Dataset for Characterizing non-permanent aquatic habitats using a Bayesian-curated remote sensing approach: insights from a rotifer case study

Remote sensing is an indispensable tool for generating consistent, standardized time-series data across broad spatial scales. However, its usefulness for ecological and evolutionary research depends on how well it captures temporally relevant environmental variation. Non-permanent aquatic systems, such as Mediterranean shallow water bodies, are particularly challenging because their characteristic abrupt wet-dry transitions occurring over short timescales increase noise and misclassification of hydrological states. Organisms inhabiting these systems are tightly coupled to the timing, duration and predictability of inundation phases. Here, we developed a Bayesian state-space approach to curate Sentinel-2 time-series data, improving the reliability of hydrological state detection in a system of shallow ponds in eastern Iberian Peninsula. The curated hydrological-state data enabled the quantification of non-permanent regimes using ecologically meaningful metrics, revealing substantial differences in water availability and predictability among neighbouring ponds under similar climatic conditions. We further demonstrate how these metrics are relevant to evolutionary ecology by linking hydrological unpredictability to bet-hedging strategies in rotifer populations. Rotifer clones originating from more unpredictable ponds exhibited lower hatching fractions, which is consistent with an adaptive risk-spreading strategy. This study highlights the importance of characterizing environmental variability at temporal scales directly relevant to organismal life cycles when linking habitat dynamics to evolutionary responses.