Mapping essential habitat of estuarine fishery species with a mechanistic SDM and Landsat data

Abstract

The identification and monitoring of essential fishery habitat (EFH) is primary and important for the conservation of estuarine fishery habitat and sustainable fisheries management. However, mapping EFH of estuarine fishery species with generally statistical species distribution model (SDM) remains difficult due to their complex life histories and the extensive spatiotemporal variability of habitat variables. Thus, we proposed a mechanistic SDM by analyzing a more direct and certain relationship at a pre-selected suitable life stage with prior knowledge of species life history, not a rough statistical analysis of overlapping cause–effect relationships operating across multiple life history stages in reverse. Landsat data was used to describe the extensive spatiotemporal variability of estuarine habitat variables. A window filtration was contributed most to select the suitable life stage with a score scheme. As mapping EFH needs to feasibly locate the representative and essential habitats over large estuarine scale, we constructed two indicators to represent “habitat importance” and “descriptor predictability” in the score scheme, which was semi-quantified with three primary factors (habitat function, duration of life history stage, inversion limitation of critical descriptors from Landsat data). Through selecting the suitable life stage for the mechanistic SDM, the usage of Landsat archives could be extended, which is beneficial from a few optically / thermally active determinants. The applicability of the mechanistic SDM was tested by mapping EFH of Eriocheir sinensis in the Yangtze River Estuary (YRE), China. The window filtration of E. sinensis’ life history stages in the YRE showed that the berry stage was filtered as the suitable stage with the highest scores both in habitat importance and descriptors predictability. Then the distribution of E. sinensis was modelled, predicated, and assessed at berry stage by comparing outward in a generally statistical way and inward with other filter window. The mechanistic model at berry stage showed the best explanatory and predictive power in both occurrence and abundance models, validating that the optimized SDM had much stronger capability in model fit and prediction. The score scheme was also verified to set reasonably and effectively through the model comparisons at different filter window stages. With the predicated distribution of E. sinensis, the core ranges of EFH were mapped at berry stage, which were mostly distributed in sub-tidal areas neighbored to wetlands, showing the habitat requirement of shallow water for hiding. Meantime, the determinants’ variance partitioning, in which dissolved oxygen (DO) and total suspended sediment (TSS) exceeded over 65%, complementally illustrated that core EFH may also require high water quality. The identification of EFH locations with our mechanistic SDM would provide spatial-explicit information and guidance to fishery habitat protection and fishery resources maintenance. Additionally, this study also explored a new way to utilize a few predictors from the extensively used Landsat data in an alternative way. This study could extend the limitations of SDM in EFH mapping and bridge gaps between spatiotemporal explicit modelling and remote sensing data accommodations, which is meaningful for fishery habitat conservation.