Highlights

•  In northern Botswana, fresh elephant carcasses increased by 593% from 2014 to 2018
• In 2018, we confirmed 156 elephants as poached for ivory based on damage to skulls
• Fresh elephant carcasses were clustered in 5 hotspots where elephants have declined
• We estimate that a minimum of 385 elephants were poached in Botswana in 2017–2018

In Brief 

Botswana holds one-third of Africa’s elephants but has experienced little poaching for ivory. Schlossberg et al. document a new elephant poaching outbreak in Botswana, with large increases in numbers of fresh elephant carcasses and 156 elephants confirmed as poached. Fresh carcasses are clustered in five hotspots where elephants have declined.

SUMMARY

Botswana holds roughly one-third of Africa’s remain- ing savannah elephants (Loxodonta africana) [1, 2] and will play a key role in the future conservation of this species. To date, Botswana has been one of the safest countries for elephants, with little poach- ing reported [3]. Here, we present evidence of a new outbreak of elephant poaching for ivory in northern Botswana. Comparing results from 2014 and 2018 aerial surveys, we found that elephant populations were stable, but numbers of elephant carcasses have increased, especially for newer car- casses dead for less than roughly 1 year. Newer car- casses were clustered in five ‘‘hotspots’’ averaging 3,522 km2 in area. We compared elephant popula- tions in hotspots to the immediately surrounding areas and found that since 2014, elephants have decreased by 16% in hotspots but increased by 10% in surrounding areas. Numbers of ‘‘old’’ car- casses, dead for more than 1 year, increased by 78% in hotspots between 2014 and 2018 but decreased by 3% in surrounding areas. To verify that poaching has been occurring, we used helicop- ters to visit 148 elephant carcasses and assess their cause of death. We confirmed poaching for all 72 newer carcasses assessed. We also confirmed poaching for 62 of 76 (82%) carcasses older than 1 year, primarily in one hotspot. Poached older car- casses were all males aged 30–60 and likely killed for their large tusks. This evidence suggests that ivory poaching on the scale of hundreds of elephants per year has been occurring in northern Botswana since 2017 or possibly earlier.

RESULTS

Regional Elephant Populations Are Stable while Carcasses Have Increased.
To monitor elephant populations, we used fixed-wing aircraft to conduct aerial surveys in northern Botswana during the 2014 and 2018 dry seasons. For the 94,000 km2 study area (Figure S1), estimated elephant populations were nearly unchanged be- tween 2014 and 2018 at approximately 122,700 elephants (two-sample Z test, Z = 0.03, p = 0.98; Figure 1). Numbers of el- ephants in breeding herds, which include females and young as well as occasional males, increased non-significantly (Z = 0.36, p = 0.72).

Elephant carcasses remain visible for years after an elephant’s death, and numbers of elephant carcasses can be an indicator of population status [4]. On surveys, we distinguished elephant car- casses by state of decomposition and likely age (Table S1) [5]. ‘‘Fresh’’ and ‘‘recent’’ carcasses still had flesh and likely died within 1 year prior to the survey. ‘‘Old’’ carcasses were mainly bones and were likely dead for >1 year. ‘‘Very old’’ carcasses had been dead for up to 10 years. Estimated numbers of all elephant carcasses increased by 21% between 2014 and 2018 (two-sample Z test, Z = 3.88, p < 0.001; Figure 1). Estimated numbers of fresh and recent carcasses increased by 593% (Z = 7.94, p < 0.001). Carcass ratios, defined as the number of carcasses divided by the sum of carcasses plus live elephants, increased significantly between 2014 and 2018 for all carcasses and for fresh and recent carcasses alone (Figure 1).

Carcasses Were Clustered in Hotspots, Where Elephants Are Declining
Past studies have shown that elephant poaching occurs in clus- ters where poachers are operating and results in carcasses that
are grouped in space [6–8]. Fresh and recent carcasses were clearly clustered on the 2018 aerial survey (Figure 2). Single-link- age cluster analysis identified five ‘‘hotspots,’’ ranging in size from 829 to 6,231 km2 (mean area = 3,522 km2), where fresh and recent carcasses were clustered (Figure 2). Density of fresh and recent carcasses in hotspots was 0.04 carcasses/km 2 versus 0.001 carcasses/km 2 in surrounding areas, defined as a 40km buffer around each hotspot.

Carcasses Were Clustered in Hotspots, Where Elephants Are Declining
Past studies have shown that elephant poaching occurs in clus- ters where poachers are operating and results in carcasses that
are grouped in space [6–8]. Fresh and recent carcasses were clearly clustered on the 2018 aerial survey (Figure 2). Single-link- age cluster analysis identified five ‘‘hotspots,’’ ranging in size from 829 to 6,231 km2 (mean area = 3,522 km2), where fresh and recent carcasses were clustered (Figure 2). Density of fresh and recent carcasses in hotspots was 0.04 carcasses/km 2 versus 0.001 carcasses/km 2 in surrounding areas, defined as a 40km buffer around each hotspot.

To learn how elephant populations are faring in the hotspots, we compared numbers of elephants in hotspots with numbers in 40-km buffers surrounding the hotspots. For all five hotspots combined, overall elephant populations decreased from 26,710 ± 2,227 to 22,391 ± 1,813 in hotspots (16%) while increasing from 58,350 ± 3,585 to 64,330 ± 3,192 in surrounding areas (+10%; Figure 3). The difference in trends between hot- spots and surroundings was significant (two-sample Z test, Z = 2.00, p = 0.04). Differences in trends were not significant for bull herds (Z = 0.86, p = 0.39) or for breeding herds (Z = 1.88, p = 0.06; Figure 3). For elephants, most individual hotspots showed patterns similar to the first row of Figure 3 (Figure S2).

We also compared hotspots and surrounding areas for change in numbers of carcasses. Old carcasses increased by 78% ± 28% in hotspots between 2014 and 2018 but decreased by 3% ± 11% in surrounding areas, a significant difference (two- sample Z test, Z = 2.71, p = 0.007; Figure 3). Differences between hotspots and surroundings for old carcasses were significant only for the NG 15 hotspot (Figure S2). For very old carcasses, overall numbers increased by 46% ± 22% in hotspots and 47% ± 12% in surrounding areas (Z = 0.04, p = 0.97). By hotspot, the only significant result for very old carcasses was that numbers increased in areas surrounding the Chinamba hotspot but decreased within the hotspot (Z = 2.41, p = 0.02). As ex- pected, because the hotspots were defined using 2018 locations of fresh and recent carcasses, numbers of fresh and recent car- casses increased substantially in hotspots but were nearly un- changed in surrounding areas (Z = 2.05, p = 0.04; Figure 3).

Carcass Assessments Verify 156 Poached Elephants

During the 2018 aerial survey, observers noted dozens of poten- tially poached elephant carcasses. To verify the cause of death for some carcasses, we used a helicopter to visit carcasses on the ground or photograph them from low altitude. We searched for signs of poaching for ivory such as skulls chopped with an axe to remove tusks or carcasses covered with brush to hide them. We first visited 72 fresh and recent carcasses, likely dead for <1 year at the time of the survey and suspected to be poached. We confirmed that all 72 carcasses were poached (Figures 4A and S3; photographs of all carcasses are in Data S3). Photographs of an additional 22 carcasses taken during the 2018 aerial survey provided sufficient evidence to determine that they had been poached. Thus, we confirmed poaching for 94 fresh or recent elephant carcasses.

We also visited old carcasses, likely dead for >1 year, to assess whether poaching may have been going on for over 1 year before the 2018 survey. We sampled old carcasses near the Linyanti River and in the NG 15 hotspot (Figure S3). Of the 76 carcasses visited, we determined that 62 (82%) were poached (Figure 4B); for three carcasses (5%), we could not determine the cause of death. In the Linyanti River area, outside any hotspot, 1 of 9 (11%) carcasses was poached; in the NG 15 hotspot, 61 of 67 carcasses were poached (91%; Figure S3). We visited 59 old carcasses on the ground and observed 17 from low altitude. The proportion of carcasses considered poached did not differ between ground visits (80%) and aerial visits (88%; Fisher’s exact test, p = 0.72). Mean estimated age at death for poached old carcasses (x = 41.4 ± 1.1 years) was significantly greater than for elephants that died of natural or unknown causes (x = 29.5 ± 3.7 years; two-sample t test, t12 =  3.12, p = 0.009;Figure 4C). All 47 poached carcasses that could be sexed were males. Remaining carcasses included an equal number of males and females/unknown-sex carcasses (Figure 4D).

Poached Elephant Estimates

Using only fresh and recent carcasses that were seen on the 2018 aerial survey and that we confirmed as poached in the hot- spots, we estimated a total of 385 ± 54 elephants were poached in the hotspots in roughly 1 year prior to the 2018 survey.

Testing Alternative Hypotheses

To test causes other than poaching for recent changes in elephant populations and carcass numbers, we compared four environmental variables between carcass hotspots and sur- rounding areas. The enhanced vegetation index (EVI), a measure of vegetation greenness and food availability for elephants, aver- aged 1.7% greater in hotspots (x = 0.284 ± 0.006) than in sur- rounding areas (x = 0.279 ± 0.007), a significant difference via meta-analysis on monthly values (effect size = 0.10 ± 0.02, Z = 5.61, p < 0.001). The Palmer Drought Severity Index (PDSI) did not differ significantly between hotspots (x =  1.87 ± 0.10) and surrounding areas (x =  1.90 ± 0.10; effect size = 0.01 ± 0.04). Elephant density in 2014 also did not differ between hot- spots (x = 1.56 ± 0.13 elephants/km 2) and surrounding areas (x = 1.34 ± 0.08 elephants/km 2; Z =  1.42, p = 0.16). Estimated human densities were 0.49 persons/km 2 in hotspots versus 0.63 persons/km 2 in surrounding areas (no test possible).

DISCUSSION

Multiple lines of evidence support the conclusion that elephant poaching has increased recently in northern Botswana. Estimated numbers of fresh and recent carcasses increased by 593% be- tween 2014 to 2018, and we verified 94 fresh/recent carcasses as poached. Thus, we have strong evidence of poaching in roughly the year prior to our 2018 survey. We also verified 62 old carcasses as poached, mainly in the NG 15 hotspot, and we observed increases in numbers of old carcasses between 2014 and 2018 in the hotspots. This suggests that poaching for ivory was going on for over a year prior to our 2018 survey, at least in one hotspot. One caveat of these findings is that age estimates for different categories of carcass are somewhat uncertain [9]. Thus, the estimates that fresh and recent carcasses are under 1 year old and that old carcasses are over 1 year old should not be considered exact.

On our 2014 aerial survey in northern Botswana, we did not re- cord any elephant carcasses suspected of being poached [10]. Of course, the fact that we did not observe suspected poached carcasses in 2014 does not indicate that no poaching was occur- ring. Data shown here, however, suggest that any poaching that was occurring in 2014 was likely much lower in frequency than in 2018.

For elephants, overall population estimates were nearly un- changed between 2014 and 2018, with a non-significant in- crease of just 0.2%. The 8% carcass ratio observed in 2018 is consistent with a stable population, as 8% represents the ex- pected boundary between growing and declining populations [4]. At the same time, numbers of carcasses and carcass ratios have increased substantially since 2014, suggesting that mortality rates have increased, even if elephant populations are not changing.

Increases in numbers of carcasses are worrisome because they can portend future increases in poaching and declines in elephant populations. In Zimbabwe’s Sebungwe ecosystem, numbers of carcasses increased in the early 2000s while elephant populations generally remained stable [11]. This stable period, however, was followed by a population collapse, with 2014 numbers down by 76% from the early 2000s [2]. Likewise,
in Mozambique’s Niassa National Reserve, increases in carcass ratios beginning in 2009 preceded a 78% decrease in elephant populations in just 5 years [12]. In Kenya’s Tsavo ecosystem, an increase in carcass ratios to $10% preceded a population crash in the 1970s (IDH and Hillman). In our hotspots, carcass ratios increased from 5% to 16% between 2014 and 2018 (Figure 1). This change may be a warning sign that Botswana’s elephant population could be at greater risk in the near future.

In Botswana, fresh and recent carcasses were clustered in five hotspots, primarily in northern parts of the study area. Past studies have shown that poaching tends to be clustered in space [6–8]. The hotspots should be thought of as coarse estimates of where poaching may be occurring. Exact boundaries of the hot- spots are somewhat arbitrary because they depend on the stop- ping rule used in the clustering algorithm. Also, because we have only been able to assess 53% of fresh and recent carcasses observed on the 2018 survey for poaching, we based our hot- spots on all fresh and recent carcasses seen on the survey.

Nonetheless, we suspect that 104 of 128 fresh and recent car- casses observed on the survey were poached (unpublished data), and all suspected carcasses that have been verified by he- licopter to date have been found to be poached. Thus, we believe that our hotspots are good approximations of where poaching is taking place.

Despite uncertainty about hotspot boundaries, we found sig- nificant differences in the status of elephant populations inside and outside of hotspots. From 2014 to 2018, elephant popula- tions decreased by 16% in hotspots but increased by 10% in surrounding areas. Estimated numbers of old carcasses increased 78% in hotspots while changing little in surrounding areas. Very old carcasses, however, showed nearly identical trends in hotspots and surrounding areas. Increases in numbers of old carcasses may be additional evidence that poaching in northern Botswana hotspots has been going on for more than 1 year but likely less than 4 years, as earlier poaching would have been noted in the 2014 survey or via changes in numbers of very old carcasses. Hotspots should be focal points for anti-poaching efforts and enhanced monitoring of elephant populations.

The one exception to the trends in hotspots was the Chinamba hotspot, where elephant populations actually increased more in the hotspot than in the surrounding area. The Chinamba area may be a false positive, where a few carcass observations in a relatively small area were joined into a cluster, and not an actual poaching hotspot. Verification of carcass origins in this area would help to determine the status of this hotspot.

To be clear, we are not claiming that poaching is entirely responsible for increases in carcass numbers or for the lack of population growth for elephants in recent years. Doing that would require far more verifications of carcass origins than we conducted. Movements of elephants likely affected observed numbers. Elephants can roam freely over most of northern Botswana, and satellite tracking has shown that elephants often move long distances within or between seasons [13, 14]. Elephants have been known to flee areas where poaching is occurring [15, 16].

Processes such as drought, food short- ages, overcrowding, and human-elephant conflict can also influence elephant population sizes and mortality rates [17–20]. Our comparison of hotspots and adjacent surrounding areas should provide spatial controls for many processes that mayaffect elephants. To test alternative explanations for recent changes, we compared hotspots and surrounding areas for drought severity (PDSI), food availability (EVI), overcrowding (2014 elephant density), and human population density, a proxy for human-elephant conflict. Botswana experienced a significant drought in 2015 and 2016 [21]. PDSI, however, did not differ significantly between hotspots and surrounding areas. Drought could, however, explain the increased numbers of very old car- casses observed in hotspots and surrounding areas in 2018 rela- tive to 2014 [20].

EVI was 1.7% greater in hotspots than surroundings, a signif- icant difference. This difference does not appear to be biologi- cally meaningful because the effect size for the meta-analysis was 0.10; effect sizes < 0.20 are considered ‘‘small’’ [22]. The dif- ference in EVI between hotspots and surroundings was also con- trary to what one might expect if EVI measures food availability, as elephants decreased the most in areas where EVI was great- est. As a further test of EVI effects, we used Bayesian errors-in- variables models to test effects of EVI on changes in elephant and carcass numbers. We found no support for effects of EVI on carcasses or elephants (unpublished data).

Elephant density in 2014 did not differ significantly between hotspots and surroundings, so density-dependent mechanisms did not likely affect our results. Human populations tended to be larger and denser in surrounding areas than in hotspots. This result is the opposite of what one might expect if human- elephant conflict were driving changes in elephant and carcass numbers. As above, we used Bayesian models to test effects of human density on changes in elephants and carcasses and found no support for this variable (unpublished data). Thus, we conclude that environmental conditions and human populations are unlikely to explain the different trajectories of elephants and carcasses in hotspots and surrounding areas.

In total, we verified 134 carcasses as poached during visits with a helicopter and 22 more carcasses using photographs taken during our aerial survey. Evidence of poaching was obvious. All poached elephants had skulls that were essentially chopped in half to facilitate removal of tusks (see Data S3). This damage to the skull remains visible long after flesh has de- composed, which allowed us to assess poaching in old car- casses. Carcasses originating from natural causes are not marred in this way, and skulls are left intact. Seventy-three percent of poached carcasses were also covered with cut branches, presumably to conceal the carcass. A few of the newer carcasses had wounds on their spines where poachers may have attempted to paralyze a wounded animal before removing tusks.

With conservative restrictions of only using carcasses verified as poached and limiting analyses to hotspots, we estimated that 385 elephants were poached in roughly 1 year prior to our 2018 survey. This is surely an underestimate because we have only been able to assess 53% of the fresh/recent carcasses seen on the survey, and few carcasses have been assessed in some of the hotspots (Figure S3). In addition, observers likely miss car- casses on aerial surveys, which would also bias our estimate of poached elephants low.

The fact that 61 old carcasses were verified as poached in the NG 15 hotspot suggests that the poaching in this area was occurring for more than roughly 1 year before our survey. The old carcasses that we assessed were not noted as poached on our survey. Rather, we searched for older carcasses in areas where we had observed poached fresh and recent carcasses as well as high densities of old carcasses during the aerial survey. Comparing the age and sex of poached carcasses with carcasses of natural origin showed that poachers were concentrating on older bulls (Figures 4C and 4D). Research has shown that poachers preferentially target older bulls for their large tusks, especially in previously unexploited populations [23, 24]. The NG 15 hotspot was the only hot- spot where numbers of old carcasses increased significantly more in the hotspot than in surrounding areas, but the other four hotspots showed similar but non-sig- nificant trends (Figure S2).

In conclusion, a new and growing poaching problem, with hundreds of ele- phants killed per year for their tusks, is ongoing in northern Botswana. To date, poaching seems to be primarily taking place in four hotspots and is mainly impacting mature bulls. Poaching in

Botswana does not appear to have reached the levels seen in central or eastern Africa in recent years, where elephant popula- tions decreased by 50% or more in some countries in under a decade [2]. In the past, Botswana has avoided serious poaching by combatting poachers with armed units [25]. Examples else- where in Africa show that such anti-poaching efforts can greatly reduce poaching [26, 27]. The hotspots that we have identified would be good locations to target for such efforts. Addressing this problem now, while poaching levels are relatively low, may aid in the conservation of elephants in Botswana.