Fig 1.
A GAM explains extracted stable densities from time series of 18 savanna elephant populations.
The component smooth functions of mean EVI, percentage of protected area within 12km of water, and PIKE of the selected best GAM are represented by solid lines transformed to the response scale of density (elephants/km2). Points represent partial residuals of the 18 extracted densities, and grey lines represent 1,000 iterations to account for uncertainty.
Fig 2.
Predicted ecological benchmark density for savanna elephant populations across 73 protected areas differed according to EVI and water availability given minimized poaching.
(A) Green fill color indicates predicted ecological benchmark density as shown in (B) the frequency distribution where densities on the x-axis represent bin centers. (C) Twenty-one countries are covered by protected areas in this assessment. See Figure E and Table F in S1 File for the 95% prediction intervals of predicted ecological benchmarks.
Fig 3.
Most recent density estimates as a percentage of predicted ecological benchmark densities for savanna elephant populations across 73 protected areas indicate most populations are far from ecological benchmarks.
(A) Fill color illustrates most recent density estimates as a percentage of predicted ecological benchmark density as indicated in (B) the frequency distribution where percentages on the x-axis represent bin centers. (C) Populations plotted according to the same color scheme show that while few populations are near their predicted ecological benchmark (indicated by the solid 1:1 line), recent density estimates increased with predicted ecological benchmark density (dashed line: y = 0.31x-0.12, F1,71 = 16.38, p<0.001, R2 = 0.19).
Fig 4.
Of 73 protected areas assessed, 72 had a deficit of elephants compared to ecological benchmarks.
(A) Subtracting predicted ecological benchmark population size from corresponding recent population estimates yields deficit or surplus estimates for each protected area, illustrated by size and color of circles (net deficit = 729,505). (B) A frequency distribution, where x-axis represents bin centers, illustrates that few populations have large deficits, while most are missing fewer than 12,000 elephants. See Table F in S1 File for the 95% prediction intervals of population deficits/surpluses.