What can we learn from 100,000 freshwater forecasts? A synthesis from the NEON Ecological Forecasting Challenge

Ecological Applications, Journal Year: 2025, Volume and Issue: 35(1)

Published: Jan. 1, 2025

Abstract Near‐term, iterative ecological forecasts can be used to help understand and proactively manage ecosystems. To date, more have been developed for aquatic ecosystems than other worldwide, likely motivated by the pressing need conserve these essential threatened increasing availability of high‐frequency data. Forecasters implemented many different modeling approaches forecast freshwater variables, which demonstrated promise at individual sites. However, a comprehensive analysis performance varying models across multiple sites is needed broader controls on performance. Forecasting challenges (i.e., community‐scale efforts generate while also developing shared software, training materials, best practices) present useful platform bridging this gap evaluate how range methods perform axes space, time, systems. Here, we analyzed from aquatics theme National Ecological Observatory Network (NEON) Challenge hosted Initiative. Over 100,000 probabilistic water temperature dissolved oxygen concentration 1–30 days ahead seven NEON‐monitored lakes were submitted in 2023. We assessed varied among with structures, covariates, sources uncertainty relative baseline null models. A similar proportion skillful both variables (34%–40%), although outperformed forecasting (10 out 29) (6 15). These top performing came classes structures. For temperature, found that skill degraded increases horizons, process‐based models, included air as covariate generally exhibited highest performance, most often accounted lower The where observations divergent historical conditions (resulting poor model performance). Overall, NEON provides an exciting opportunity intercomparison learn about strengths diverse suite advance our understanding ecosystem predictability.

Language: Английский

---

Interval combination prediction of mine tunnel air volume using improved spiking neural network for error correction

International Journal of Mining Reclamation and Environment, Journal Year: 2025, Volume and Issue: unknown, P. 1 - 25

Published: Feb. 28, 2025

Language: Английский

---

Investor sentiment and stock returns: Wisdom of crowds or power of words? Evidence from Seeking Alpha and Wall Street Journal

Journal of Financial Markets, Journal Year: 2025, Volume and Issue: unknown, P. 100970 - 100970

Published: March 1, 2025

Language: Английский

---

A multi-model approach to estimate excess mortality in the Nordics, 2020–2023

Public Health, Journal Year: 2025, Volume and Issue: 242, P. 131 - 138

Published: March 9, 2025

Excess mortality has been one of the commonly used measures population health effects COVID-19 pandemic. Denmark, Finland, Norway and Sweden share several socioeconomic characteristics but adopted different control experienced varying degrees case hospitalization burden during Using trends between 2001 2019 a combination models, we estimated compared annual monthly excess in these countries nationally as well stratified by age, sex subnational regions 2020 2023. Multi-model study. Three methods were to estimate mortality: i) Bayesian spatial model with random effect component for dependence among subregions trend seasonality terms; ii) GAMM terms (a thin-plate spline) within-year cyclic cubic spline); and, iii) autoregressive exponential smoothing methods. Estimates from approaches combined using averaging. Based on age-standardized rates (per 100,000 population) averaged estimates: Finland had highest cumulative four countries; older age groups (70+ year) accounted nearly all mortality; men higher than women; capital some lowest relative other each country. With few exceptions, 2023 returned pre-pandemic levels. Model verification indicated good calibration superior skill its models. We believe our approach better quantifies uncertainty individual estimates are comprehensive, spatially, temporally demographically well-resolved, can support further association studies.

Language: Английский

---

Comparative analysis and evaluation of ageing forecasting methods for semiconductor devices in online health monitoring

Engineering Applications of Artificial Intelligence, Journal Year: 2025, Volume and Issue: 150, P. 110545 - 110545

Published: March 20, 2025

Language: Английский

---