Volume 49, Issue 14 e2022GL098663

Research Letter

Open Access

Quantifying the Effect of Climate Change on Midlatitude Subseasonal Prediction Skill Provided by the Tropics

Kirsten J. Mayer,

Corresponding Author

Kirsten J. Mayer

[email protected]

orcid.org/0000-0003-3582-8337

Department of Atmospheric Science, Colorado State University, Fort Collins, CO, USA

Correspondence to:

K. J. Mayer,

[email protected]

Contribution: Conceptualization, Methodology, Formal analysis, Investigation, Data curation, Writing - original draft, Writing - review & editing, Visualization, Funding acquisition

Search for more papers by this author

Elizabeth A. Barnes,

Elizabeth A. Barnes

orcid.org/0000-0003-4284-9320

Department of Atmospheric Science, Colorado State University, Fort Collins, CO, USA

Contribution: Investigation, Writing - review & editing, Supervision, Funding acquisition

Search for more papers by this author

Kirsten J. Mayer,

Corresponding Author

Kirsten J. Mayer

[email protected]

orcid.org/0000-0003-3582-8337

Department of Atmospheric Science, Colorado State University, Fort Collins, CO, USA

Correspondence to:

K. J. Mayer,

[email protected]

Contribution: Conceptualization, Methodology, Formal analysis, Investigation, Data curation, Writing - original draft, Writing - review & editing, Visualization, Funding acquisition

Search for more papers by this author

Elizabeth A. Barnes,

Elizabeth A. Barnes

orcid.org/0000-0003-4284-9320

Department of Atmospheric Science, Colorado State University, Fort Collins, CO, USA

Contribution: Investigation, Writing - review & editing, Supervision, Funding acquisition

Search for more papers by this author

First published: 15 June 2022

https://doi.org/10.1029/2022GL098663

Share a link

Email
Facebook
Twitter
LinkedIn
Reddit
Wechat

Abstract

Subseasonal timescales (∼2 weeks–2 months) are known for their lack of predictability, however, specific Earth system states known to have a strong influence on these timescales can be harnessed to improve prediction skill (known as “forecasts of opportunity”). As the climate continues warming, it is hypothesized these states may change and consequently, their importance for subseasonal prediction may also be impacted. Here, we examine changes to midlatitude subseasonal prediction skill provided by the tropics under anthropogenic warming using artificial neural networks to quantify skill. The network is tasked to predict the sign of the 500 hPa geopotential height for historical and future time periods in the Community Earth System Model Version 2 - Large Ensemble across the Northern Hemisphere at a 3 week lead using tropical precipitation. We show prediction skill changes substantially in key midlatitude regions and these changes appear linked to changes in seasonal variability with the largest differences in accuracy occurring during forecasts of opportunity.

Key Points

Neural networks can be used to evaluate subseasonal predictability under future climate change scenarios
In community earth system model version 2 - Large ensemble (CESM2-LE), largest differences in subseasonal predictability provided by the tropics mainly occur during forecasts of opportunity
Changes in Northern Hemisphere subseasonal prediction skill appear mainly linked to changes to seasonal variability in CESM2-LE

Plain Language Summary

Predictions on 2 weeks to 2 months (subseasonal) timescales are important for the public and private sectors due to the increased preparation time provided to save lives and property. In the current climate, signals initiated in the tropics can overpower noise in the midlatitudes and ultimately lead to enhanced midlatitude subseasonal prediction skill. However, it has been hypothesized that increasing global temperatures due to climate change may impact these signals and their sources in the future. Therefore, it is important to understand how subseasonal predictability provided by the tropics will be affected. Here, we utilize a type of machine learning known as a neural network to investigate this question. We find that subseasonal prediction skill provided by the tropics changes throughout the Northern Hemisphere in a warmer climate and these changes appear mainly linked to changes in seasonal variability. In addition, we see that the largest differences in accuracy occur during opportunities for enhanced subseasonal prediction skill.

1 Introduction

Accurate predictions on subseasonal timescales (2 weeks–2 months) are important for many public and private sectors such as water management and agriculture (White et al., 2022). This is because prediction on these timescales provides pivotal lead times for saving lives and property in these sectors (White et al., 2022). The tropics is of particular importance for this timescale because of intraseasonal phenomena like the Madden-Julian Oscillation (MJO; Madden & Julian, 1971, 1972). Quasi-stationary Rossby waves generated by upper level divergence associated with MJO convection (Hoskins & Ambrizzi, 1993) can modulate midlatitude circulation in the following weeks (e.g., Baggett et al., 2017; Henderson et al., 2016; Hoskins & Karoly, 1981; Sardeshmukh & Hoskins, 1988; Zheng et al., 2018) and these tropical-extratropical teleconnections are known to lead to enhanced midlatitude prediction skill on subseasonal lead times (Tseng et al., 2018). Phenomena like the El Niño Southern Oscillation (ENSO), an interannual oceanic mode in the tropical Pacific Ocean (Trenberth, 1997), can also impact subseasonal prediction. It can do so through modulation of the MJO (e.g., Hendon et al., 1999; Kessler, 2001; Pohl & Matthews, 2007) or modulation of the large-scale background state (e.g., Moon et al., 2011; Namias, 1986; Takahashi & Shirooka, 2014), and both can ultimately impact teleconnection propagation (e.g., Arcodia et al., 2020; Henderson & Maloney, 2018; Stan et al., 2017; Tseng et al., 2020) and subseasonal prediction skill (e.g., Johnson et al., 2014; L. Wang & Robertson, 2019). Therefore, when phenomena like the MJO and ENSO are present, they can provide a predictable signal above climate noise and be used to enhance subseasonal prediction skill, known as forecasts of opportunity (Mariotti et al., 2020).

The current understanding of the importance of the tropics on midlatitude subseasonal predictability is rooted in our knowledge of the historical climate. However, with the climate continuously warming, it is unclear how transferable this knowledge will be to a future, warmer climate. Therefore, research on subseasonal timescales has examined how the MJO (Maloney et al., 2018) and ENSO (Cai et al., 2021) will change in the future as well as the subsequent changes to their teleconnections (e.g., Beverley et al., 2021; Cui & Li, 2021; Drouard & Cassou, 2019; Fereday et al., 2020; Samarasinghe et al., 2021; W. Zhou et al., 2020; Meehl et al., 2007; Z.-Q. Zhou et al., 2014). It stands to reason that these changes will likely impact subseasonal predictability across the Northern Hemisphere, but little work has been done in this area (an example being Sheshadri et al., 2021). Here, we utilize the Community Earth System Model Version 2 - Large Ensemble (CESM2-LE; Rodgers et al., 2021) and simple artificial neural networks to identify changes in subseasonal predictability provided by the tropics under future warming.

In recent years, neural networks have been successfully applied to weather and climate prediction (e.g., Chapman et al., 2021; Gordon et al., 2021; Ham et al., 2019; Labe & Barnes, 2022; Martin et al., 2022; Rasp & Thuerey, 2021; Weyn et al., 2021) due to their ability to extract nonlinear relationships from large amounts of data. This makes them advantageous for learning nonlinear relationships in the climate system. In addition, recent advances in explainability techniques and their application to climate sciences demonstrate than neural networks can identify physical relationships in the Earth system (e.g., Davenport & Diffenbaugh, 2021; Mayer & Barnes, 2021; McGovern et al., 2019; Toms et al., 2020). For example, Mayer and Barnes (2021) demonstrate that neural networks can be used to identify subseasonal forecasts of opportunity through the neural network's confidence in a given prediction. They further show that the network identifies physically meaningful sources of subseasonal predictability for the North Atlantic.

Here we use artificial neural networks to quantify how subseasonal prediction skill provided by the tropics may change under future climate warming. Given the importance of forecasts of opportunity for subseasonal prediction in the current climate, we examine both total changes to overall prediction skill as well as changes to skill during forecasts of opportunity, in particular. The artificial neural networks identify subseasonal prediction skill changes across the Northern Hemisphere in the CESM2-LE. In particular, there is an increase in prediction skill over the North Atlantic and western North America as well as a decrease over the North Pacific. In addition, this approach shows that the greatest changes in skill occur during forecasts of opportunity and that these changes appear linked to changes in seasonal variability in the CESM2-LE.

2 Data and Methods

2.1 Data

Here, we examine midlatitude subseasonal prediction skill changes using the first 10 members from the coupled Community Earth System Model Version 2 - Large Ensemble (CESM2-LE; Rodgers et al., 2021). CESM2 has both a well represented MJO (Ahn et al., 2020) and MJO teleconnections (J. Wang et al., 2022) and thus, is ideal for this analysis. We note that the results presented are specifically for the CESM2-LE, and as with all model-based results, are dependent on specific model biases (e.g., Sea surface temperature (SST) biases; Danabasoglu et al., 2020). We use the years 1970–2015 as our ‘“historical period” to represent a climate similar to today and compare it to the latter half of the century (2055–2100; “future period”) under the SSP3-7.0 climate change scenario. We find that 10 members are sufficient for this analysis as the network skill plateaus when at least five ensemble members are used for training, depending on location and time period (Figure S1 and Text S1 in Supporting Information S1). While additional ensemble members could be used, we believe our conclusions would remain unaffected, as the sign of the change in prediction skill of the 20% most confident predictions remains consistent regardless of the number of members examined here.

The CESM2-LE members #1–10 are split into training (members #1–8), validation (member #9) and testing data (member #10). To simultaneously detrend and remove the seasonal cycle for each grid point, the third order polynomial fit of the training and validation members' ensemble mean is subtracted from every ensemble member individually for each day of the year. We find the conclusions are insensitive to the specific members assigned to training, validation and testing (Figure S2 in Supporting Information S1).

We utilize the CESM2-LE tropical precipitation (28.5°S–28.5°N) and geopotential height at 500 hPa (z500; 31.25°N–88.75°N) during the extended boreal winter (November-March) since this is when MJO teleconnections tend to be strongest (Madden, 1986). Tropical precipitation anomalies are computed for each member and grid point by standardizing with the training data mean and standard deviation. For computational purposes, the z500 field is partitioned into non-overlapping 5° × 5° boxes, where the average of these values is assigned to the center grid point latitude and longitude. This decreases the z500 resolution from 2.5° × 2.5° to 7.5° × 7.5°, however, given the large scale structure of z500, we do not expect the resolution reduction to impact the conclusions. The sign of the z500 anomalies are defined by subtracting the training data median from the training, validation and testing data and converting the anomalies into 0 and 1s depending on the sign (negative and positive, respectively).

Sea surface temperatures from the first 10 members of the CESM2-LE are also used to calculate the Niño 3.4 index for each member, following the NCAR Climate Data Guide (2020). The trend and seasonal cycle is removed simultaneously as aforementioned, and a 5 month running mean is applied prior to standardizing the SSTs with each member's mean and standard deviation. An El Niño/La Niña event is therefore defined as a standardized Niño 3.4 index value of greater/less than ±1σ. We use this index to examine any possible role that ENSO may play in the identified changes to subseasonal predictability.

2.2 Neural Network Architecture and Application

The neural network ingests daily tropical precipitation anomalies and makes a prediction of the sign of z500 at a given grid point at a lead of 21 days (Week 3; Figure 1a). Prediction of the sign of z500 at each grid point allows the network freedom to learn important patterns and relationships between tropical precipitation and z500. The number of input nodes is equal to the number of precipitation grid points (N = 3,456). The first and second layer of the network consist of 128 and 8 nodes, respectively. A softmax activation function is applied to the output layer of 2 nodes which transforms the network output into values which sum to one. These transformed values represent a network estimation of likelihood, which we refer to as “model confidence”, where the predicted category is defined as a value greater than 0.5. As shown in Mayer and Barnes (2021), when prediction skill increases with model confidence, higher model confidence can be used to identify subseasonal forecasts of opportunity.

Details are in the caption following the image — **Figure 1**
Open in figure viewer PowerPoint

(a) The artificial neural network input (tropical precipitation), architecture (first hidden layer: 128 nodes, second hidden layer: 8 nodes) and output (sign of z500 hPa at a location “x”). (b and c) Timeseries of the *correct* sign predictions of z500 in ensemble member #10 for the historical (left column) and future (right column) for (b) the North Pacific and (c) the North Atlantic. Red (blue) dots indicate positive (negative) predictions. Darker dots denote the 20% most confident predictions, and the gray shading indicates when the standardized Niño 3.4 index exceeds ±1σ.

We use this network architecture because it has some of the highest validation skill for both the historical and the future time periods in the North Atlantic and also performs well in the North Pacific (Figures S3 and S4 in Supporting Information S1). We note that slight variations of the hyperparameters (i.e., network depth, nodes per layer, learning rate, ridge regression parameter) show similar skill. While one could optimize the architecture and hyperparameters for every gridpoint individually, we have not done this due to the considerable computational resources necessary and find it unlikely to lead to substantially different conclusions. For additional information on the network architecture and hyperparameters see Text S2 in Supporting Information S1.

Example correct network predictions for the testing ensemble member #10 are shown in Figures 1b and 1c for the historical (left column) and the future (right column) periods in (b) the North Pacific and (c) the North Atlantic. The color denotes the sign of the prediction and the darker colors denote the (20% most) confident predictions. The vertical gray shading indicates periods of ENSO events. Figures 1b and 1c demonstrates that the networks can accurately and confidently predict both sign anomalies. In addition, it shows a possible relationship between confident subseasonal predictions and ENSO events, but the amount which confident predictions coincide with ENSO events depends on location and time period. This relationship will be addressed further in Section 3.2.

3 Results

3.1 Changes in Subseasonal Prediction Skill

To examine how subseasonal prediction skill provided by tropical-extratropical teleconnections changes in a warmer climate, 100 networks are trained for the North Pacific (41.25°N, 205°E) and the North Atlantic (41.25°N, 325°E) for both the historical and future periods. These two locations are chosen because they encompass regions known to be significantly impacted by the MJO (e.g., Cassou, 2008; Lin et al., 2009; Mori & Watanabe, 2008) and ENSO (e.g., Wallace & Gutzler, 1981; Zhang et al., 1996) teleconnections, which subsequently have North American and European impacts. The 100 networks are created by varying their random seed to test the sensitivity of the network to the random initialized weights.

Accuracies binned by various model confidence thresholds are shown in Figure 2. Accuracy increases with model confidence (moving from left to right), suggesting the network is identifying forecasts of opportunity for these regions. In addition, we find that all networks at almost every confidence level perform better than random chance (Figure 2, Text S3 in Supporting Information S1). The North Pacific (Figure 2a) has higher accuracy compared to the North Atlantic (Figure 2c), likely due to the strong influence of tropical phenomena like the MJO and ENSO in modulating the circulation in the North Pacific (e.g., Mori & Watanabe, 2008; Riddle et al., 2013; Roundy et al., 2010; Wallace & Gutzler, 1981; Zhang et al., 1996). In the future, subseasonal prediction skill increases in the North Atlantic (Figure 2c) and decreases in the North Pacific (Figure 2a) in the CESM2-LE, and this is most evident at higher confidence values. If one examines the accuracy for all (100% most confident) predictions, the North Atlantic and North Pacific accuracies exhibit almost no difference between the two time periods. It is when we focus on the higher confidence predictions that a clear signal emerges. In other words, the changes in subseasonal prediction skill are most evident during forecasts of opportunity in these regions.

Histograms of the accuracies at the 20% most confident threshold (Figures 2b and 2d) further show that the future period has substantially shifted away from the historical period in both regions. The majority of the future North Atlantic accuracies exceed the 95th percentile of the historical accuracies, and all of the future North Pacific accuracies lie below the fifth percentile of the historical accuracies.

To explore whether the results in Figure 2 hold for other regions, we train 10 neural networks for each grid point and time period across the Northern Hemisphere. We train 10 networks instead of 100 for computational efficiency. To test whether these changes in skill in the North Atlantic and North Pacific could be seen with only 10 networks, we conducted a bootstrapping analysis (Text S4 and Figure S5 in Supporting Information S1) following the method used to create Figure 3, and find that 10 networks are sufficient for identifying these changes. Figure 3 shows the resulting mean testing accuracy of the top three of the 10 networks for each location. The top three networks are defined as the networks with the three highest 20% most confident validation accuracies. We use the top three networks so that the mean accuracies for each region are not as influenced by models that learn very little or not at all.

For all predictions (Figures 3a and 3b) and 20% most confident predictions (“confident predictions” from here on; Figures 3d and 3e), the locations of highest skill are in regions associated with the Pacific/North America pattern (PNA; Wallace & Gutzler, 1981). The higher accuracies over PNA regions suggests the network is most likely identifying forecasts of opportunity associated with teleconnections from the MJO and/or ENSO (e.g., Mori & Watanabe, 2008; Riddle et al., 2013; Roundy et al., 2010; Wallace & Gutzler, 1981; Zhang et al., 1996). We also find that the spatial coherence in accuracies across networks corresponds to the networks correctly predicting many of the same days for neighboring grid points (not shown). In the future period (Figures 3b and 3e), there is an additional region of higher accuracies spread across Asia and the North Atlantic. Overall, the confident predictions have higher accuracies than all predictions, indicating that higher model confidence predictions exhibit greater skill.

In the future, spatially coherent increases in skill are seen across Asia, along the west coast of North America, across the southern United States and throughout the North Atlantic (Figures 3c and 3f) while decreases are seen over the North Pacific, Canada and western Europe. While the change in skill over East Asia is substantial, it appears that the overall skill in East Asia for both time periods does not harness any subseasonal variability, but rather comes about exclusively from seasonal variability or longer timescales (Figures S8 and S9 in Supporting Information S1). As a result, these changes in skill are not addressed further here. The difference plots for both all and the confident predictions (Figures 3c and 3f) have similar spatial patterns of changes in accuracy, however, the confident predictions show the largest changes in skill. Specifically, the absolute maximum change in skill for all predictions is about 5% while the absolute maximum change in skill for confident predictions is about 10%. This further demonstrates that the greatest changes to subseasonal prediction skill provided by the tropics occur during forecasts of opportunity across the Northern Hemisphere, consistent with Figure 2.

3.2 Tropical Drivers of Changing Midlatitude Skill

Seasonal variability can have a large influence on subseasonal variability and prediction skill. In the tropics, ENSO can modulate the MJO (e.g., Hendon et al., 1999; Kessler, 2001; Pohl & Matthews, 2007) and the basic state (e.g., Moon et al., 2011; Namias, 1986; Takahashi & Shirooka, 2014), and ENSO teleconnections can (de)constructively interfere with MJO teleconnections (e.g., Arcodia et al., 2020; Henderson et al., 2020; Henderson & Maloney, 2018; Stan et al., 2017; Tseng et al., 2020). Recent studies have identified possible changes to both MJO and ENSO variability (Cai et al., 2021; Maloney et al., 2018) as well as their teleconnections (e.g., Beverley et al., 2021; Fredriksen et al., 2020; W. Zhou et al., 2020) under future climate warming. Thus, the changes in midlatitude subseasonal prediction skill seen in Figures 2 and 3 could be a reflection of changes to subseasonal variability, seasonal variability, or through a combination of changes to both.

We find that the increase in skill along the west coast of North America and in the North Atlantic is supported by previous research on MJO and ENSO teleconnections in a warmer climate. In particular, the subseasonal skill increase along the west coast of North America (Figure 3f) appears to be associated with a north-eastward shift of higher accuracies over the North Pacific in the future (Figures 3d and 3e). This is consistent with research showing that PNA patterns initiated by ENSO (e.g., Beverley et al., 2021; Fredriksen et al., 2020; Meehl & Teng, 2007; Meehl et al., 2007; Müller & Roeckner, 2008; Kug et al., 2010; Z.-Q. Zhou et al., 2014) and the MJO (J. Wang et al., 2022; Wolding et al., 2017; W. Zhou et al., 2020) are projected to shift eastward in a warmer climate in a variety of climate models, including CESM2 (Fredriksen et al., 2020; J. Wang et al., 2022). In the North Atlantic, increased skill is also consistent with research suggesting that the North Atlantic may become more sensitive to MJO teleconnections (Samarasinghe et al., 2021) and that the ENSO-North Atlantic Oscillation (NAO) teleconnection may strengthen (Drouard & Cassou, 2019; Fereday et al., 2020) in the future. The decrease in skill over the North Pacific is also consistent with recent research using a variety of CMIP6 models that suggests the ENSO teleconnection amplitude over the North Pacific may weaken in a warmer climate (e.g., Beverley et al., 2021; Fredriksen et al., 2020).

To gain insight into the neural network's identified sources of predictability, we apply explainable AI to create heatmaps of the relevant regions of the input tropical precipitation the network uses to make confident and correct predictions (see Text S5 in Supporting Information S1; Bach et al., 2015; Montavon et al., 2019). In the North Pacific and North Atlantic, the network tends to focus on the tropical equatorial Pacific, typically associated with ENSO (Figure S6 in Supporting Information S1). In the North Pacific, the future decrease in skill is associated with a decrease in relevance of the ENSO region (Figure S6a–S6d in Supporting Information S1). For the North Atlantic, the future increase in skill is associated with an increase in relevance of the ENSO region (Figure S6e–S6h in Supporting Information S1). These explainability results suggest that the changes in subseasonal prediction skill may be related to changes in the importance of the ENSO region (i.e., seasonal variability), even though both subseasonal and seasonal variability are contributing to the total skill (Figures S8 and S9 in Supporting Information S1). This changing role of ENSO in both regions is also evident in the prediction timeseries in Figure 1. In the North Atlantic (Figure 1c), the confident predictions in the historical period are scattered throughout the years, whereas in the future period, the confident predictions correspond more frequently with ENSO events (darker dots mainly occur in the gray shading). The opposite is seen for the North Pacific (Figure 1b). Given the results of this analysis, we next examine if the changes in midlatitude subseasonal prediction skill are related to changes in ENSO teleconnections.

We analyze the relationship between ENSO teleconnections and subseasonal prediction skill changes across the Northern Hemisphere by calculating how often a positive z500 anomaly occurs 21 days following an El Niño/La Niña event (Figure 4). This metric quantifies the consistency of specific teleconnections following ENSO events and thus, demonstrates the downstream influence of ENSO on specific regions. Therefore, any regional changes to the consistency between the two time periods implies changes to the impact of ENSO in that region. Over the North Pacific, the consistency of the z500 sign following both ENSO phases decreases (Figures 4c and 4f), suggestive of a reduction in the influence of ENSO teleconnections. Furthermore, the large decrease in skill over Canada (Figure 3f) aligns with the decrease in El Niño teleconnection consistency in the future (Figure 4f). Over the North Atlantic, there is a slight increase in ENSO teleconnection consistency which may be related to the projected strengthening of the ENSO-NAO teleconnection in the future (Drouard & Cassou, 2019; Fereday et al., 2020). Lastly, the increase in skill along the west coast of North America (Figure 3f) aligns with an increase in consistency of La Niña teleconnections (Figure 4c). Thus, we hypothesize that the substantial changes in subseasonal prediction skill in regions across the Northern Hemisphere are connected to changes in ENSO teleconnections in the CESM2-LE.

We provide further evidence of the role of seasonal variability in changes to subseasonal prediction skill through an additional neural network analysis in the North Pacific and North Atlantic. We filter out 60+ day variability from the z500 anomalies (Text S6 in Supporting Information S1) to remove low-frequency signals such as those from ENSO teleconnections. With this filtering, there is almost no change in skill between the historical and future period in the North Pacific (Figure S7c and S7d in Supporting Information S1). This demonstrates that the decrease in skill in this region is mainly a result of changes to seasonal variability. In the North Atlantic, the increase in skill is still seen, but to a reduced degree when the lower frequencies are removed (Figure S7e and S7f in Supporting Information S1). This suggests that seasonal variability is playing a role in subseasonal prediction skill changes in this region, however, there is also likely a contribution from subseasonal variability to these changes. This is consistent with research that suggests the North Atlantic may become more sensitive to MJO teleconnections in the future (Samarasinghe et al., 2021).

The influence of seasonal variability on subseasonal prediction skill changes can be further examined in the North Pacific and North Atlantic by training the neural networks to instead predict the sign of unfiltered z500 anomalies on seasonal lead times. In the North Pacific, we find that changes in skill at 60 and 90 day leads are similar to that for a lead of 21 days. This again implies that the changes in subseasonal prediction skill seen in the North Pacific are due to changes in seasonal variability. In the North Atlantic, the change in skill for the seasonal lead time is larger than the 21 day lead time. This difference in the change suggests that the network is focusing on different sources of predictability for the 21 day prediction compared to the 60 or 90 day predictions, implying again that the change in skill in the North Atlantic is not purely due to seasonal variability changes in the future (Figures S8 and S9 in Supporting Information S1).

4 Conclusions

While accurate subseasonal predictions are important for society (White et al., 2022), this timescale is known to exhibit limited predictability (Vitart et al., 2017). One method to improve prediction skill on subseasonal timescales is to utilize Earth system states which are known to provide enhanced subseasonal predictability when they are present (forecasts of opportunity; Mariotti et al., 2020). Previous research has examined how specific Earth system states important for subseasonal prediction (e.g., MJO and ENSO) and their teleconnections may change in a warmer climate (e.g., Cai et al., 2021; Maloney et al., 2018; J. Wang et al., 2022). To address whether these projected changes ultimately impact subseasonal predictability, we use the CESM2-LE and simple artificial neural networks to quantify and understand how subseasonal predictability provided by the tropics may change in a warmer climate. We find that there are changes to subseasonal prediction skill across the Northern Hemisphere and the largest differences in skill mainly occur during forecasts of opportunity.

Our results are supported by recent research on changes to MJO and ENSO teleconnections. In particular, the increase in skill along the west coast of North America is consistent with the projected eastward shift of MJO and ENSO teleconnections in the future (e.g., Beverley et al., 2021; Fredriksen et al., 2020; Jenney et al., 2021; J. Wang et al., 2022). In addition, our results suggest there is a contribution from both subseasonal and seasonal variability changes to the increase in prediction skill in the North Atlantic. This is consistent with research suggesting the North Atlantic becomes more sensitive to the MJO (Samarasinghe et al., 2021) and ENSO (Drouard & Cassou, 2019; Fereday et al., 2020) in the future. We also identify a substantial decrease in skill over the North Pacific and from our analysis, hypothesize that this decrease is mainly driven by a reduced influence of ENSO teleconnections to this region in the future. Overall, while both MJO and ENSO teleconnections are projected to change in the future, our analysis demonstrates that changes to ENSO and its teleconnections (e.g., seasonal variability) at least partially explain substantial changes in subseasonal prediction skill across the North Hemisphere in the CESM2-LE. Changes to subseasonal variability may still play a role in changes to subseasonal prediction skill in certain locations (e.g., North Atlantic), but further work is needed to understand and quantify its contribution. In addition, we only explored the Niño 3.4 index as a metric for ENSO variability, and future work could further extend this to other metrics that capture ENSO dynamics and that may also account for possible changes in ENSO variability under climate change.

Using the CESM2-LE, we show that neural networks are a useful tool for identifying and understanding future changes in predictability. In addition, we find that changes in subseasonal prediction skill across the Northern Hemisphere are often largest during forecasts of opportunity, suggesting that future research on prediction skill changes should focus on periods of enhanced predictability. While this research addresses changes in boreal wintertime subseasonal predictability provided by the tropics, future research should also examine how other seasons and sources of predictability may be affected in a warmer climate. This could include identifying possible changes to the importance of the stratosphere for subseasonal prediction or changes to boreal summer subseasonal predictability due to changes to the importance of the boreal summer intraseasonal oscillation (B. Wang & Rui, 1990). Furthermore, although this work examines subseasonal predictability changes by the end of the century, examining how quickly these changes may be detected is also worthy of study. Ultimately, this research demonstrates the utility of neural networks to quantify and gain physical insight into changes in subseasonal predictability in future climates.

Acknowledgments

This research is partially funded by the National Science Foundation Graduate Research Fellowship under Grant 006784 and partially funded by the Regional and Global Model Analysis program area of the Department of Energy's Office of Biological and Environmental Research as part of the Program for Climate Model Diagnosis and Intercomparison Project.

Conflict of Interest

The authors declare no conflicts of interest relevant to this study.

Open Research

Data Availability Statement

CESM2 Large Ensemble data (precipitation, SSTs and z500) are provided by the University Corporation for Atmospheric Research/National Center for Atmospheric Research (https://www.cesm.ucar.edu/projects/community-projects/LENS2/data-sets.html).

Supporting Information

References

Ahn, M., Kim, D., Kang, D., Lee, J., Sperber, K. R., Gleckler, P. J., et al. (2020). MJO propagation across the maritime continent: Are CMIP6 models better than CMIP5 models? Geophysical Research Letters, 47(11), 741. https://doi.org/10.1029/2020gl087250
Arcodia, M. C., Kirtman, B. P., & Siqueira, L. S. P. (2020). How MJO teleconnections and ENSO interference impacts US precipitation. Journal of Climate, 33(11), 4621–4640. https://doi.org/10.1175/JCLI-D-19-0448.1
Bach, S., Binder, A., Montavon, G., Klauschen, F., Müller, K.-R., & Samek, W. (2015). On Pixel-Wise explanations for Non-Linear classifier decisions by Layer-Wise relevance propagation. PLoS One, 10(7), e0130140. https://doi.org/10.1371/journal.pone.0130140
Baggett, C. F., Barnes, E. A., Maloney, E. D., & Mundhenk, B. D. (2017). Advancing atmospheric river forecasts into subseasonal-to-seasonal time scales. Geophysical Research Letters, 44(14), 7528–7536. https://doi.org/10.1002/2017gl074434
Beverley, J. D., Collins, M., Hugo Lambert, F., & Chadwick, R. (2021). Future changes to el niño teleconnections over the north Pacific and North America. Journal of Climate, 34(15), 6191–6205. https://doi.org/10.1175/JCLI-D-20-0877.1
Cai, W., Santoso, A., Collins, M., Dewitte, B., Karamperidou, C., Kug, J.-S., et al. (2021). Changing el niño–southern oscillation in a warming climate. Nature Reviews Earth & Environment, 2(9), 628–644. https://doi.org/10.1038/s43017-021-00199-z
Cassou, C. (2008). Intraseasonal interaction between the Madden-Julian oscillation and the north Atlantic oscillation. Nature, 455(7212), 523–527. https://doi.org/10.1038/nature07286
Chapman, W. E., Monache, L. D., Alessandrini, S., Subramanian, A. C., Martin Ralph, F., Xie, S.-P., et al. (2021). Probabilistic predictions from deterministic atmospheric river forecasts with deep learning. Monthly Weather Review, 150(1), 215–234. https://doi.org/10.1175/MWR-D-21-0106.1
Cui, J., & Li, T. (2021). Changes in MJO characteristics and impacts in the past century. Journal of Climate, 35(2), 577–590. https://doi.org/10.1175/JCLI-D-21-0306.1
Danabasoglu, G., Lamarque, J. F., Bacmeister, J., Bailey, D. A., DuVivier, A. K., Edwards, J., et al. (2020). The community Earth system model version 2 (CESM2). Journal of Advances in Modeling Earth Systems, 12(2), 106. https://doi.org/10.1029/2019MS001916
Davenport, F. V., & Diffenbaugh, N. S. (2021). Using machine learning to analyze physical causes of climate change: A case study of U.S. Midwest extreme precipitation. Geophysical Research Letters, 48(15), e2021GL093787. https://doi.org/10.1029/2021GL093787
Drouard, M., & Cassou, C. (2019). A modeling- and Process-Oriented study to investigate the projected change of ENSO-Forced wintertime teleconnectivity in a warmer world. Journal of Climate, 32(23), 8047–8068. https://doi.org/10.1175/jcli-d-18-0803.1
Fereday, D. R., Chadwick, R., Knight, J. R., & Scaife, A. A. (2020). Tropical rainfall linked to stronger future ENSO-NAO teleconnection in CMIP5 models. Geophysical Research Letters, 47(22), e2020GL088664. https://doi.org/10.1029/2020GL088664
Fredriksen, H.-B., Berner, J., Subramanian, A. C., & Capotondi, A. (2020). How does el niño–southern oscillation change under global warming—A first look at CMIP6. Geophysical Research Letters, 47(22), e2020GL090640. https://doi.org/10.1029/2020gl090640
Gordon, E. M., Barnes, E. A., & Hurrell, J. W. (2021). Oceanic harbingers of Pacific decadal oscillation predictability in CESM2 detected by neural networks. Geophysical Research Letters, 48(21), e2021GL095392. https://doi.org/10.1029/2021gl095392
Ham, Y.-G., Kim, J.-H., & Luo, J.-J. (2019). Deep learning for multi-year ENSO forecasts. Nature, 573(7775), 568–572. https://doi.org/10.1038/s41586-019-1559-7
Henderson, S. A., & Maloney, E. D. (2018). The impact of the Madden–Julian oscillation on High-Latitude winter blocking during el niño–southern oscillation events. Journal of Climate, 31(13), 5293–5318. https://doi.org/10.1175/jcli-d-17-0721.1
Henderson, S. A., Maloney, E. D., & Barnes, E. A. (2016). The influence of the Madden–Julian oscillation on northern hemisphere winter blocking. Journal of Climate, 29(12), 4597–4616. https://doi.org/10.1175/jcli-d-15-0502.1
Henderson, S. A., Vimont, D. J., & Newman, M. (2020). The critical role of Non-Normality in partitioning tropical and extratropical contributions to PNA growth. Journal of Climate, 33(14), 6273–6295. https://doi.org/10.1175/jcli-d-19-0555.1
Hendon, H. H., Zhang, C., & Glick, J. D. (1999). Interannual variation of the Madden–Julian oscillation during austral summer. Journal of Climate, 12(8), 2538–2550. https://doi.org/10.1175/1520-0442(1999)012<2538:ivotmj>2.0.co;2
Hoskins, B. J., & Ambrizzi, T. (1993). Rossby wave propagation on a realistic longitudinally varying flow. Journal of the Atmospheric Sciences, 50(12), 1661–1671. https://doi.org/10.1175/1520-0469(1993)050<1661:rwpoar>2.0.co;2
Hoskins, B. J., & Karoly, D. J. (1981). The steady linear response of a spherical atmosphere to thermal and orographic forcing. Journal of the Atmospheric Sciences, 38(6), 1179–1196. https://doi.org/10.1175/1520-0469(1981)038<1179:tslroa>2.0.co;2
Jenney, A. M., Randall, D. A., & Barnes, E. A. (2021). Drivers of uncertainty in future projections of Madden–Julian oscillation teleconnections. Weather and Climate Dynamics, 2(3), 653–673. https://doi.org/10.5194/wcd-2-653-2021
Johnson, N. C., Collins, D. C., Feldstein, S. B., L’Heureux, M. L., & Riddle, E. E. (2014). Skillful wintertime North American temperature forecasts out to 4 weeks based on the state of ENSO and the MJO. Weather and Forecasting, 29(1), 23–38. https://doi.org/10.1175/WAF-D-13-00102.1
Kessler, W. S. (2001). EOF representations of the Madden–Julian oscillation and its connection with ENSO. Journal of Climate, 14(13), 3055–3061. https://doi.org/10.1175/1520-0442(2001)014<3055:erotmj>2.0.co;2
Kug, J.-S., An, S.-I., Ham, Y.-G., & Kang, I.-S. (2010). Changes in el niño and la niña teleconnections over north Pacific–America in the global warming simulations. Theoretical and Applied Climatology, 100(3), 275–282. https://doi.org/10.1007/s00704-009-0183-0
Labe, Z. M., & Barnes, E. A. (2022). Predicting slowdowns in decadal climate warming trends with explainable neural networks. Geophysical Research Letters. https://doi.org/10.1029/2022GL098173
Lin, H., Brunet, G., & Derome, J. (2009). An observed connection between the north Atlantic oscillation and the Madden–Julian oscillation. Journal of Climate, 22(2), 364–380. https://doi.org/10.1175/2008jcli2515.1
Madden, R. A. (1986). Seasonal variations of the 40-50 day oscillation in the tropics. Journal of the Atmospheric Sciences, 43(24), 3138–3158. https://doi.org/10.1175/1520-0469(1986)043〈3138:SVOTDO〉2.0.CO;2
Madden, R. A., & Julian, P. R. (1971). Detection of a 40–50 day oscillation in the zonal wind in the tropical Pacific. Journal of the Atmospheric Sciences, 28(5), 702–708. https://doi.org/10.1175/1520-0469(1971)028<0702:doadoi>2.0.co;2
Madden, R. A., & Julian, P. R. (1972). Description of Global-Scale circulation cells in the tropics with a 40–50 day period. Journal of the Atmospheric Sciences, 29(6), 1109–1123. https://doi.org/10.1175/1520-0469(1972)029<1109:dogscc>2.0.co;2
Maloney, E. D., Adames, Á. F., & Bui, H. X. (2018). Madden–Julian oscillation changes under anthropogenic warming. Nature Climate Change, 9(1), 26–33. https://doi.org/10.1038/s41558-018-0331-6
Mariotti, A., Baggett, C., Barnes, E. A., Becker, E., Butler, A., Collins, D. C., et al. (2020). Windows of opportunity for skillful forecasts subseasonal to seasonal and beyond. Bulletin of the American Meteorological Society, 101(5), E608–E625. https://doi.org/10.1175/BAMS-D-18-0326.1
Martin, Z. K., Barnes, E. A., & Maloney, E. (2022). Using simple, explainable neural networks to predict the Madden-Julian oscillation. Journal of Advances in Modeling Earth Systems, 14(5), e2021MS002774. https://doi.org/10.1029/2021MS002774
Mayer, K. J., & Barnes, E. A. (2021). Subseasonal forecasts of opportunity identified by an explainable neural network. Geophysical Research Letters, 48(10), e2020GL092092. https://doi.org/10.1029/2020GL092092
McGovern, A., Lagerquist, R., Gagne, D. J., Eli Jergensen, G., Elmore, K. L., Homeyer, C. R., & Smith, T. (2019). Making the black box more transparent: Understanding the physical implications of machine learning. Bulletin of the American Meteorological Society, 100(11), 2175–2199. https://doi.org/10.1175/bams-d-18-0195.1
Meehl, G. A., Tebaldi, C., Teng, H., & Peterson, T. C. (2007). Current and future U.S. weather extremes and el niño. Geophysical Research Letters, 34(20), L20704. https://doi.org/10.1029/2007gl031027
Meehl, G. A., & Teng, H. (2007). Multi-model changes in el niño teleconnections over North America in a future warmer climate. Climate Dynamics, 29(7–8), 779–790. https://doi.org/10.1007/s00382-007-0268-3
Montavon, G., Binder, A., Lapuschkin, S., Samek, W., & Müller, K.-R. (2019). Layer-Wise relevance propagation: An overview. In W. Samek, G. Montavon, A. Vedaldi, L. K. Hansen, & K.-R. Müller (Eds.), Explainable AI: Interpreting, explaining and visualizing deep learning (pp. 193–209). Springer International Publishing.
Moon, J.-Y., Wang, B., & Ha, K.-J. (2011). ENSO regulation of MJO teleconnection. Climate Dynamics, 37(5), 1133–1149. https://doi.org/10.1007/s00382-010-0902-3
Mori, M., & Watanabe, M. (2008). The growth and triggering mechanisms of the PNA: A MJO-PNA coherence. Journal of the Meteorological Society of Japan. Ser. II, 86(1), 213–236. https://doi.org/10.2151/jmsj.86.213
Müller, W. A., & Roeckner, E. (2008). ENSO teleconnections in projections of future climate in ECHAM5/MPI-OM. Climate Dynamics, 31(5), 533–549. https://doi.org/10.1007/s00382-007-0357-3
Namias, J. (1986). Persistence of flow patterns over North America and adjacent ocean sectors. Monthly Weather Review, 114(7), 1368–1383. https://doi.org/10.1175/1520-0493(1986)114<1368:pofpon>2.0.co;2
NCAR climate data guide. (2020). Retrieved from https://climatedataguide.ucar.edu/climate-data/nino-sst-indices-nino-12-3-34-4-oni-and-tni
Pohl, B., & Matthews, A. J. (2007). Observed changes in the lifetime and amplitude of the Madden–Julian oscillation associated with interannual ENSO sea surface temperature anomalies. Journal of Climate, 20(11), 2659–2674. https://doi.org/10.1175/jcli4230.1
Rasp, S., & Thuerey, N. (2021). Data-driven medium-range weather prediction with a resnet pretrained on climate simulations: A new model for WeatherBench. Journal of Advances in Modeling Earth Systems, 13(2), e2020MS002405. https://doi.org/10.1029/2020ms002405
Riddle, E. E., Stoner, M. B., Johnson, N. C., L’Heureux, M. L., Collins, D. C., & Feldstein, S. B. (2013). The impact of the MJO on clusters of wintertime circulation anomalies over the North American region. Climate Dynamics, 40(7), 1749–1766. https://doi.org/10.1007/s00382-012-1493-y
Rodgers, K. B., Lee, S.-S., Rosenbloom, N., Timmermann, A., Danabasoglu, G., Deser, C., et al. (2021). Ubiquity of human-induced changes in climate variability. Earth System Dynamics Discussions, 12(4), 1–22. https://doi.org/10.5194/esd-12-1393-2021
Roundy, P. E., MacRitchie, K., Asuma, J., & Melino, T. (2010). Modulation of the global atmospheric circulation by combined activity in the Madden–Julian oscillation and the el niño–southern oscillation during boreal winter. Journal of Climate, 23(15), 4045–4059. https://doi.org/10.1175/2010jcli3446.1
Samarasinghe, S. M., Connolly, C., Barnes, E. A., Ebert-Uphoff, I., & Sun, L. (2021). Strengthened causal connections between the MJO and the north Atlantic with climate warming. Geophysical Research Letters, 48(5), e2020GL091168. https://doi.org/10.1029/2020gl091168
Sardeshmukh, P. D., & Hoskins, B. J. (1988). The generation of global rotational flow by steady idealized tropical divergence. Journal of the Atmospheric Sciences, 45(7), 1228–1251. https://doi.org/10.1175/1520-0469(1988)045<1228:tgogrf>2.0.co;2
Sheshadri, A., Borrus, M., Yoder, M., & Robinson, T. (2021). Midlatitude error growth in atmospheric GCMs: The role of eddy growth rate. Geophysical Research Letters, 48(23), e2021GL096126. https://doi.org/10.1029/2021gl096126
Stan, C., Straus, D. M., Frederiksen, J. S., Lin, H., Maloney, E. D., & Schumacher, C. (2017). Review of Tropical-Extratropical teleconnections on intraseasonal time scales: The subseasonal to seasonal (S2S) teleconnection Sub-Project. Reviews of Geophysics, 55(4), 902–937. https://doi.org/10.1002/2016rg000538
Takahashi, C., & Shirooka, R. (2014). Storm track activity over the north Pacific associated with the Madden-Julian oscillation under ENSO conditions during boreal winter. Journal of Geophysical Research, 119(18), 10663–10683. https://doi.org/10.1002/2014jd021973
Toms, B. A., Barnes, E. A., & Ebert-Uphoff, I. (2020). Physically interpretable neural networks for the geosciences: Applications to Earth system variability. Journal of Advances in Modeling Earth Systems, 12(9), e2019MS002002. https://doi.org/10.1029/2019ms002002
Trenberth, K. E. (1997). The definition of el nino. Bulletin of the American Meteorological Society, 78(12), 2771–2778. https://doi.org/10.1175/1520-0477(1997)078<2771:tdoeno>2.0.co;2
Tseng, K.-C., Barnes, E. A., & Maloney, E. D. (2018). Prediction of the midlatitude response to strong Madden-Julian oscillation events on S2S time scales: Prediction of Z500 at S2S time scales. Geophysical Research Letters, 45(1), 463–470. https://doi.org/10.1002/2017gl075734
Tseng, K.-C., Maloney, E., & Barnes, E. A. (2020). The consistency of MJO teleconnection patterns on interannual time scales. Journal of Climate, 33(9), 3471–3486. https://doi.org/10.1175/jcli-d-19-0510.1
Vitart, F., Ardilouze, C., Bonet, A., Brookshaw, A., Chen, M., Codorean, C., et al. (2017). The subseasonal to seasonal (S2S) prediction project database. Bulletin of the American Meteorological Society, 98(1), 163–173. https://doi.org/10.1175/bams-d-16-0017.1
Wallace, J. M., & Gutzler, D. S. (1981). Teleconnections in the geopotential height field during the northern hemisphere winter. Monthly Weather Review, 109(4), 784–812. https://doi.org/10.1175/1520-0493(1981)109<0784:titghf>2.0.co;2
Wang, B., & Rui, H. (1990). Synoptic climatology of transient tropical intraseasonal convection anomalies: 1975–1985. Meteorology and Atmospheric Physics, 44(1), 43–61. https://doi.org/10.1007/bf01026810
Wang, J., Kim, H., & DeFlorio, M. J. (2022). Future changes of PNA-like MJO teleconnections in CMIP6 models: Underlying mechanisms and uncertainty. Journal of Climate, 35(11), 1–40. https://doi.org/10.1175/jcli-d-21-0445.1
Wang, L., & Robertson, A. W. (2019). Week 3–4 predictability over the United States assessed from two operational ensemble prediction systems. Climate Dynamics, 52(9), 5861–5875. https://doi.org/10.1007/s00382-018-4484-9
Weyn, J. A., Durran, D. R., Caruana, R., & Cresswell-Clay, N. (2021). Sub-seasonal forecasting with a large ensemble of deep-learning weather prediction models. Journal of Advances in Modeling Earth Systems, 13(7), e2021MS002502. https://doi.org/10.1029/2021MS002502
White, C. J., Domeisen, D. I. V., Acharya, N., Adefisan, E. A., Anderson, M. L., Aura, S., et al. (2022). Advances in the application and utility of subseasonal-to-seasonal predictions. Bulletin of the American Meteorological Society, 1, 1–57. https://doi.org/10.1175/BAMS-D-20-0224.1
Wolding, B. O., Maloney, E. D., Henderson, S., & Branson, M. (2017). Climate change and the madden-julian oscillation: A vertically resolved weak temperature gradient analysis. Journal of Advances in Modeling Earth Systems, 9(1), 307–331. https://doi.org/10.1002/2016ms000843
Zhang, Y., Wallace, J. M., & Iwasaka, N. (1996). Is climate variability over the north Pacific a linear response to ENSO? Journal of Climate, 9(7), 1468–1478. https://doi.org/10.1175/1520-0442(1996)009<1468:icvotn>2.0.co;2
Zheng, C., Kar-Man Chang, E., Kim, H.-M., Zhang, M., & Wang, W. (2018). Impacts of the Madden–Julian oscillation on Storm-Track activity, surface air temperature, and precipitation over north America. https://doi.org/10.1175/JCLI-D-17-0534.1
Zhou, W., Yang, D., Xie, S.-P., & Ma, J. (2020). Amplified Madden–Julian oscillation impacts in the Pacific–North America region. Nature Climate Change, 10(7), 654–660. https://doi.org/10.1038/s41558-020-0814-0
Zhou, Z.-Q., Xie, S.-P., Zheng, X.-T., Liu, Q., & Wang, H. (2014). Global Warming–Induced changes in el niño teleconnections over the north Pacific and North America. Journal of Climate, 27(24), 9050–9064. https://doi.org/10.1175/jcli-d-14-00254.1

References From the Supporting Information

Cassou, C., Kushnir, Y., Hawkins, E., Pirani, A., Kucharski, F., Kang, I.-S., & Caltabiano, N. (2018). Decadal climate variability and predictability: Challenges and opportunities. Bulletin of the American Meteorological Society, 99(3), 479–490. https://doi.org/10.1175/bams-d-16-0286.1
Friedman, J. H. (2012). Fast sparse regression and classification. International Journal of Forecasting, 28(3), 722–738. https://doi.org/10.1016/j.ijforecast.2012.05.001
Goodfellow, I., Bengio, Y., Courville, A., & Bengio, Y. (2016). Deep learning (Vol. 1). MIT press.
Kingma, D. P., & Ba, J. (2014). Adam: A method for stochastic optimization. https://doi.org/10.48550/arXiv.1412.6980
Mamalakis, A., Ebert-Uphoff, I., & Barnes, E. A. (2022). Neural network attribution methods for problems in geoscience: A novel synthetic benchmark dataset. Environmental Data Science, 1, E8. https://doi.org/10.1017/eds.2022.7
Nielsen, M. A. (2015). Neural networks and deep learning (Vol. 25). Determination press.
Simpson, I. R., Deser, C., McKinnon, K. A., & Barnes, E. A. (2018). Modeled and observed multidecadal variability in the north Atlantic jet stream and its connection to sea surface temperatures. Journal of Climate, 31(20), 8313–8338. https://doi.org/10.1175/jcli-d-18-0168.1
Welch, B. L. (1947). The generalisation of student’s problems when several different population variances are involved. Biometrika, 34(1–2), 28–35. https://doi.org/10.2307/2332510

Volume49, Issue14

28 July 2022

e2022GL098663

Quantifying the Effect of Climate Change on Midlatitude Subseasonal Prediction Skill Provided by the Tropics

Abstract

Key Points

Plain Language Summary

1 Introduction

2 Data and Methods

2.1 Data

2.2 Neural Network Architecture and Application

3 Results

3.1 Changes in Subseasonal Prediction Skill

3.2 Tropical Drivers of Changing Midlatitude Skill

4 Conclusions

Acknowledgments

Conflict of Interest

Open Research

Data Availability Statement

Supporting Information

References

References From the Supporting Information

Editor Highlight—Quantifying Changes in Midlatitude Subseasonal Prediction Skill

Figures

References

Information

RESOURCES

PUBLICATION INFO

About Wiley Online Library

Help & Support

Opportunities

Connect with Wiley

Quantifying the Effect of Climate Change on Midlatitude Subseasonal Prediction Skill Provided by the Tropics

Abstract

Key Points

Plain Language Summary

1 Introduction

2 Data and Methods

2.1 Data

2.2 Neural Network Architecture and Application

3 Results

3.1 Changes in Subseasonal Prediction Skill

3.2 Tropical Drivers of Changing Midlatitude Skill

4 Conclusions

Acknowledgments

Conflict of Interest

Open Research

Data Availability Statement

Supporting Information

References

References From the Supporting Information

Editor Highlight—Quantifying Changes in Midlatitude Subseasonal Prediction Skill

Figures

References

Related

Information

RESOURCES

PUBLICATION INFO

Editor Highlight—Quantifying Changes in Midlatitude Subseasonal Prediction Skill