The hitherto difficult task of reliably estimating populations of wide-ranging megafauna has been enabled by advances in capture-recapture methodology. Here we combine photographic sampling with a Bayesian spatially-explicit capture-recapture (SCR) model to estimate population parameters for the endangered Asian elephant Elephas maximus in the productive floodplain ecosystem of Kaziranga National Park, India. Posterior density estimates of herd-living adult females and sub-adult males and females (herd-adults) was 0.68 elephants/km2 (95% Credible Intervals, CrI = 0.56-0.81) while that of adult males was 0.24 elephants/km2 (95% CrI = 0.18-0.30), with posterior density estimates highlighting spatial heterogeneity in elephant distribution. Estimates of the space-usage parameter suggested that herd-adults ([Formula: see text] = 5.91 km, 95% CrI = 5.18-6.81) moved around considerably more than adult males ([Formula: see text] = 3.64 km, 95% CrI = 3.09-4.34). Based on elephant movement and age-sex composition, we derived the population that contributed individuals sampled in Kaziranga to be 908 herd-adults, 228 adult males and 610 young (density = 0.46 young/km2, SD = 0.06). Our study demonstrates how SCR is suited to estimating geographically open populations, characterising spatial heterogeneity in fine-scale density, and facilitating reliable monitoring to assess population status and dynamics for science and conservation.