skip to main content


Title: Skillful Coupled Atmosphere‐Ocean Forecasts on Interannual to Decadal Timescales Using a Linear Inverse Model
Abstract

There are two major challenges to improving interannual to decadal forecasts: (a) consistently initializing the coupled system so that variability is not dominated by initial imbalances, and (b) having a large sample of different initial conditions on which to test forecast skill. The second challenge requires consideration of time periods not only outside the recent period of intensive ocean observation, but also before the instrumental era, which increases the importance of the first challenge. Forecasts prior to the 1850s isolate internally generated sources of variability by removing the majority of anthropogenic forcing, and the sparse observational record during this time period motivates the use of paleoclimate proxy data. We address these issues by using a linear inverse model (LIM) approach and a recent proxy‐based reconstruction over the last millennium at annual resolution. The reconstruction is used to train, initialize, and validate LIM forecasts. The LIM trained on paleo‐data assimilated using a LIM trained on global climate model (GCM) simulation data outperforms a LIM trained on raw GCM data at forecast leads longer than 2 years for in‐sample experiments, and beyond 4‐year leads in most out‐of‐sample experiments validated on instrumental data. The most skillful normal mode of the paleo‐data LIM for the instrumental experiment represents a persistent pattern with a longer decay time than for the GCM‐LIM's modes, which accounts for the outperformance at longer leads. The paleo‐data LIM is consequently more sensitive to ocean initialization, which is reflected in forecasts during the instrumental era where ocean reanalyses exhibit large uncertainty.

 
more » « less
NSF-PAR ID:
10409285
Author(s) / Creator(s):
 ;  
Publisher / Repository:
DOI PREFIX: 10.1029
Date Published:
Journal Name:
Earth and Space Science
Volume:
10
Issue:
4
ISSN:
2333-5084
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. Abstract

    Forecasting the El Niño-Southern Oscillation (ENSO) has been a subject of vigorous research due to the important role of the phenomenon in climate dynamics and its worldwide socioeconomic impacts. Over the past decades, numerous models for ENSO prediction have been developed, among which statistical models approximating ENSO evolution by linear dynamics have received significant attention owing to their simplicity and comparable forecast skill to first-principles models at short lead times. Yet, due to highly nonlinear and chaotic dynamics (particularly during ENSO initiation), such models have limited skill for longer-term forecasts beyond half a year. To resolve this limitation, here we employ a new nonparametric statistical approach based on analog forecasting, called kernel analog forecasting (KAF), which avoids assumptions on the underlying dynamics through the use of nonlinear kernel methods for machine learning and dimension reduction of high-dimensional datasets. Through a rigorous connection with Koopman operator theory for dynamical systems, KAF yields statistically optimal predictions of future ENSO states as conditional expectations, given noisy and potentially incomplete data at forecast initialization. Here, using industrial-era Indo-Pacific sea surface temperature (SST) as training data, the method is shown to successfully predict the Niño 3.4 index in a 1998–2017 verification period out to a 10-month lead, which corresponds to an increase of 3–8 months (depending on the decade) over a benchmark linear inverse model (LIM), while significantly improving upon the ENSO predictability “spring barrier”. In particular, KAF successfully predicts the historic 2015/16 El Niño at initialization times as early as June 2015, which is comparable to the skill of current dynamical models. An analysis of a 1300-yr control integration of a comprehensive climate model (CCSM4) further demonstrates that the enhanced predictability afforded by KAF holds over potentially much longer leads, extending to 24 months versus 18 months in the benchmark LIM. Probabilistic forecasts for the occurrence of El Niño/La Niña events are also performed and assessed via information-theoretic metrics, showing an improvement of skill over LIM approaches, thus opening an avenue for environmental risk assessment relevant in a variety of contexts.

     
    more » « less
  2. Abstract

    We use online data assimilation to combine information from a linear inverse model of coupled atmosphere‐ocean dynamics with proxy records to create a new annual‐resolution reconstruction of atmosphere and ocean fields over the last millennium. Instrumental validation of reconstructed sea‐surface temperature and 0–700 m ocean heat content shows broad regions of positive spatial correlations, and high correlations (∼0.6–0.9) for global averages and indices of large‐scale modes of atmospheric variability. Compared to previous reconstructions, the online reconstructions show global and hemispheric averages with little‐to‐no millennial‐scale trend and global‐mean temperatures ∼0.25–0.5 K cooler during early periods (1000–1400 C.E.). The spatial anomaly differences of average temperature between an early (1000–1250 C.E.) and later (1400–1700 C.E.) period show warm anomalies over high‐latitude Europe and cool tropical conditions in partial agreement with previous assessments. The addition of online data assimilation, which provides dynamical memory to climate proxy information, is shown to be crucial for adequately characterizing decadal‐to‐centennial‐scale variability of 0–700 m ocean heat content. Furthermore, the climate forecasts provide model‐based physical constraints for atmosphere–ocean interaction, which become increasingly important during early periods when less proxy information is available for assimilation.

     
    more » « less
  3. Abstract

    Skillfully predicting the North Atlantic Oscillation (NAO), and the closely related northern annular mode (NAM), on ‘subseasonal’ (weeks to less than a season) timescales is a high priority for operational forecasting centers, because of the NAO’s association with high-impact weather events, particularly during winter. Unfortunately, the relatively fast, weather-related processes dominating total NAO variability are unpredictable beyond about two weeks. On longer timescales, the tropical troposphere and the stratosphere provide some predictability, but they contribute relatively little to total NAO variance. Moreover, subseasonal forecasts are only sporadically skillful, suggesting the practical need to identify the fewer potentially predictable events at the time of forecast. Here we construct an observationally based linear inverse model (LIM) that predicts when, and diagnoses why, subseasonal NAO forecasts will be most skillful. We use the LIM to identify those dynamical modes that, despite capturing only a fraction of overall NAO variability, are largely responsible for extended-range NAO skill. Predictable NAO events stem from the linear superposition of these modes, which represent joint tropical sea-surface temperature-lower stratosphere variability plus a single mode capturing downward propagation from the upper stratosphere. Our method has broad applicability because both the LIM and the state-of-the-art European Centre for Medium-Range Weather Forecasts Integrated Forecast System (IFS) have higher (and comparable) skill for the same set of predicted high skill forecast events, suggesting that the low-dimensional predictable subspace identified by the LIM is relevant to real-world subseasonal NAO predictions.

     
    more » « less
  4. null (Ed.)
    Abstract Using an assemblage of four ice cores collected around the Pacific basin, one of the first basinwide histories of Pacific climate variability has been created. This ice core–derived index of the interdecadal Pacific oscillation (IPO) incorporates ice core records from South America, the Himalayas, the Antarctic Peninsula, and northwestern North America. The reconstructed IPO is annually resolved and dates to 1450 CE. The IPO index compares well with observations during the instrumental period and with paleo-proxy assimilated datasets throughout the entire record, which indicates a robust and temporally stationary IPO signal for the last ~550 years. Paleoclimate reconstructions from the tropical Pacific region vary greatly during the Little Ice Age (LIA), although the reconstructed IPO index in this study suggests that the LIA was primarily defined by a weak, negative IPO phase and hence more La Niña–like conditions. Although the mean state of the tropical Pacific Ocean during the LIA remains uncertain, the reconstructed IPO reveals some interesting dynamical relationships with the intertropical convergence zone (ITCZ). In the current warm period, a positive (negative) IPO coincides with an expansion (contraction) of the seasonal latitudinal range of the ITCZ. This relationship is not stationary, however, and is virtually absent throughout the LIA, suggesting that external forcing, such as that from volcanoes and/or reduced solar irradiance, could be driving either the ITCZ shifts or the climate dominating the ice core sites used in the IPO reconstruction. 
    more » « less
  5. “Active structures” are physical structures that incorporate real-time monitoring and control. Examples include active vibration damping or blast mitigation systems. Evaluating physics-based models in real-time is generally not feasible for such systems having high-rate dynamics which require microsecond response times, but data-driven machine-learning-based models can potentially offer a solution. This paper compares the cost and performance of two FPGA-based implementations of real-time, continuously-trained models for forecasting timeseries signals with non-stationarities, with one using HighLevel Synthesis (HLS) and the other a programmable overlay architecture. The proposed model accepts a uni-variate vibration signal and seeks to forecast future samples to inform highrate controllers. The proposed forecasting method performs two concurrent neural inference operations. One inference forecasts the state of the signal f samples into the future as a function of the most recent h samples, while the other forecasts the current sample given h samples starting from h + f − 1 samples into the past. The first forecast produces the forecast while the second forecast allows the system to calculate the model’s loss and perform an immediate model update before the next sample period. 
    more » « less