Hybrid forecasting: blending climate predictions with AI models

Hydrology and earth system sciences, Journal Year: 2023, Volume and Issue: 27(9), P. 1865 - 1889

Published: May 15, 2023

Abstract. Hybrid hydroclimatic forecasting systems employ data-driven (statistical or machine learning) methods to harness and integrate a broad variety of predictions from dynamical, physics-based models – such as numerical weather prediction, climate, land, hydrology, Earth system into final prediction product. They are recognized promising way enhancing the skill meteorological variables events, including rainfall, temperature, streamflow, floods, droughts, tropical cyclones, atmospheric rivers. now receiving growing attention due advances in climate at subseasonal decadal scales, better appreciation strengths AI, expanding access computational resources methods. Such attractive because they may avoid need run computationally expensive offline land model, can minimize effect biases that exist within dynamical outputs, benefit learning, learn large datasets, while combining different sources predictability with varying time horizons. Here we review recent developments hybrid outline key challenges opportunities for further research. These include obtaining physically explainable results, assimilating human influences novel data sources, integrating new ensemble techniques improve predictive skill, creating seamless schemes merge short long lead times, incorporating initial surface ocean/ice conditions, acknowledging spatial variability landscape forcing, increasing operational uptake schemes.

Language: Английский

---

On the Sensitivity of Standardized-Precipitation-Evapotranspiration and Aridity Indexes Using Alternative Potential Evapotranspiration Models

Hydrology, Journal Year: 2023, Volume and Issue: 10(3), P. 64 - 64

Published: March 6, 2023

This paper examines the impacts of three different potential evapotranspiration (PET) models on drought severity and frequencies indicated by standardized precipitation index (SPEI). The precipitation-evapotranspiration is a recent approach to operational monitoring analysis severity. combines temperature data, quantifying as difference in timestep between PET. thus represents hydrological processes that drive events more realistically than at expense additional computational complexity increased data demands. principally due need estimate PET within each time step. was originally defined using Thornthwaite model. However, numerous researchers have demonstrated sensitive model adopted. requiring sparse meteorological inputs, such model, particular utility for scarce environments. aridity (AI) investigates spatiotemporal changes hydroclimatic system. It ratio precipitation. used characterize wet (humid) dry (arid) regions. In this study, sensitivity indexes carried out models; namely, Penman–Monteith temperature-based parametric undertaken six gauge stations California region where long-term occurred. Having estimating index, our findings highlight presence uncertainty defining drought, especially large timescales (12 months 48 months), preferable both indexes. latter outcome worth further consideration when climatic studies are under development areas full required variables assessment not available.

Language: Английский

---

Optimizing machine learning for agricultural productivity: A novel approach with RScv and remote sensing data over Europe

Agricultural Systems, Journal Year: 2024, Volume and Issue: 218, P. 103955 - 103955

Published: April 29, 2024

Language: Английский

---

Artificial intelligence for climate prediction of extremes: State of the art, challenges, and future perspectives

Wiley Interdisciplinary Reviews Climate Change, Journal Year: 2024, Volume and Issue: unknown

Published: Sept. 3, 2024

Abstract Extreme events such as heat waves and cold spells, droughts, heavy rain, storms are particularly challenging to predict accurately due their rarity chaotic nature, because of model limitations. However, recent studies have shown that there might be systemic predictability is not being leveraged, whose exploitation could meet the need for reliable predictions aggregated extreme weather measures on timescales from weeks decades ahead. Recently, numerous been devoted use artificial intelligence (AI) study make climate predictions. AI techniques great potential improve prediction uncover links large‐scale local drivers. Machine deep learning explored enhance prediction, while causal discovery explainable tested our understanding processes underlying predictability. Hybrid combining AI, which can reveal unknown spatiotemporal connections data, with models provide theoretical foundation interpretability physical world, improving skills extremes climate‐relevant possible. challenges persist in various aspects, including data curation, uncertainty, generalizability, reproducibility methods, workflows. This review aims at overviewing achievements subseasonal decadal timescale. A few best practices identified increase trust these novel techniques, future perspectives envisaged further scientific development. article categorized under: Climate Models Modeling > Knowledge Generation The Social Status Change Science Decision Making

Language: Английский

---

STAT-LSTM: A multivariate spatiotemporal feature aggregation model for SPEI-based drought prediction

Earth Science Informatics, Journal Year: 2025, Volume and Issue: 18(3)

Published: Feb. 25, 2025

Language: Английский

---

Advances and gaps in the science and practice of impact‐based forecasting of droughts

Wiley Interdisciplinary Reviews Water, Journal Year: 2023, Volume and Issue: 11(2)

Published: Oct. 25, 2023

Abstract Advances in impact modeling and numerical weather forecasting have allowed accurate drought monitoring skilful forecasts that can drive decisions at the regional scale. State‐of‐the‐art early‐warning systems are currently based on statistical indicators, which do not account for dynamic vulnerabilities, hence neglect socio‐economic initiating actions. The transition from conventional physical of droughts toward impact‐based (IbF) is a recent paradigm shift early warning services, to ultimately bridge gap between science action. demand generate predictions “what will do” underpins rising interest IbF across all weather‐sensitive sectors. Despite large expected benefits, migrating this new presents myriad challenges. In article, we provide comprehensive overview IbF, outlining progress made field. Additionally, present road map highlighting current challenges limitations practice possible ways forward. We identify seven scientific practical challenges/limitations: contextual challenge (inadequate accounting spatio‐sectoral dynamics vulnerability exposure), human‐water feedbacks (neglecting how human activities influence propagation drought), typology (oversimplifying meteorological), model (reliance mainstream machine learning models), data (mainly textual) with linked sectoral geographical limitations. Our vision facilitate its use making informed timely mitigation measures, thus minimizing impacts globally. This article categorized under: Science Water > Extremes Methods Environmental Change

Language: Английский

---

Adapting subseasonal-to-seasonal (S2S) precipitation forecast at watersheds for hydrologic ensemble streamflow forecasting with a machine learning-based post-processing approach

Journal of Hydrology, Journal Year: 2024, Volume and Issue: 631, P. 130643 - 130643

Published: Jan. 22, 2024

Language: Английский

---

Significant relationships between drought indicators and impacts for the 2018–2019 drought in Germany

Environmental Research Letters, Journal Year: 2023, Volume and Issue: 19(1), P. 014037 - 014037

Published: Nov. 29, 2023

Abstract Despite the scientific progress in drought detection and forecasting, it remains challenging to accurately predict corresponding impact of a event. This is due complex relationships between (multiple) indicators adverse impacts across different places/hydroclimatic conditions, sectors, spatiotemporal scales. In this study, we explored these by analyzing severe 2018–2019 central European event Germany. We first computed standardized precipitation index (SPI), evaporation (SPEI), soil moisture (SSMI) streamflow (SSFI) over various accumulation periods, then related sectorial losses from report inventory (EDII) media sources. To cope with uncertainty associated both data, developed fuzzy method categorize them. Lastly, applied at region level (EU NUTS1) correlating monthly time series. Our findings revealed strong significant albeit some cases region-specific time-variant. Furthermore, our analysis established interconnectedness which displayed systematically co-occurring impacts. As such, work provides new framework explore indicators-impacts dependencies space, time, addition, emphasizes need leverage available data better forecast

Language: Английский

---

Characterizing drought prediction with deep learning: A literature review

MethodsX, Journal Year: 2024, Volume and Issue: 13, P. 102800 - 102800

Published: June 13, 2024

Drought prediction is a complex phenomenon that impacts human activities and the environment. For this reason, predicting its behavior crucial to mitigating such effects. Deep learning techniques are emerging as powerful tool for task. The main goal of work review state-of-the-art characterizing deep used in drought results suggest most widely climate indexes were Standardized Precipitation Index (SPI) Evapotranspiration (SPEI). Regarding multispectral index, Normalized Difference Vegetation (NDVI) indicator utilized. On other hand, countries with higher production scientific knowledge area located Asia Oceania; meanwhile, America Africa regions few publications. Concerning methods, Long-Short Term Memory network (LSTM) algorithm implemented task, either canonically or together (hybrid methods). In conclusion, reveals need more about using indices Africa; therefore, it an opportunity characterize developing countries.

Language: Английский

---

A New Multi-Objective Genetic Programming Model for Meteorological Drought Forecasting

Water, Journal Year: 2023, Volume and Issue: 15(20), P. 3602 - 3602

Published: Oct. 14, 2023

Drought forecasting is a vital task for sustainable development and water resource management. Emerging machine learning techniques could be used to develop precise drought models. However, they need explicit simple enough secure their implementation in practice. This article introduces novel model, called multi-objective multi-gene genetic programming (MOMGGP), meteorological that addresses both the accuracy simplicity of model applied. The proposed considers two objective functions: (i) root mean square error (ii) expressional complexity during its evolution. While former increase at training phase, latter assigned decrease achieve parsimony conditions. evolution verification procedure were demonstrated using standardized precipitation index obtained Burdur City, Turkey. comparison with benchmark (GP) (MGGP) models showed MOMGGP provides same more Thus, it suggested utilize practical forecasting.

Language: Английский

---