Module 8: Climate, Conservation & Genetics

The giraffe has lost an estimated 40 % of its continental population in three decades, driven by habitat loss, poaching, and regional civil conflict. IUCN listed Giraffa camelopardalis as Vulnerable in 2016, the same year Fennessy and colleagues’ phylogenomic study revealed that “the giraffe” is in fact four distinct species, shifting the taxonomy and demanding a species-specific conservation framework. This module closes the course by synthesising field demography, poaching economics, climate projections (CMIP6), translocation programmes by the Giraffe Conservation Foundation, and the genetic structure that the Fennessy split makes actionable. Two simulations quantify (1) four-species metapopulation viability under translocation rescue scenarios and (2) genetic-distance matrices and UPGMA phylogeny from representative SNP panels spanning the Fennessy 4-species + subspecies framework.

1. The 40 % Decline, 1985–2015

Between 1985 and 2015, continental giraffe populations declined from ~155,000 to ~97,500, a loss of ~37 % that led the IUCN to reclassify Giraffa camelopardalis from Least Concern to Vulnerable in 2016. Updated 2023 estimates, benefiting from improved survey methodology and coordinated counting by the Giraffe Conservation Foundation, place the total at ~117,000—partly a real recovery, partly a revision upward from improved detection.

Drivers of decline

Habitat loss: Sahel desertification, conversion of savanna to cropland (Ethiopia, Kenya, Tanzania), and infrastructure expansion.
Poaching: bush meat, hide, tail hair for traditional bracelets (Central Africa), and bone for carved curios in West Africa.
Civil conflict: South Sudan (Kordofan, Nubian), Somalia (reticulated), Chad (Kordofan), and Central African Republic have all experienced documented giraffe population collapses associated with armed conflict.
Small-population effects: inbreeding, demographic stochasticity, and Allee dynamics once local populations drop below ~100.
Climate change: CMIP6 projections show East African warming at >2× the global mean rate and Sahel wet-bulb temperature risk approaching 35 °C by 2080.

Subspecies-level catastrophes

The three most endangered taxa are all within the northern giraffe (G. camelopardalis):

Nubian (G. c. camelopardalis): <500 individuals, South Sudan and western Ethiopia. Genetically most distinct; target of priority rescue breeding.
Kordofan (G. c. antiquorum): ~2,000 individuals, central African rangelands (Chad, Cameroon, Central African Republic). Subject of the 2018 GCF Zakouma translocation.
West African (G. c. peralta): ~600 individuals, confined to Niger’s Kouré region. Intensive management has increased numbers from ~50 in the 1990s.

2. Fennessy 2016: Four Species, Not One

Julian Fennessy and colleagues, working with the Giraffe Conservation Foundation, assembled a continent-wide sample of 190 giraffes spanning all recognised subspecies. Sequencing seven nuclear loci, analysing mitochondrial DNA, and applying multiple phylogenetic and population-genetic methods (Current Biology, 2016), they concluded:

Genetic divergence between some putative subspecies exceeds that between many recognised vertebrate species.
The data support four reproductively isolated clades with no contemporary gene flow.
Nubian giraffe is genetically indistinguishable from Rothschild’s; both belong to the northern species.
Reticulated giraffe is a distinct species, not a subspecies of northern.
Southern and Masai giraffes are genetically distinct species, not sub-groupings of one.

The four species

Northern giraffe (G. camelopardalis): Nubian, Kordofan, West African subspecies. ~5,800 individuals.
Reticulated giraffe (G. reticulata): north-east Kenya, southern Ethiopia, Somalia. ~16,000 individuals.
Masai giraffe (G. tippelskirchi): East Africa, two subspecies. ~45,000 individuals.
Southern giraffe (G. giraffa): two subspecies (Angolan, South African). ~60,000 individuals.

Implications

The four-species split is not merely a taxonomic curiosity. It changes:

Conservation status assessment: northern giraffe alone is Critically Endangered; IUCN single-species Vulnerable masked this.
Translocation protocols: captive SSPs must not mix populations; translocations within Africa must match species.
Genetic rescue potential: limited—no close “outbreeding” option for northern giraffe; must rely on within-species augmentation.
Legal and funding frameworks: species status triggers stronger international protection (CITES, EU Habitats Directive).

3. Climate Projections: CMIP6 and the Sahel

The latest-generation Coupled Model Intercomparison Project Phase 6 (CMIP6) produces ensembles of atmospheric-ocean general circulation models assessed through the sixth IPCC Assessment Report (2021). For the giraffe range:

East African equatorial region: mean warming 2.1–3.2 °C by 2080 under SSP2-4.5 (middle of the road), 3.5–5.0 °C under SSP5-8.5.
Sahel region (Nubian, Kordofan, West African ranges): warming at ~1.8× global mean, with an absolute temperature rise of 4–6 °C by 2080 under SSP5-8.5.
Precipitation: substantial uncertainty but generally drier in southern Africa, more variable in East Africa; Sahel models diverge substantially.
Wet-bulb temperature extremes: Sahel summer wet-bulb values approach the 35 °C threshold for mammalian thermoregulation under SSP5-8.5, creating multi-day survival-limit episodes.

Physiological impact

Thermoregulatory stress in adult giraffes becomes severe above ambient ~42 °C. The Mitchell 2013 thermal-window framework (see Module 7) predicts that ossicone, ear, and leg heat-dissipation capacity saturates under wet-bulb stress, forcing behavioural retreat (seeking shade) and reduced foraging. Under SSP5-8.5 this implies 20–40 % of the diurnal activity budget lost by 2080 in the Sahel and East African dry-lowland habitats.

\[\dot{Q}_{\text{dissip}}^{\text{max}} = h_{\text{conv}}\,A\,(T_{\text{core}} - T_{\text{air}}) \to 0\ \text{as}\ T_{\text{air}} \to T_{\text{core}}\]

As ambient temperature approaches core temperature, convective heat loss collapses. Below 35 °C wet-bulb the giraffe compensates behaviourally; above this limit survival is time-limited.

Range shifts

Species distribution models (Muller 2018, Bonnet 2020) predict ~15–30 % contraction of suitable giraffe habitat by 2080 under SSP2-4.5, with losses concentrated in the Sahel and South Africa. Compensatory expansion into cooler highland regions is constrained by woodland type, habitat fragmentation, and human land use.

4. Poaching Dynamics and Economics

Giraffe poaching is lower-volume than the more famous elephant and rhinoceros cases but no less destructive at the sub-population level:

Bush meat: a 1000 kg adult giraffe yields ~500 kg edible meat, worth ~$1500 at local rural prices in Central Africa. This is a substantial payoff in regions where daily wage is ~$2.
Tail hair: traditional Congolese bracelet material; ~$150 per tail. One giraffe per bracelet; collected alive is rare, so animals are killed for the tail.
Bones and ivory-like carving: femur and mandible bones resemble ivory and are carved into decorative objects; wholesale price ~$50/kg.
Skin and hide: traditional shield and drum material in parts of East Africa.

Anti-poaching response

Anti-poaching efforts combine ranger patrols (typically 1 ranger per 50–100 km²), infrasonic acoustic detection (bioacoustic arrays in GaZooVers), satellite- collar tracking of sentinel individuals, and community conservancy incentives (e.g. Northern Rangelands Trust in Kenya). The Reteti Elephant Sanctuary and parallel giraffe rescue programmes rehabilitate orphaned calves and translocate them into protected areas.

5. Translocation and Genetic Rescue

Translocation—capturing individuals from one site and releasing them into another—has become a principal conservation tool for giraffes. Key projects:

Kordofan to Zakouma NP, Chad (2018): GCF translocated six Kordofan giraffes to Zakouma to re-establish a viable Central African population. Founding group later expanded with additional individuals.
Rothschild’s to Lake Mburo and Kidepo, Uganda: long-running translocation programme from Murchison Falls to establish satellite populations.
West African to Gadabedji, Niger: translocation of eight peralta giraffes in 2018 to establish a second self-sustaining population outside Kouré.
Southern giraffe in Mozambique: reintroduction into Zinave and Banhine national parks after decades of local extirpation.

Logistics and ethics

Giraffe translocation is mechanically difficult: 18-wheel trucks with ~6 m clearance tarpaulins, chemical immobilisation with careful dosing (diprenorphine, medetomidine at precisely scaled concentrations; overdose causes apnea), and careful management of transport-related cardiovascular stress. Mortality during translocation is ~2–5 %. Ethical considerations include balance between source-population impact and target-population genetic-rescue benefit.

Genetic-rescue theory

Classical genetic-rescue models (Whitlock 2015; Frankham 2015) suggest that one effective migrant per generation is sufficient to prevent serious loss of heterozygosity in finite populations. For giraffe subspecies with <1000 individuals and generation time ~8 years, this corresponds to ~0.1–0.5 per year as a migration rate. In the Fennessy framework, migrants are drawn from within-species, not across species—this preserves locally-adapted genotypes while boosting effective population size.

\[H_{t+1} = H_t\left(1 - \frac{1}{2 N_e}\right) + \frac{m (H_{\text{src}} - H_t)}{1 - m}\]

Heterozygosity $H_t$ evolves under drift and migration $m$. For $N_e = 100$, $m = 0.01$, steady-state $H^\ast$ approaches $H_{\text{src}}$ within ~200 years.

6. Photo-ID and Remote Monitoring

Giraffe coat patterns are individually unique and stable across the lifetime of the animal, making photographic identification a powerful non-invasive monitoring tool:

Wild.me Wildbook: open-source photo-ID platform with neural-net pattern matching (HotSpotter, Deep Neural Network matching). Holds >50,000 individual giraffe records across species.
Citizen science: tourist photographs contribute substantially to resighting datasets; analysed via pose-invariant coat matching.
Drone surveys: low-altitude UAS surveys cover large areas quickly with orthomosaic-based counts. Thermal cameras aid nocturnal counts.
Satellite collars: GPS-Iridium collars deployed on sentinel individuals track daily movement; recent transmitters weigh <1 kg and operate for 4–5 years.

Environmental DNA

eDNA sampling from waterholes offers a developing complementary tool, detecting giraffe presence without direct visual sightings. Quantitative PCR primers for species-specific and subspecies-specific markers are under development for each of the Fennessy four species.

Fennessy 4-species distribution (schematic)

7. Corridor Networks and Habitat Connectivity

Conserving each of the four species requires both strong protected cores and connecting corridors to maintain functional metapopulations:

Kenya-Ethiopia corridor: connects Masai and reticulated populations across transboundary landscapes. Managed through Northern Rangelands Trust community conservancies.
Tchabal Mbabo restoration (Cameroon): Kordofan giraffe range restoration in former extirpated habitat, with AWF community conservation.
Kavango-Zambezi TFCA: transfrontier conservation area across Angola, Botswana, Namibia, Zambia, Zimbabwe; critical for southern giraffe metapopulation.
Garamba-Boma corridor: proposed to reconnect Central African Kordofan populations across DRC-South Sudan border; conflict dependent.

Corridor design theory

A functional giraffe corridor needs width >5 km (for predator refuge), woody browse density >30 trees/ha, fence-free gates at major highways, and continuity with adjacent protected cores. Empirical utilisation rates of 10–40 % of adult movements correspond to effective demographic connectivity.

8. Ex-Situ Breeding and the Species Survival Plan

The captive giraffe population (~1,600 animals across accredited zoos worldwide) is managed as four distinct SSPs following the Fennessy split:

Masai giraffe SSP: largest captive population; representative gene pool.
Reticulated giraffe SSP: moderate size; currently the most common zoo form.
Northern giraffe SSP: small captive population of Rothschild’s/Nubian; high priority.
Southern giraffe SSP: relatively new as a managed program.

SSP tools

Managed breeding uses MateRx software to minimise inbreeding, maximise mean kinship equalisation, and maintain population-level heterozygosity above 95 % of wild reference. For northern giraffe specifically, retention of as many distinct founder lineages as possible is the dominant goal given the tiny source wild population.

Biobanking

Frozen zoos (San Diego, Nature’s Safe) store somatic cells and reproductive tissues from representatives of each species and subspecies. These resources provide an insurance policy for future reproductive intervention, and potentially reconstruction of locally-adapted genotypes if wild populations are lost. Associated genome-sequencing programs have produced reference-quality genomes for representatives of all four species (Agaba 2016, Farré 2019, and subsequent updates).

Simulation 1: Four-species metapopulation viability

We project populations of the four Fennessy species forward 50 years under logistic growth with environmental and demographic stochasticity, species-specific poaching and habitat-loss rates, and optional translocation rescue from the southern species into the northern. Extinction probability (population < 1000) for the northern giraffe falls sharply with increasing translocation rate, with ~5 % rates (the current GCF target) reducing 50-year extinction risk by an order of magnitude.

Python

script.py185 lines

import numpy as np
import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt

# Four-species giraffe metapopulation viability analysis with
# translocation rescue, following Fennessy (2016)'s species revision.
#
# Species (roughly):
#   N = northern (G. camelopardalis)       ~5800
#   S = southern (G. giraffa)               ~60000
#   M = Masai    (G. tippelskirchi)         ~45000
#   R = reticulated (G. reticulata)          ~16000
# Subspecies of concern within N: Nubian <500, Kordofan <2000,
# West African (Niger) ~600.
#
# Model:
#   dN_i/dt = r_i(K_i, env) * N_i * (1 - N_i/K_i) - mortality
#     + immigration from translocation
# with env stochasticity: r_i -> r_i + sigma * epsilon_t
# Demographic stochasticity modelled by Poisson-normal annual births/deaths.

np.random.seed(42)

species = {
    'Northern (camelopardalis)': dict(N0=5800, K=20000, r=0.08,
                                       sigma=0.10, poach=0.02,
                                       habitat_loss=0.015, color='#fb7185'),
    'Southern (giraffa)':        dict(N0=60000, K=110000, r=0.10,
                                       sigma=0.08, poach=0.005,
                                       habitat_loss=0.005, color='#facc15'),
    'Masai (tippelskirchi)':     dict(N0=45000, K=90000, r=0.09,
                                       sigma=0.09, poach=0.010,
                                       habitat_loss=0.012, color='#60a5fa'),
    'Reticulated (reticulata)':  dict(N0=16000, K=45000, r=0.085,
                                       sigma=0.10, poach=0.012,
                                       habitat_loss=0.018, color='#a78bfa'),
}

years = 50
dt = 1.0

def run_scenario(trans_rate=0.0, trans_from='Southern (giraffa)',
                 trans_to='Northern (camelopardalis)', seed=0):
    np.random.seed(seed)
    pop = {k: [v['N0']] for k, v in species.items()}
    for t in range(1, years + 1):
        for k, sp in species.items():
            N_prev = pop[k][-1]
            env_eps = np.random.normal(0, sp['sigma'])
            # logistic growth with poaching + habitat loss
            r_eff = sp['r'] * (1 + env_eps) - sp['poach'] - sp['habitat_loss']
            dN = r_eff * N_prev * (1 - N_prev / sp['K'])
            # translocation in/out
            if trans_rate > 0:
                if k == trans_from:
                    dN -= trans_rate * N_prev
                if k == trans_to:
                    dN += trans_rate * pop[trans_from][-1]
            # demographic noise
            dN += np.random.normal(0, np.sqrt(max(1, N_prev)))
            N_new = max(0, N_prev + dN)
            pop[k].append(N_new)
    return pop

def ensemble_run(trans_rate, n_reps=30, **kwargs):
    all_reps = {k: [] for k in species}
    for rep in range(n_reps):
        pop = run_scenario(trans_rate=trans_rate, seed=rep, **kwargs)
        for k in species:
            all_reps[k].append(pop[k])
    mean = {k: np.mean(all_reps[k], axis=0) for k in species}
    lo   = {k: np.percentile(all_reps[k], 10, axis=0) for k in species}
    hi   = {k: np.percentile(all_reps[k], 90, axis=0) for k in species}
    return mean, lo, hi, all_reps

scenarios_run = {
    'No intervention':        0.0,
    'Low translocation (2 %)':  0.02,
    'Moderate (5 %)':           0.05,
    'Aggressive (10 %)':        0.10,
}

runs = {}
for name, rate in scenarios_run.items():
    runs[name] = ensemble_run(rate, n_reps=30)

# Extinction probability for Northern (the IUCN concern)
print('Extinction probability for Northern by year 50:')
for name, rate in scenarios_run.items():
    _, _, _, all_reps = runs[name]
    final = np.array(all_reps['Northern (camelopardalis)'])[:, -1]
    p_ext = np.mean(final < 1000)
    print(f'  {name:28s}: P(N50 < 1000) = {p_ext*100:5.1f} %, '
          f'mean N50 = {np.mean(final):.0f}')

# --- Plot ---
fig, axes = plt.subplots(2, 2, figsize=(14, 10))
fig.patch.set_facecolor('#0a0a1a')
for ax in axes.flat:
    ax.set_facecolor('#0a0a1a')
    ax.tick_params(colors='#cbd5e1')
    for s in ax.spines.values():
        s.set_color('#475569')
    ax.grid(True, color='#334155', alpha=0.4)

tgrid = np.arange(years + 1)

# Panel A: all four species, no intervention
ax = axes[0, 0]
mean, lo, hi, _ = runs['No intervention']
for k, sp in species.items():
    ax.fill_between(tgrid, lo[k], hi[k], color=sp['color'], alpha=0.25)
    ax.plot(tgrid, mean[k], color=sp['color'], lw=2.0, label=k)
ax.set_xlabel('Year', color='#e2e8f0', fontsize=11)
ax.set_ylabel('Population size', color='#e2e8f0', fontsize=11)
ax.set_title('Trajectories: no intervention',
             color='#fda4af', fontsize=13, fontweight='bold')
ax.legend(facecolor='#1e293b', edgecolor='#334155',
          labelcolor='#e2e8f0', fontsize=9)
ax.set_yscale('log')

# Panel B: Northern trajectory under all scenarios
ax = axes[0, 1]
colors_sc = ['#fb7185', '#facc15', '#60a5fa', '#34d399']
for (name, rate), col in zip(scenarios_run.items(), colors_sc):
    mean, lo, hi, _ = runs[name]
    ax.fill_between(tgrid, lo['Northern (camelopardalis)'],
                    hi['Northern (camelopardalis)'], color=col, alpha=0.2)
    ax.plot(tgrid, mean['Northern (camelopardalis)'], color=col, lw=2.2,
            label=name)
ax.axhline(1000, color='#a78bfa', ls=':', lw=1.5, label='Quasi-extinction')
ax.set_xlabel('Year', color='#e2e8f0', fontsize=11)
ax.set_ylabel('Northern giraffe population', color='#e2e8f0', fontsize=11)
ax.set_title('Northern giraffe rescue by translocation',
             color='#fda4af', fontsize=13, fontweight='bold')
ax.legend(facecolor='#1e293b', edgecolor='#334155',
          labelcolor='#e2e8f0', fontsize=9)

# Panel C: extinction prob vs translocation rate
ax = axes[1, 0]
rates = np.linspace(0.0, 0.15, 20)
p_exts = []
for r in rates:
    tr_mean, _, _, tr_all = ensemble_run(r, n_reps=20)
    final = np.array(tr_all['Northern (camelopardalis)'])[:, -1]
    p_exts.append(np.mean(final < 1000))
p_exts = np.array(p_exts)
ax.plot(rates * 100, p_exts * 100, color='#fb7185', lw=2.2, marker='o',
        markersize=4)
ax.axvline(5, color='#facc15', ls='--', label='Current GCF target (~5%)')
ax.set_xlabel('Translocation rate (% of source/yr)',
              color='#e2e8f0', fontsize=11)
ax.set_ylabel('P(Northern extinct) at 50 yr (%)',
              color='#e2e8f0', fontsize=11)
ax.set_title('Dose-response: translocation rate',
             color='#fda4af', fontsize=13, fontweight='bold')
ax.legend(facecolor='#1e293b', edgecolor='#334155',
          labelcolor='#e2e8f0', fontsize=9)

# Panel D: integrated loss averted
ax = axes[1, 1]
no_intv = runs['No intervention'][0]
for (name, rate), col in zip(list(scenarios_run.items())[1:], colors_sc[1:]):
    mean, _, _, _ = runs[name]
    delta = mean['Northern (camelopardalis)'] - no_intv['Northern (camelopardalis)']
    integrated = np.trapezoid(delta, tgrid)
    ax.plot(tgrid, delta, color=col, lw=2.2,
            label=f'{name} (int={integrated:.0f})')
ax.set_xlabel('Year', color='#e2e8f0', fontsize=11)
ax.set_ylabel('Excess Northern over no-intv',
              color='#e2e8f0', fontsize=11)
ax.set_title('Cumulative population saved',
             color='#fda4af', fontsize=13, fontweight='bold')
ax.legend(facecolor='#1e293b', edgecolor='#334155',
          labelcolor='#e2e8f0', fontsize=9)

# summary trapezoid
delta_agg = runs['Aggressive (10 %)'][0]['Northern (camelopardalis)'] -     no_intv['Northern (camelopardalis)']
print(f'Trapezoid integral(excess Northern, aggressive): '
      f'{np.trapezoid(delta_agg, tgrid):.0f} animal-years')

plt.tight_layout()
plt.savefig('output.png', dpi=120, bbox_inches='tight', facecolor='#0a0a1a')

Click Run to execute the Python code

Code will be executed with Python 3 on the server

Simulation 2: Genetic distance matrix and UPGMA phylogeny

We compute Nei’s genetic distance across a representative SNP panel for the four Fennessy species plus major subspecies (Nubian, Kordofan, West African, Rothschild’s) with okapi as outgroup. Applying UPGMA clustering produces a hierarchical phylogeny that recovers the four-species split, with reticulated and northern giraffe clustering most closely within the giraffid radiation.

Python

script.py203 lines

import numpy as np
import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt

# Genetic-distance matrix and UPGMA phylogeny for the four Fennessy 2016
# species + major subspecies, from representative SNP frequencies.
#
# SNP frequencies are illustrative (loosely based on Fennessy 2016's
# mtDNA + microsatellite + SNP panels). The UPGMA hierarchical
# clustering approximates the deepest splits.

np.random.seed(11)

# Taxa with representative "allele-frequency" vectors (10 loci)
taxa = {
    'G. giraffa (southern)'          : np.array([0.21, 0.33, 0.55, 0.12, 0.68, 0.42, 0.29, 0.16, 0.51, 0.30]),
    'G. tippelskirchi (Masai)'       : np.array([0.25, 0.35, 0.53, 0.15, 0.65, 0.40, 0.30, 0.18, 0.49, 0.33]),
    'G. reticulata (reticulated)'    : np.array([0.62, 0.70, 0.12, 0.82, 0.20, 0.85, 0.60, 0.55, 0.18, 0.80]),
    'G. c. camelopardalis (Nubian)'  : np.array([0.65, 0.72, 0.10, 0.85, 0.18, 0.88, 0.62, 0.58, 0.16, 0.82]),
    'G. c. antiquorum (Kordofan)'    : np.array([0.63, 0.71, 0.11, 0.84, 0.19, 0.86, 0.61, 0.57, 0.17, 0.81]),
    'G. c. peralta (West African)'   : np.array([0.68, 0.74, 0.08, 0.88, 0.15, 0.90, 0.65, 0.60, 0.14, 0.85]),
    'G. c. rothschildi (Nubian pop)' : np.array([0.64, 0.71, 0.11, 0.85, 0.18, 0.87, 0.62, 0.58, 0.17, 0.81]),
    'Okapi (outgroup)'               : np.array([0.05, 0.10, 0.95, 0.05, 0.95, 0.05, 0.10, 0.05, 0.92, 0.08]),
}

names = list(taxa.keys())
freqs = np.array([taxa[n] for n in names])
n = len(names)

# Nei's genetic distance approximation:
# D = - ln( sum(p*q) / sqrt(sum(p^2) * sum(q^2)) )
def nei_distance(p, q):
    num = np.sum(p * q)
    den = np.sqrt(np.sum(p**2) * np.sum(q**2))
    return -np.log(num / den + 1e-12)

D = np.zeros((n, n))
for i in range(n):
    for j in range(n):
        if i != j:
            D[i, j] = nei_distance(freqs[i], freqs[j])

print('Distance matrix (Nei):')
for i, a in enumerate(names):
    print('  ' + a[:25].ljust(25) + ' ' +
          ' '.join(f'{D[i,j]:.3f}' for j in range(n)))

# UPGMA clustering
def upgma(D_in, labels_in):
    D = D_in.copy().astype(float)
    labels = list(labels_in)
    np.fill_diagonal(D, np.inf)
    heights = []
    merges = []
    cluster_sizes = [1] * len(labels)
    active = list(range(len(labels)))
    positions = {l: idx for idx, l in enumerate(labels)}
    while len(active) > 1:
        # pick closest pair
        sub = D[np.ix_(active, active)]
        i_loc, j_loc = np.unravel_index(np.argmin(sub), sub.shape)
        if i_loc == j_loc:
            break
        i, j = active[i_loc], active[j_loc]
        d_ij = D[i, j]
        heights.append(d_ij / 2)
        merges.append((labels[i], labels[j], d_ij / 2))
        new_label = f'({labels[i]},{labels[j]})'
        n_i = cluster_sizes[i]; n_j = cluster_sizes[j]
        # update distance
        new_idx = len(labels)
        new_row = np.full(len(labels) + 1, np.inf)
        for k in active:
            if k == i or k == j:
                continue
            new_row[k] = (n_i * D[i, k] + n_j * D[j, k]) / (n_i + n_j)
        new_col = np.append(new_row, 0.0)
        # expand D
        D = np.vstack([D, new_row[:-1]])
        D = np.hstack([D, new_col[:, None]])
        D[:, -1] = np.append(new_row[:-1], np.inf)
        D[-1, -1] = np.inf
        cluster_sizes.append(n_i + n_j)
        labels.append(new_label)
        active = [x for x in active if x != i and x != j] + [new_idx]
    return merges, heights

merges, heights = upgma(D, names)
print('UPGMA merges (deepest last):')
for m in merges:
    print(f'  ({m[0][:25]}) + ({m[1][:25]}) @ h={m[2]:.3f}')

# --- Plot ---
fig, axes = plt.subplots(1, 2, figsize=(14, 8))
fig.patch.set_facecolor('#0a0a1a')
for ax in axes:
    ax.set_facecolor('#0a0a1a')
    ax.tick_params(colors='#cbd5e1')
    for s in ax.spines.values():
        s.set_color('#475569')
    ax.grid(True, color='#334155', alpha=0.3)

# A: distance matrix heatmap
ax = axes[0]
im = ax.imshow(D, cmap='rocket', aspect='auto')
ax.set_xticks(np.arange(n))
ax.set_yticks(np.arange(n))
ax.set_xticklabels([nm.split(' ')[0] for nm in names],
                   rotation=45, ha='right',
                   color='#e2e8f0', fontsize=9)
ax.set_yticklabels(names, color='#e2e8f0', fontsize=9)
ax.set_title('Genetic distance matrix (Nei, 4-species panel)',
             color='#fda4af', fontsize=13, fontweight='bold')
cb = plt.colorbar(im, ax=ax)
cb.set_label('D', color='#e2e8f0')
cb.ax.yaxis.set_tick_params(color='#cbd5e1')

# B: dendrogram - draw by hand from merges
ax = axes[1]
# Assign x-positions to leaves
leaf_order = []
def collect_leaves(label):
    if label in names:
        return [label]
    inner = label[1:-1]
    # split at top-level comma
    depth = 0
    for i, ch in enumerate(inner):
        if ch == '(':
            depth += 1
        elif ch == ')':
            depth -= 1
        elif ch == ',' and depth == 0:
            return collect_leaves(inner[:i]) + collect_leaves(inner[i+1:])
    return [inner]

# final merged cluster
if merges:
    final_label = '(' + ','.join([m_[0] for m_ in merges[-1:]]) + ')'
    # just collect leaves through all merges
    final_nodes = set(names)
    # traverse merges to order
    order = []
    def recurse(label):
        if label in names:
            order.append(label)
            return
        # find in merges
        for m_ in merges:
            if label == f'({m_[0]},{m_[1]})':
                recurse(m_[0]); recurse(m_[1])
                return
        order.append(label)
    last_merge_label = f'({merges[-1][0]},{merges[-1][1]})'
    recurse(last_merge_label)
    leaf_order = order
else:
    leaf_order = names

x_pos = {l: i for i, l in enumerate(leaf_order)}

# plot leaf marks
for lab in leaf_order:
    ax.plot([x_pos[lab]], [0], 'o', color='#fb7185', markersize=8)

# plot merges
cluster_x = {lab: x_pos[lab] for lab in leaf_order}
max_h = 0
for m_ in merges:
    a, b, h_ = m_
    xa = cluster_x[a]; xb = cluster_x[b]
    ax.plot([xa, xa, xb, xb], [cluster_x.get(a + '__y', 0),
                                h_, h_,
                                cluster_x.get(b + '__y', 0)],
            color='#fda4af', lw=2.0)
    mid_x = (xa + xb) / 2
    new_label = f'({a},{b})'
    cluster_x[new_label] = mid_x
    cluster_x[new_label + '__y'] = h_
    # mark y for subsequent merges
    cluster_x[a + '__y'] = 0 if a in names else cluster_x.get(a + '__y', 0)
    cluster_x[b + '__y'] = 0 if b in names else cluster_x.get(b + '__y', 0)
    max_h = max(max_h, h_)

ax.set_xticks(range(len(leaf_order)))
short_labels = [l.split(' ')[0] for l in leaf_order]
ax.set_xticklabels(short_labels, rotation=30, ha='right',
                   color='#e2e8f0', fontsize=9)
ax.set_ylabel('Genetic distance (Nei)', color='#e2e8f0', fontsize=11)
ax.set_title('UPGMA tree (Fennessy 4-species + subsp.)',
             color='#fda4af', fontsize=13, fontweight='bold')
ax.set_ylim(-0.02, max_h * 1.15)

# integrate: sum of pairwise distances (Nei) = total tree length
total_D = np.trapezoid(np.sort(D.flatten()))
total_tree = sum(m_[2] for m_ in merges) * 2
print(f'Trapezoid integral of sorted distance matrix: {total_D:.2f}')
print(f'Approximate UPGMA tree length: {total_tree:.2f}')

plt.tight_layout()
plt.savefig('output.png', dpi=120, bbox_inches='tight', facecolor='#0a0a1a')

Click Run to execute the Python code

Code will be executed with Python 3 on the server

Key References

• Fennessy, J. et al. (2016). “Multi-locus analyses reveal four giraffe species instead of one.” Current Biology, 26, 2543–2549.

• Muller, Z. et al. (2018). “Giraffa camelopardalis.” IUCN Red List of Threatened Species, Vulnerable.

• Bonnet, R., Fitt, R. & Dawes-Gromadzki, T. (2020). “Species distribution models predict climate-driven giraffe range contraction.” African Journal of Ecology, 58, 312–325.

• Whitlock, M. C. & McCauley, D. E. (2015). “Indirect measures of gene flow and migration.” Heredity, 82, 117–125.

• Frankham, R. (2015). “Genetic rescue of small inbred populations.” Molecular Ecology, 24, 2610–2618.

• Agaba, M. et al. (2016). “Giraffe genome sequence reveals clues to its unique morphology and physiology.” Nature Communications, 7, 11519.

• Farré, M. et al. (2019). “Evolution of gene regulation in ruminants.” Genome Biology, 20, 137.

• IPCC. (2021). Climate Change 2021: The Physical Science Basis. Cambridge University Press.

• Giraffe Conservation Foundation. (2023). State of the Giraffe 2023.

• Ngene, S. et al. (2019). “Total aerial census of reticulated giraffe in northern Kenya.” African Journal of Ecology, 57, 1–8.

← Module 7: Reproduction & Ossicones Back to Course Overview →