Man-made barriers impede fish movement up and downstream, between the channel and floodplain, and from marginal habitats to those that promote survival and reproductive output. Barriers also increase genetic isolation resulting in loss of genetic variation. Species with drifting eggs and larvae, like the endangered Rio Grande silvery minnow (Hybognathus amarus), are especially susceptible to disruptions to demographic and genetic connectivity. Using archived Rio Grande silvery minnow DNA from 25 temporal collections spanning 38 years, and targeted amplicon sequencing, we show that unidirectional stream flow and longitudinal position of populations affect patterns of N_e and genetic diversity. N_e is reduced in the upstream-most reach (Angostura) and there is a strong correlation between upstream and range-wide N_e suggesting impacts to the entire population. In the absence of population augmentation, allelic diversity was reduced upstream but stocking restored diversity. Long-term genetic analysis indicates: (1) that there is no longer sufficient spawning and rearing to maintain natural diversity in the Angostura reach; (2) upstream movement of adults is insufficient to recover diversity; and/or (3) there is higher variance in reproductive success in resident fish resulting in small values of N_e; all consequences of a highly engineered and disconnected river corridor. Similar impacts are predicted for species with pelagic life histories in altered rivers, underscoring the need for fish passage structures that facilitate connectivity and habitat restoration aimed toward propagule retention in upstream reaches. Long-term genetic data also show periodic recruitment bottlenecks compound losses to genetic diversity imposed by river regulation that cannot be restored by hatchery stocking alone.

A total of 2,687 wild-collected samples from 25 temporal collections spanning 38 years were used, including a subset of archived DNA samples (~96 samples per year [32 samples per reach] when available; Table 1). Sampling and DNA isolation methods are described in Osborne et al. (2012, 2023) and a brief description is also provided in the supplementary material. Libraries were prepared following the method of Campbell et al. (2015) with two modifications described in Caeiro-Dias et al. (in review). The GT-seq panel contains 283 neutral loci comprising 352 SNPs distributed across Rio Grande silvery minnow genome. We refer to genotypes at these loci as microhaplotypes because a locus may contain more than one variable SNP. The sex-specific marker HAM6 described in Caeiro-Dias et al. (2023) was also included in the development of the GT-seq panel (Caeiro-Dias et al. in review).

Microhaplotypes were identified with GTscore pipeline v. 1.3 (https://github.com/gjmckinney/GTscore). Briefly, the AmpliconRadCounter.pl script was used to count the number of unique reads per individual, to identify on-target reads by aligning the locus forward primer and the in-silico probe for each SNP allele to each unique read and to count the number of reads containing each SNP allele for every individual. Primer and probe sequences are provided in Caeiro-Dias et al. (in prep). Then the matrix with counts of reads containing a SNP allele for each individual outputted by AmpliconRadCounter.pl script was used for microhaplotype genotyping with the maximum likelihood algorithm described in McKinney et al. (2018) and implemented in GTscore.R script. Individuals with more than 30 % of missing data were excluded. In a few cases, particularly in older samples, library preparation and sequencing were repeated to increase coverage until missing data per locus was smaller than 30 %.