A Remote Sensing Baseline and Time Sequence of Land Cover Change for the Conservation of Rainbowfish (Melanotaenia spp.) from the Bird’s Head Peninsula, Western New Guinea

Rainbowfish of the genus Melanotaenia are highly endemic freshwater fishes found only in Australia and New Guinea. Although widespread, most species have narrow geographic ranges, making them particularly vulnerable to environmental change. Currently, 43 described (and many undescribed) Melanotaenia species occur in the Bird’s Head and Bird’s Neck region of Western New Guinea, 29 of which are currently classified as critically endangered, endangered, or vulnerable by the IUCN Red List, including two that may be extinct in the wild. We generated a high-spatial-resolution baseline land cover classification of rainbowfish habitats using low-cloud Planet Labs quarterly basemap mosaics and compared it with a moderate-resolution Landsat 8 OLI-derived classification to assess how spatial resolution influences land cover classification. Using the full 40-year Landsat archive, we quantified decadal land cover change around species type localities and identified localized disturbance events that may affect rainbowfish habitats. For species described from large rivers and lakes, changes in water-body extent over time were quantified. Deforestation varied widely, ranging from little or no detectable change in remote, difficult-to-access locations (e.g., M. misoolensis, M. sneideri), to landscapes heavily modified by logging, urbanization, mining, and agriculture (e.g., M. boesemani, M. arfakensis). Around the type localities, from the high-resolution imagery, we detected ~2939 ha of cleared land, whereas from the Landsat classification we identified only 31 ha of clearing, indicating that most of the fine-scale deforestation was not resolved at the Landsat scale. Time-sequence analyses indicate that over one-third of type localities experienced one or more localized disturbance events over the last 40 years. Land cover change in this region is highly dynamic and differs from commonly studied frontier deforestation patterns elsewhere. It also underscores a critical conservation challenge where rainbowfish species are being discovered in landscapes that are simultaneously undergoing rapid, spatially heterogeneous change. The same infrastructure that enables biological exploration also accelerates habitat modification. These changes threaten the persistence of highly endemic rainbowfish and underscore the value of multi-scale spatial and temporal remote sensing approaches for assessing habitat change in remote, biodiverse regions. The framework presented here is also broadly applicable to other narrowly distributed endemic taxa.