Aquatic ecosystems are in peril. World Heritage estates such as Kakadu National Park are not immune to anthropogenic pressures. To better protect freshwater fish communities and obtain relevant ecological data, there has been a shift towards environmental (e)DNA metabarcoding. In tropical lentic environments, seasonal variation drives significant environmental change, particularly the resuspension of particulate matter and phytoplankton productivity. These factors constrain the volume of water that can be sampled, potentially affecting the detection capacity of eDNA metabarcoding. Hence, information about the volume of samples required to robustly detect fish is critical prior to developing metabarcoding biomonitoring programs. In Kakadu, we sampled two contrasting systems with distinct fish communities. Employing two mt-rRNA primers (12S and 16S), we examined the detection of fish sampled from 1, 2, 5, and 10 L. Leveraging MinION sequencing, 15 fish taxa were confidently assigned to the species-level. Despite both systems containing unique fish communities (perMANOVA: F = 5.73, P < 0.001), the most salient finding was that fish richness and community structure were not influenced by sample volume. In tropical systems where volume is inhibited, sampling a consistently lower volume may still be suitable for fish detection. Several species rarely observed using traditional monitoring approaches were detected, including Porochilus rendahli and Oxyeleotris lineolata. Our findings emphasise the need for multiple primers to capture a wider diversity of fish. This study provides the foundations for a routine eDNA monitoring program optimised for Australian tropical lentic freshwaters, which may assist in the management of tropical fish communities.