.55 \end{bmatrix} \qquad \begin{bmatrix} 0.18 & 0.27 \\ 0.27 & 0.27 \end{bmatrix}\]

  1. Chick survival \(s_0\) enters two entries. Compute the sensitivity of λ to \(s_0\) with the chain rule.
  2. Predator control could raise chick survival from 0.20 to 0.25; a hunting ban could raise adult survival from 0.50 to 0.55. Estimate the change in λ for each.
  3. Which would you recommend, and what else would you want to know?
  1. \(\partial\lambda/\partial s_0 = 0.45 \times 2 + 0.45 \times 3 \approx 2.27\).
  2. Chick survival: \(2.27 \times 0.05 \approx 0.11\). Adult survival: \(0.55 \times 0.05 \approx 0.03\).
  3. Predator control moves λ about four times as much, but cost, feasibility, and whether chick deaths are compensated (density dependence) also decide.

1d (10 points). Chick survival varies a lot from year to year with the weather in July (standard deviation 0.10); adult survival varies little (0.05). Using the formula sheet, which source of variation lowers long-run growth more? Does this species fit the idea that the rates with most leverage vary least?

Chick survival: \(0.10^2 \times 2.27^2 / 2 \approx 0.026\); adult survival: \(0.05^2 \times 0.55^2 / 2 \approx 0.0004\). Chick survival costs far more. It does not fit buffering: here the rate with most leverage is the variable one, which is typical of fast species, whose early survival depends on weather.

1e (10 points). Hunting takes place in autumn, after the chicks have fledged. A hunter argues: “Most young birds we shoot would have died in winter anyway, so hunting does not reduce the population.” When is he right, and what would you measure to test it?

He is right if winter mortality is density dependent, so that survivors live better when hunting removes birds (compensatory mortality), and only up to the deaths that would have happened anyway. Test: compare winter survival of marked birds in years or areas with different hunting pressure.

Case 2 — Two colonies of Alpine ibex (40 points)

Alpine ibex were reintroduced to the Swiss Alps from a few founders. Two colonies are now monitored: one still at low density, one that has filled its range. Females give birth from three years of age; the model simplifies this to two states counted before the birth season: yearling females and adult females. Each adult female raises \(f\) female kids, which survive their first year with probability \(s_K\); yearlings survive with \(s_Y\), adults with \(s_A\).

Rate Low-density colony High-density colony
\(f\), female kids per adult female 0.45 0.40
\(s_K\), kid survival 0.70 0.45
\(s_Y\), yearling survival 0.90 0.85
\(s_A\), adult female survival 0.95 0.94
λ 1.19 1.08

Answer 2a and three of 2b–2e (10 points each).

2a (10 points, compulsory). Draw the life cycle and write the matrix for the low-density colony, with symbols and with numbers. Why is there no arrow from yearlings back to yearlings?

Yearling Adult sY = 0.90 f sK = 0.315 sA = 0.95

\[\mathbf{A} = \begin{bmatrix} 0 & f s_K \\ s_Y & s_A \end{bmatrix} = \begin{bmatrix} 0 & 0.315 \\ 0.90 & 0.95 \end{bmatrix}\]

Yearlings do not breed; adults produce next year’s yearlings through \(f s_K\).

2b (10 points). An LTRE of the high-density colony against the low-density one gives these contributions to the difference in λ: \(f\) −0.019, \(s_K\) −0.070, \(s_Y\) −0.009, \(s_A\) −0.009. In the low-density colony, the elasticity of adult survival is 0.67 and that of kid survival 0.17. Check the LTRE, and explain why adult survival explains so little of the difference despite its higher elasticity.

The contributions sum to −0.107; the observed difference is −0.11. A contribution is difference × sensitivity: adult survival barely differs (0.95 against 0.94), while kid survival fell from 0.70 to 0.45. Elasticity says what would matter; the LTRE says what did change.

2c (10 points). A ranger fits a regression of yearly growth on the colony’s counts over 20 years and finds a negative slope. She concludes that the colony is regulated by density dependence. Why is that not enough, and how could you test it?

Counts that are high by chance tend to be followed by lower ones, and counting errors add to this, so the regression finds negative slopes even without density dependence. Test: simulate random-walk populations with the colony’s mean and variation of growth and see how often their slope is as steep; or compare kid survival in the same colony across years of different density.

2d (10 points). Kid survival comes from counts of females with a kid in July and of yearlings the next spring. Name two ways these counts could give a wrong estimate of \(s_K\), and what data would avoid them.

Detection below 1: kids or yearlings are missed, so survival is confused with detection. Emigration: a yearling that left counts as dead (apparent survival). Marked individuals resighted repeatedly separate survival from detection (capture–recapture, EEE323).

2e (10 points). To slow the high-density colony, the canton allows 10 females a year to be shot. In that colony, the reproductive value of an adult female is 1.27 times that of a yearling. Which class should hunters take to slow growth most, and why might the colony fall less than the cull suggests?

Adult females: 10 adults remove about 12.7 yearling-equivalents of future population, 10 yearlings remove 10; adults also carry the highest elasticity. The colony may fall less because kid survival is density dependent (2b): fewer animals means better kid survival, which partly replaces the animals shot.

Question 3 — Judge the claims (20 points)

Choose four of the five claims (5 points each). For each, say whether it holds, and why, in a few sentences. Name the rate, the mechanism or the assumption that decides it. Only your first four answers are marked.

3a. A hunting authority estimates the maximum sustainable yield of a duck population at 4,000 birds a year from a logistic model of recruitment, and sets a fixed quota of 4,000 birds every year: “Since the quota equals the MSY, the harvest is sustainable.” Use a sketch of recruitment against population size.

It does not hold. At MSY the quota line just touches the top of the recruitment curve: one equilibrium at \(K/2\), with no margin. A run of bad years or an overestimated curve pushes the population below \(K/2\), where the fixed quota takes more than the population adds, and it declines ever faster. Better: a fixed harvest rate set below the rate at MSY, from annual counts, which takes fewer birds when numbers are low; with a threshold below which no harvest is taken.

Sketch: hump-shaped recruitment against N; the quota as a horizontal line touching the top; a rate as a line through the origin crossing the curve once, on its right side.

3b. “In this wetland the local growth rate of the reed warblers is below 1. It is a sink, so draining it will not harm the regional population.”

Not necessarily. Local growth counts only the births and deaths of the wetland’s own birds; a place can have local λ < 1 and still export many young that breed elsewhere. Whether it is a source or a sink depends on its contribution, \(C = R + E\): the birds it keeps (\(R\)) plus the ones it sends out (\(E\)) per bird. It may also be a pseudo-sink: crowded by immigrants, its local growth looks low only because of density. And even a true sink holds birds that buffer the regional population in bad years elsewhere.

3c. “Our PVA of the capercaillie gives a 2% risk that the population falls below 50 females within 20 years. The population is safe and needs no action.”

Too strong. The risk depends on the threshold, the horizon and the model: over 50 years, or with a threshold of 100 females, it could be much higher. A PVA is only as good as its rates and its assumptions: uncertainty in the estimates, a missing trend in the environment, correlated bad years or density dependence can each change it. A low risk is a reason to keep monitoring and to compare scenarios, not to stop looking.

3d. “Over 10 years, the arithmetic mean of the yearly λ of a marmot colony was 1.02. The colony will grow, and in 50 years a typical colony like it will be larger.”

It does not hold. Growth multiplies: the long-run rate is the geometric mean of the yearly λ, which is lower than the arithmetic mean whenever λ varies, and can be below 1 while the arithmetic mean is above 1. The average of many futures is pulled up by a few colonies that do very well; the typical (median) colony follows the geometric mean. Some of the variation may also be counting error rather than real change.

3e. “Building a corridor between two populations of a lizard always lowers their risk of extinction.”

Not always. Connection lets immigrants rescue a population after a crash, and that lowers risk. But if both populations experience the same weather, they crash in the same years (synchrony), and there is no one left to rescue the other. A corridor can also spread disease or predators, or lead dispersers into poor habitat where they die. It helps most when the two populations fluctuate independently.