Seasonal prediction of Horn of Africa long rains using machine learning: The pitfalls of preselecting correlated predictors

Deman, Victoria; Koppa, Akash; Waegeman, Willem; MacLeod, David; Singer, Michael Bliss; Miralles, Diego G.

doi:10.3389/frwa.2022.1053020

Cited by 8 publications

(4 citation statements)

References 54 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…At seasonal and subseasonal scales, the source regions derived with the Lagrangian analysis have the potential to improve the predictability of the long and short rains through both better selection of predictors (Deman et al., 2022) and improving the understanding of their drivers. Therefore, our results can be used to augment drought and flood early warning systems, which are important tools in the region to prepare appropriate mitigation measures (C. Funk et al., 2019).…”

Section: Discussionmentioning

confidence: 99%

“…This study represents a step toward improving our understanding of the drivers of the recent and future interannual changes in the long and short rains using Lagrangian analysis. Furthermore, accurately delineating the source regions of moisture can aid in a more efficient selection of rainfall predictors for seasonal forecasting, and thus may potentially enable more accurate forecasts in the region (Deman et al., 2022).…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

A Lagrangian Analysis of the Sources of Rainfall Over the Horn of Africa Drylands

et al. 2023

Self Cite

View full text Add to dashboard Cite

The Horn of Africa drylands (HAD) are among the most vulnerable regions to hydroclimatic extremes. The two rainfall seasons—long and short rains—exhibit high intraseasonal and interannual variability. Accurately simulating the long and short rains has proven to be a significant challenge for the current generation of weather and climate models, revealing key gaps in our understanding of the drivers of rainfall in the region. In contrast to existing climate modeling and observation‐based studies, here we analyze the HAD rainfall from an observationally‐constrained Lagrangian perspective. We quantify and map the region's major oceanic and terrestrial sources of moisture. Specifically, our results show that the Arabian Sea (through its influence on the northeast monsoon circulation) and the southern Indian Ocean (via the Somali low‐level jet) contribute ∼80% of the HAD rainfall. We see that moisture contributions from land sources are very low at the beginning of each season, but supply up to ∼20% from the second month onwards, that is, when the oceanic‐origin rainfall has already increased water availability over land. Further, our findings suggest that the interannual variability in the long and short rains is driven by changes in circulation patterns and regional thermodynamic processes rather than changes in ocean evaporation. Our results can be used to better evaluate, and potentially improve, numerical weather prediction and climate models, and have important implications for (sub‐)seasonal forecasts and long‐term projections of the HAD rainfall.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

A Lagrangian Analysis of the Sources of Rainfall Over the Horn of Africa Drylands

et al. 2023

Self Cite

View full text Add to dashboard Cite

show abstract

“…To tackle this issue, further research needs to be conducted to enhance the accuracy of Long rains forecasts. This could involve the integration of new statistical approaches, ensemble member selection (Heinrich-Mertsching et al, 2023b), ongoing research on exploring the utilization of machine learning techniques (Deman et al, 2022) and hybrid methods that take into consideration the different forecasting approaches.…”

Section: Research Modeling and Predictionmentioning

confidence: 99%

Advances, gaps and way forward in provision of climate services over the Greater Horn of Africa

Gudoshava,

Otieno,

Koech

et al. 2024

Front. Clim.

View full text Add to dashboard Cite

The Greater Horn of Africa is prone to extreme climatic conditions, thus, making climate services increasingly important in supporting decision-making processes across a range of climate sensitive sectors. This study aims to provide a comprehensive review of the recent advances, gaps and challenges in the provision of climate services over the region, for each of the components of the Global Framework for Climate Services. The study explores various milestones that have been achieved toward climate service delivery. The achievements include improvement of station network coverage, and enhancing the capacity of member states to utilize various tools in data analysis and generate routine climate products. The advancement in science, and availability of High-Performance Computing has made it possible for forecast information to be provided from nowcasting to seasonal timescales. Moreover, operationalizing of the objective forecasting method for monthly and seasonal forecasts has made it possible to translate tercile forecasts for applications models. Additionally, innovative approaches to user engagement through co-production, communication channels, user-friendly interfaces, and dissemination of climate information have also been developed. Despite the significant progress that has been made in the provision of climate services, there are still many challenges and gaps that need to be overcome in order to ensure that these services are effectively meeting the needs of users. The research of the science underpinning climate variability, capacity building and stakeholder engagement, as well as improved data management and quality control processes are some of the gaps that exist over the region. Additionally, communication and dissemination of climate information, including timely warnings and risk communication, require improvement to reach diverse user groups effectively. Addressing these challenges will require strengthened partnerships, increased investment in capacity building, enhanced collaboration between the climate information producers and stakeholders, and the development of user-friendly climate products. Bridging these gaps will foster greater resilience to climate-related hazards and disasters in the Greater Horn of Africa and support sustainable development in the region.

show abstract

“…These problem classes simplify and help abstract theoretical and practical problems to the computational field where ML techniques can act. For example, among the algorithms available to solve regression and classification problems, artificial neural networks (ANNs), deep learning (DL), support vector machines (SVMs), k-nearest neighbors (kNNs), decision trees (DTs), and random forests (RFs) stand out (Balaji et al, 2021;Yu and Haskins, 2021;Deman and Miralles, 2022). Clustering issues can be solved using algorithms such as k-means, hierarchical cluster analysis (HCA), and density-based spatial clustering of applications with noise (DBSCAN) (Tang et al, 2022;Manoj Stanislaus et al, 2023).…”

Section: ML Pipeline For Rainfall Forecastingmentioning

confidence: 99%

Precipitation forecasting: from geophysical aspects to machine learning applications

Oliveira,

Nogueira Neto,

Santos

et al. 2023

Front. Clim.

View full text Add to dashboard Cite

Intense precipitation events pose a significant threat to human life. Mathematical and computational models have been developed to simulate atmospheric dynamics to predict and understand these climates and weather events. However, recent advancements in artificial intelligence (AI) algorithms, particularly in machine learning (ML) techniques, coupled with increasing computer processing power and meteorological data availability, have enabled the development of more cost-effective and robust computational models that are capable of predicting precipitation types and aiding decision-making to mitigate damage. In this paper, we provide a comprehensive overview of the state-of-the-art in predicting precipitation events, addressing issues and foundations, physical origins of rainfall, potential use of AI as a predictive tool for forecasting, and computational challenges in this area of research. Through this review, we aim to contribute to a deeper understanding of precipitation formation and forecasting aided by ML algorithms.

show abstract

Seasonal prediction of Horn of Africa long rains using machine learning: The pitfalls of preselecting correlated predictors

Cited by 8 publications

References 54 publications

A Lagrangian Analysis of the Sources of Rainfall Over the Horn of Africa Drylands

A Lagrangian Analysis of the Sources of Rainfall Over the Horn of Africa Drylands

Advances, gaps and way forward in provision of climate services over the Greater Horn of Africa

Precipitation forecasting: from geophysical aspects to machine learning applications

Contact Info

Product

Resources

About