The impact of resource exploitation by ancient human communities on Madagascar’s environment is an area of intense debate. A fundamental question in the archaeology of Madagascar is the extent to which arrival of settlers, introduction of non-native plants and animals, and subsequent human exploitation of island biota, which catalyzed declines in biodiversity and significantly degraded environmental conditions. Fine-grained datasets, including zooarchaeological, archaeobotanical and other ecological evidence, are needed to assess the relationship between human resource exploitation and environmental change. On Madagascar, the resolution of zooarchaeological datasets is often reduced by poor preservation of faunal remains, making precise taxonomic identifications difficult, and few projects to-date have comprehensively assessed zooarchaeological data. Here, we present zooarchaeological data from three coastal villages in the Velondriake Marine Protected Area in southwest Madagascar, where human occupation spans from ca. 1400 BP to the present. Faunal remains from the Late Holocene sites of Antsaragnagnangy and Antsaragnasoa were identified using morphological analysis of remains, and a PCR-based bulk bone metabarcoding approach was applied at Andamotibe to molecularly identify fish and other vertebrates in a faunal assemblage that was particularly fragmented. Results were interpreted and contextualized using modern data on local fish diversity, climate and anthropogenic impacts on marine and estuarine habitats, as well as modern fishing practices (including preferred fishing grounds, tackle, taxonomic representation and volume of catch). Our use of multiple analytical and interpretative approaches has provided the most highly resolved view to date of past human subsistence in coastal southwest Madagascar. We contend that future research into human-environment dynamics on Madagascar should make use of diverse analytical methods, in order to more comprehensively evaluate past interactions between human communities and the native biota. Furthermore, we encourage an historical ecological approach, so that long-term perspectives on changing human-environment dynamics may be used to contextualize modern trends.