Thu, Aug 18, 2022: 4:00 PM-4:15 PM
512E
Background/Question/MethodsAll organisms experience fluctuations in environmental conditions that affect their population dynamics through time and their distribution in space. We can retrospectively explain how a particular sequence of environmental conditions led to observed changes in population size, composition, or persistence. But prospectively, the sequence of environments is fundamentally uncertain and may differ systematically from observations that were made in a different place or time. As such, our ability to accurately forecast population dynamics and fates is severely limited. We sought to improve population forecasts by characterizing climate-demography relationships from our long-term, range-wide demographic censuses of the dioecious perennial plant, Valeriana edulis. We then predicted (i) the current spatial distribution of populations and (ii) near-term change in population composition by pairing these climate-demography relationships with remotely-sensed microclimate data and downscaled climate projection model projections for growing season temperature, precipitation, and date of snow melt. Finally, we quantified the forecast skill of our models by comparing projections with ground-truth measurements of the current spatial distribution and composition of populations.
Results/ConclusionsWe found that our population forecasts based on climate-demography relationships generally had higher skill than null models in both space and time. The impact of the climate drivers differed across the life cycle and among demographic vital rates. Aggregated together, these yielded a slight decline in per-capita population growth (λ) with increasing growing season temperature, a strong decline with increasing precipitation, and a moderate increase with later snow melt date. In space, our calibrated population distribution predictions had modest sensitivity and high specificity, resulting in a moderately positive True Skill Statistic of 0.345. We believe that improving the spatial resolution of our climate data could substantially improve this score. In time, population composition (operational sex ratios) varied strongly among years in some populations and very little in others. Highly stable operational sex ratios were fundamentally easier to predict forecast, but gave little opportunity for climate-demography relationships to improve forecast skill. We argue that embracing an iterative near-term forecasting approach will offer opportunities to rapidly improve our ability to forecast population dynamics.
Results/ConclusionsWe found that our population forecasts based on climate-demography relationships generally had higher skill than null models in both space and time. The impact of the climate drivers differed across the life cycle and among demographic vital rates. Aggregated together, these yielded a slight decline in per-capita population growth (λ) with increasing growing season temperature, a strong decline with increasing precipitation, and a moderate increase with later snow melt date. In space, our calibrated population distribution predictions had modest sensitivity and high specificity, resulting in a moderately positive True Skill Statistic of 0.345. We believe that improving the spatial resolution of our climate data could substantially improve this score. In time, population composition (operational sex ratios) varied strongly among years in some populations and very little in others. Highly stable operational sex ratios were fundamentally easier to predict forecast, but gave little opportunity for climate-demography relationships to improve forecast skill. We argue that embracing an iterative near-term forecasting approach will offer opportunities to rapidly improve our ability to forecast population dynamics.