Interpretable Deep Learning Approach Reveals the Critical Role of Sea Surface Salinity in Long-Lead ENSO Forecasting

Recent advancements in climate modeling have uncovered the pivotal role of Sea Surface Salinity (SSS) in enhancing long-lead ENSO forecasting. Utilizing a deep learning model have extended the ENSO forecast lead time to 24 months, emphasizing post-2000 El Niño events.
Like

Share this post

Choose a social network to share with, or copy the URL to share elsewhere

This is a representation of how your post may appear on social media. The actual post will vary between social networks

       Recent advancements in climate modeling have taken a significant leap forward with a new study highlighting the crucial role of Sea Surface Salinity (SSS) in improving the long-lead forecasting of the El Niño-Southern Oscillation (ENSO). This groundbreaking research, conducted by a team of scientists from the Institute of Oceanology, Chinese Academy of Sciences, and Duke University, introduces a novel deep learning (DL) model that successfully extends the effective ENSO forecast lead time to 24 months, with a particular focus on post-2000 El Niño events.

       ENSO, a climate phenomenon characterized by periodic fluctuations in sea surface temperatures (SST) across the equatorial Pacific, has long been a critical factor in global climate variability, impacting weather patterns, agriculture, and ecosystems worldwide. Despite decades of research, predicting ENSO events with long lead times has remained a significant challenge, particularly due to the influence of the Spring Predictability Barrier (SPB), which limits the forecast skill during and after boreal spring.

       The new study addresses these challenges by incorporating SSS data into their DL model, named the Spatio-Temporal Pyramid Network (STPNet). The model uses a multiscale-pyramid structure to capture diverse spatiotemporal features, combining SST and SSS data to enhance ENSO forecast accuracy. The inclusion of SSS, which influences ocean stratification and heat redistribution, allows the model to maintain a high correlation between predicted and observed ENSO indices, even for long-lead forecasts starting from the challenging spring months.

Figure. 1. The structure and performance of the STPNet ENSO forecast model. (a) Structure of the STPNet. The STPNet uses a spatial and temporal feature extraction module to extract the corresponding independent features, fused through a feature fusion block to obtain the joint spatiotemporal features. The predicted Niño3.4 index is obtained through the global average pooling (GAP) and fully connected layer. The model uses the global monthly anomalies of SSTA and SSSA for the current and previous 23 months to predict the Niño3.4 index for the next 24 months. (b) The ENSO prediction skill in terms of model-observation correlation during 2002-2021 for STPNet with the input variables of SSSA and SSTA (solid blue line) and the input variable of SSTA alone (solid red line). For comparison, the ENSO prediction skill for the CNN model (solid green line) and the ResCNN model (solid orange line) with input variables of SSTA and SSSA is also shown.

       One of the key findings of the study is the differential importance of SST and SSS in ENSO forecasting. While SST remains critical for short-to-medium lead forecasts (less than one year), SSS becomes increasingly important for medium-to-long lead forecasts (greater than six months). This discovery underscores the potential of SSS as a valuable predictor in climate models, especially with the growing availability of satellite-based SSS observations.

Figure. 2. The role of ocean inter-basin and tropical-extratropical interactions in ENSO forecast as revealed by the STPNet model. (a) Hovmöller diagrams showing the zonal propagation of the ENSO-related SST signals averaged between 10°S and 10°N. (b) Hovmöller diagrams showing the meridional propagation of the ENSO-related SST signals averaged at all longitudes. (c) Same as panel a but for  the ENSO-related SSS signals. (d) Same as panel b but for the ENSO-related SSS signals. (e,f) The ENSO prediction skill in terms of model-observation correlation during 2002-2021 for STPNet using the input variables of SST and SSS in different ocean basins and (f) different latitudinal bands. PO: Pacific Ocean; AO: Atlantic Ocean; IO: Indian Ocean; Tropical: Tropical Ocean; NH: Northern Hemisphere extratropical ocean; SH: Southern Hemisphere extratropical ocean.

       The research also explores the role of ocean inter-basin interactions, such as those between the Pacific, Indian, and Atlantic Oceans, in ENSO development. The STPNet model reveals that these interactions, along with tropical-extratropical dynamics, play a significant role in the predictability of ENSO events, providing new insights into the complex processes driving this climate phenomenon.

       As the availability of high-resolution SSS data continues to improve, the study's findings offer a promising avenue for enhancing operational ENSO forecasts, potentially leading to better preparedness for the global impacts of El Niño and La Niña events. The researchers plan to further refine their model by incorporating subsurface temperature and salinity data, which they believe will further improve long-lead ENSO predictions.

       This research represents a significant step forward in climate science, demonstrating the power of integrating advanced deep learning techniques with comprehensive oceanic data to tackle one of the most challenging aspects of climate prediction. The implications of this work extend beyond academia, offering practical benefits for sectors reliant on accurate climate forecasting, from agriculture to disaster management.

About the Authors:

Haoyu Wang is a Ph.D. student in Physical Oceanography at the University of Chinese Academy of Sciences. His research interests include ENSO forecasting and estimation of oceanic subsurface variables using remote sensing and artificial intelligence techniques.

For more information on this study, please contact [lixf@qdio.ac.cn].

 

Please sign in or register for FREE

If you are a registered user on Research Communities by Springer Nature, please sign in

Follow the Topic

Ocean Sciences
Physical Sciences > Earth and Environmental Sciences > Earth Sciences > Ocean Sciences
Climate Sciences
Physical Sciences > Earth and Environmental Sciences > Earth Sciences > Climate Sciences
Artificial Intelligence
Mathematics and Computing > Computer Science > Artificial Intelligence

Related Collections

With collections, you can get published faster and increase your visibility.

Tropical Cyclones in a Warming World

This Collection aims to address the complex dynamics between tropical cyclones and climate change, exploring how these interactions shape the characteristics and behaviors of tropical cyclones in a warming world.

Publishing Model: Open Access

Deadline: Dec 31, 2024

Secondary Organic Aerosols in Polluted Atmospheres

Investigating secondary organic aerosols in different environment by field observations, laboratory studies, and modeling approaches.

Publishing Model: Open Access

Deadline: Dec 31, 2024