Noninvasive fecal DNA (fDNA)-based spatially explicit capture-recapture (SCR) approaches have proven valuable for estimating abundance of deer (Odocoileus spp) and other wildlife. The utility of SCR approaches for more gregarious species, such as elk (Cervus canadensis), is less clear because grouping behavior formally violates assumptions of statistical independence. We sought to understand the extent of spatial clustering in tule elk (C. c. nannodes) and to assess robustness of SCR estimates to such spatial dependence. Using GPS telemetry of 32 male and 34 female tule elk from 3 northern California populations during Jun–Aug 2017–19, we found strong clustering of female elk, but no detectable clustering by males. Concurrently, we conducted fDNA sampling, obtaining >1,000 genotypes of 425 sex-typed individuals, which we used for conventional CR analysis of the 66 physically captured individuals and for SCR based solely on the fDNA samples. Estimates of abundance from these approaches incorporating both sexes were statistically equivalent, although precision was considerably higher for SCR. Single-sex SCR estimates agreed with these estimates in 2 of the populations but differed in one population for which female spatial clustering was most extreme: in that case, the estimate of female abundance was much higher than in the two-sex model, the conventional CR model, and in an independent study of that population later that same year. We conclude that SCR of both sexes was robust to gregariousness in elk because inclusion of males was sufficient to offset spatial dependence stemming from aggregation of females.