An evolutionary programming postprocessor, using coevolution in a predator–prey ecosystem model, is developed and applied both to 72-h, 2-m temperature forecasts for the conterminous United States and southern Canada and to 60-min nowcasts of convection occurrence for the United States east of 94°W. The new approach improves deterministic and probabilistic forecasts of surface temperature relative to bias-corrected numerical weather prediction forecasts and to an earlier version of evolutionary programming forecasts for these same data. The new method also improves deterministic performance for an artificial neural network trained and evaluated for these same data. Additionally, the new approach substantially improves these forecasts’ reliability, as evidenced by reductions in the occurrence of excessive outliers in the rank histogram. The coevolutionary postprocessor also improves deterministic nowcasts of convection occurrence when compared to those produced by the National Weather Service’s AutoNowCaster system and to those obtained using multiple logistic regression. Notably, the degree of improvement relative to traditional methods appears to be problem dependent, while the training and implementation of such a system requires additional effort. However, the coevolutionary system is shown to be robust to imbalances between the frequency of positive and null events in the training data, unlike many postprocessing methods; to be implementable and effective in an adaptive mode, removing the need for retraining as inputs (such as numerical weather prediction model data) change; and to provide a useful, alternative perspective on the likelihood of event occurrence when used in combination with other methods.
AutoNowcaster (ANC) is an automated system that nowcasts thunderstorms, including thunderstorm initiation. However, its parameters have to be tuned to regional environments, a process that is time consuming, labor intensive, and quite subjective. When the National Weather Service decided to explore using ANC in forecast operations, a faster, less labor-intensive, and objective mechanism to tune the parameters for all the forecast offices was sought. In this paper, a genetic algorithm approach to tuning ANC is described. The process consisted of choosing datasets, employing an objective forecast verification technique, and devising a fitness function. ANC was modified to create nowcasts offline using weights iteratively generated by the genetic algorithm. The weights were generated by probabilistically combining weights with good fitness, leading to better and better weights as the tuning process proceeded. The nowcasts created by ANC using the automatically determined weights are compared with the nowcasts created by ANC using weights that were the result of manual tuning. It is shown that nowcasts created using the automatically tuned weights are as skilled as the ones created through manual tuning. In addition, automated tuning can be done in a fraction of the time that it takes experts to analyze the data and tune the weights.
Life history theory suggests that long‐lived, pond‐breeding amphibians should have low and highly variable early life‐stage survival rates, but this theoretical expectation is often untested and the causes of variation are usually unknown. We evaluated the impact of hydroperiod, presence of a pathogen (Batrachochytrium dendrobatidis [Bd]), presence of a potential predator (cutthroat trout Oncorhychus clarki stomias), and whether animals had been reintroduced into a site on survival of early life stages of boreal toads (Anaxyrus boreas boreas). We used a multistate mark‐recapture framework to estimate survival of boreal toad embryos from egg to metamorphosis at four sites over 5 years. We found substantial spatial and temporal variation in survival to metamorphosis and documented some evidence that monthly tadpole survival was lower in sites with Bd, without trout, and at permanent sites. Our results support theories of amphibian life history, aid in the management of this species of conservation concern, and contribute to our knowledge of the ecology of the species. Additionally, we present methodology that allows practitioners to account for different lengths of time between sampling periods when estimating survival probabilities which is especially applicable to organisms with distinct biological stages.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.