- The Science
- CLIVAR Research Foci
- CLIVAR Imperatives
- CLIVAR Endorsed Projects & Activities
- CLIVAR Objectives
- CLIVAR Successes
- WCRP Science
- Panels and Working Groups
- PAGES/CLIVAR Intersection Working Group
- CCl/CLIVAR/JCOMM Expert Team on Climate Change Detection and Indices (ETCCDI)
- Global Synthesis and Observations Panel (GSOP)
- WGCM/CLIVAR Working Group on Ocean Model Development (WGOMD)
- Asian-Australian Monsoon Panel (AAMP)
- Atlantic Implementation Panel (AIP)
- CLIVAR-GEWEX Africa Climate Panel (ACP)
- CLIVAR/IOC-GOOS Indian Ocean Panel (IOP)
- Pacific Panel (PP)
- CLIVAR/CliC/SCAR Southern Ocean Panel (SOP)
- Variability of the American Monsoon Systems (VAMOS)
- National Programmes
- Job opportunities
- Early Career Scientists
Frontier 3: Intra-seasonal and seasonal predictability and prediction
Identify and understand phenomena that offer some degree of intra-seasonal to inter-annual predictability, to skilfully predict these climate fluctuations and trends and to increase interactions between scientists, operational forecasters and decision makers
Prepared by: B. Kirtman, H. Hendon, E. H. Berbery and K. R. Sperber
Our ability to predict the seasonal variations of climate dramatically improved from the early 1980s to the late 1990s with the advent of dynamical-coupled model forecast systems, primarily due to their ability to predict El Niño/Southern Oscillation and some of its global teleconnections. After the late 1990s, our ability to predict climate fluctuations reached a plateau with only modest subsequent improvement in skill. This is particularly true for seasonal fluctuations of the monsoon systems throughout the globe.
Robust simulation of subseasonal (or intraseasonal) variability has remained elusive, and in fact difficulties on the sub-seasonal time-scale are likely impacting our ability to predict the seasonal monsoon fluctuations. Given the plateau of forecast quality on seasonal time-scales and the relatively slow progress in improving the representation of sub-seasonal phenomena, new prediction strategies have emerged, largely based on multi-institutional international collaborations. These include: (i) the multi-model ensemble strategy, which may be the best current approach for minimising model error and forecast uncertainty; and (ii) the need to include all the interactions among the relevant components of the climate system (i.e. atmosphere (including stratosphere), land, cryosphere, and ocean) for sub-seasonal and seasonal prediction.
CLIVAR has taken the lead in fostering activities to understand the limits of predictability, such as the Climate System Historical Forecast Project (CHFP). This project represents a major scientific research crosscut throughout WCRP (e.g. interactions among SPARC, CliC, GEWEX, and CLIVAR), and because it includes a subseasonal focus, collaboration between WCRP and WWRP will be enabled. Clearly, continuing and expanding international involvement in this project is a core activity of CLIVAR that has large implications for WCRP.
About a third of the world’s population lives in countries influenced significantly by climate anomalies. Many of these countries have economies that are largely dependent upon their agricultural and fishery sectors. The climate forecast successes of the 1980s and 1990s brought great promise for societal benefit in the use and application of seasonal forecast information. However this promise has not been fully realised partly because there have not been adequate interactions between the physical scientists involved in seasonal prediction research and production, applications scientists, decision makers, and operational seasonal prediction providers, but also because there is still scope for improving forecasts. Improving forecasts, particularly at intraseasonal time scales requires a better understanding of the processes that are involved in initiating and maintaining an intraseasonal signal, and more broadly in accounting all the critical interactions among all the elements of the climate system (ocean-atmosphere-biosphere-cryosphere). Additional considerations include the need for a well-maintained and expanded suite of observations, and the ability to adequately blend (assimilate) the observations with the models to make the best possible forecasts. An additional consideration is the non-stationarity of climate and the role that natural or anthropogenic forcing may play in modifying the predictability of the climate system.
Scientific Background and Major Challenges
Feasibility of seasonal prediction: The feasibility of seasonal prediction rests on the existence of slow, and predictable, variations in the Earth's boundary conditions. Within the paradigm of atmospheric predictability due to external forcing, the potential for skilful forecasts depends on the ratio of the externally forced signal relative to the atmospheric generated internal noise. The majority of external variance is known to originate from tropical sea surface temperature variations. Less is known about the seasonal signals due to other external forcings of the total climate system, such as soil moisture, land use, sea ice, atmospheric chemical composition and aerosols. Additional skill due to atmospheric initial conditions is expected for certain slow modes of the atmosphere (for instance, annular modes), but there is little evidence that atmospheric initial conditions contribute to skill for lead times beyond a few weeks. Seasonal forecast quality can be improved by taking into account processes in the cryosphere, land surface, and stratosphere - including, but not limited to, the following factors which have the potential of improving the predictability of variability at seasonal timescales.
Sea ice: Sea ice is highly coupled to the ocean-atmosphere system from synoptic to decadal timescales with large sea ice anomalies tending to persist due to positive feedback in the ocean-atmosphere-ice system. Sea ice anomalies in the Southern Hemisphere can be predicted statistically at seasonal timescales by a linear Markov model and cross-validation with observed estimates can yield correlations of 0.5, even at 12 month lead times. Land ice and snow cover in the Northern Hemisphere is a highly variable surface condition, both spatially and temporally, and can be related to atmospheric variability.
Soil moisture: Soil moisture anomalies, which can persist for weeks to months, can generate rainfall and air temperature anomalies in transitional zones between wet and dry regions. Other potential land-based sources of predictive skill, in addition to snow cover, are subsurface heat reservoirs and vegetation health (leafiness).
Stratosphere: The stratosphere acts as a boundary condition for the troposphere since the characteristic time scales for stratospheric circulation variations are much longer than those in the troposphere. In particularly sensitive areas, such as northern Europe in winter, model results suggest that the influence of stratospheric variability on land surface temperature can exceed the local effect of SST.
Monsoons: The monsoon regions of the world, where over half of the global population lives, are especially challenging. Interannual variability of the mean monsoon rainfall is relatively low (~10% of the mean), but seasonal variations of the monsoon have a profound impact on agriculture and water availability. Predictable variations of the monsoons associated with El Niño are typically confined to pre-monsoon and post monsoon, while most of the variability of the main monsoon appears to be associated with internally generated (i.e. independent of slow boundary forcing) intraseasonal variations. Monsoon failure (or extreme drought) is often a result of extended intraseasonal monsoon breaks. Monsoon intraseasonal variability is not well simulated or predicted in climate models of the sort used for seasonal prediction, which compounds the problem of trying to predict relatively low interannual variability together with the modest relationship with El Niño. Monsoon variations have also been tied to other components of the climate system that are not well simulated or predicted including local air-sea interactions, and variations of the cryosphere, aerosols, and land surface.
Monsoon variability is associated with the large-scale dynamics, but during its early stages, when the surface is not sufficiently wet, soil moisture anomalies may also modulate the onset and development of precipitation. Likewise, there is evidence that vegetation conditions previous to the monsoon may delay or advance the date of the monsoon onset. When the soil is not too dry or not too wet, the soil conditions can control the amount of water being evaporated, and also can produce fundamental changes in the PBL structure that affects the development of convection and precipitation. Adequate representation of the land surface conditions should be carefully included in seasonal simulations.
Subseasonal prediction: Subseasonal climate prediction is a promising area for evaluating the progressing utility and application of climate forecasts, especially for the many regions of the globe where slow variations of tropical SST may not have a large impact, but where impacts from slow variations of internal modes (e.g. the annular modes and the Madden-Julean Oscillation, MJO) are significant. This is especially true in the monsoons, where seasonal variability and predictability is relatively low but intraseasonal variability is large. Monsoon intraseasonal prediction may be feasible because of the strong signature of the MJO and other Monsoon Intraseasonal Oscillations (MISO), whose intrinsic timescale is much longer than extratropical weather, but much shorter than El Niño and other modes of coupled ocean-atmosphere climate variability. Predictability of monsoon intraseasonal variability (ISV) associated with the MJO and other convectively coupled MISO’s is largely unknown, due mainly to the inadequate simulation of the MJO/MISO, and, in general, the inadequate simulation of the interaction of organized tropical convection with large-scale circulation.
The MJO: A multitude of forecast techniques have been developed to provide experimental forecasts of the MJO. These range from empirical and statistical models, to dynamical forecasts. The approaches include linear regression, linear inverse modelling, frequency-wavenumber decomposition and extrapolation of forecasts, and projection of forecast data onto observed MJO EOF’s. The US CLIVAR MJO Working Group (MJOWG) has established an experimental forecast system based on the combined EOF’s of near-equatorial averaged outgoing longwave radiation and the zonal wind at 200hPa and 850hPa. With the endorsement of the WCRP, seven NWP centers are providing output for making real-time experimental MJO forecasts using this methodology. Thus, for the first time a standard assessment of skill in dynamically forecasting the MJO will be obtained by this effort. Importantly, this multi-national effort will facilitate the benefit of using of a multi-model ensemble to predict the MJO. The newly established WCRP/WWRP MJO Task Force will carry out this work, and that of improving our understanding of basic processes that are important for the initiation and maintenance of the MJO, which is the follow-on of the MJOWG. Additionally, CLIVAR has facilitated cooperation among international partners in the development of (1) an ocean observing system in the Indian Ocean (IndOOS), (2) process studies for better understanding of monsoon intraseasonal variability (e.g. Cooperative Indian Ocean experiment on intraseasonal variability in the Year 2011, CINDY 2001), and (3) numerical experimentation to assess the predictability of the MJO (e.g. Hindcast Experiment for Intraseasonal Prediction). The figure below shows an example of how the MJO impacts week 2-3 forecast quality. Here we see significant improvements throughout the tropics when the initial condition includes MJO information.
ROC area of the probability that precipitation averaged over days 8-21 is in the upper tercile of ONDJFM starts using POAMA (1980-2007) for cases with an MJO in the initial conditions (top) and with no MJO in the initial conditions (bottom). ROC areas sigificant at the 5% significance level using a Mann-Whitney U test are shaded
Model uncertainties: Clearly, the maximum predictability of the climate system has yet to be achieved in operational intraseasonal-seasonal forecasting. Model error, particularly in the tropics, continues to limit forecasting skill and, since not all the interactions in the climate system (such as land-atmosphere interactions or atmosphere-cryosphere interactions) are currently fully resolved, there may still be untapped sources of predictability. Uncertainty due to model formulation can be improved by multi-model methodologies though the approach is currently ad hoc since the choice of models has not been optimised and there is no best strategy on how to combine models. Forecast initialisation with ocean data assimilation also improves forecast quality and coupled initialisation continues to be an area that requires active research. It is an entirely open question of how climate change impacts seasonal prediction. The observational requirements for seasonal prediction are not being adequately met. Dynamical model forecasts can also be improved by calibration and the synergistic use of empirical techniques. CLIVAR is ideally suited and well positioned to foster progress on all of these issues.
Model resolution: Another relevant consideration is that current climate models have been limited by resolution and physical parameterisations in their representations of the statistics of internal atmospheric (e.g. synoptic weather systems and tropical waves) and oceanic (e.g. poorly resolved tropical instability waves) dynamics and therefore the interactions of these intrinsic motions with climate. Moreover, the specification of accurate initial conditions in the full climate system may be critical to accurately capture the high frequency phenomena of relevance (e.g. the dependence of the Madden-Julian Oscillation (MJO) on the upper ocean state). For example, was the failure to capture the early onset and the extreme amplitude of the 1997/98 ENSO event because all the models fail to adequately capture the MJO and the associated sub-seasonal variability? Additionally, it would be desirable to be able to credibly simulate coupled features that depend on the initial state with affects on regional weather and climate such as El Niño or Pacific decadal variability. It would seem that a way forward to improve predictions on timescales longer than weeks would be to better resolve the weather-climate link. The issue then becomes what are the important missing elements of the simulation of day-to-day weather, and what is the best strategy for being able to better represent them in the AOGCMs. Another example indicating the potential importance of ocean resolution is given in the figure below. Here we see that the resolving ocean eddies in climate models significantly improves the climatological rainfall distribution.
30-year mean rainfall simulation from CCSM4 with ocean eddy resolving resolution (shaded, HRC) and eddy permitting resolution (contours, LRC). The bottom panel show observed rainfall for comparison
CHFP experiment: Complete the CHFP core experiment with extension to the challenge of intraseasonal prediction. This extension necessarily needs to be well coordinated with THORPEX activities and YOTC.
Intraseasonal variability: Coordinated and targeted focus on simulation and prediction of slow internal, intraseasonal variability, including the convectively coupled tropical modes such as the MJO but also the extratropical annular modes that may be responding to external forcing. This may require a focused task force or working group that can bring together researchers from WWRP and WCRP.
Climate system interactions: Foster the numerical experimentation on climate system interactions, including modes of variability that are poorly represented that are associated with organised tropical convection. CLIVAR needs to maintain mechanisms to ensure that the pan-WCRP momentum established by the task force for seasonal prediction is maintained.
Data assimilation: Foster research on coupled climate system data assimilation through workshops and meetings.
Model improvements: Accelerate model improvement efforts – this may require the establishment of a task force or working group on tropical biases, for example. The approach envisaged would build on previous activities, and would additionally attempt to encourage observationalists, process modellers and large scale modellers to work together on the problem of how best to develop a multipronged attack for reducing tropical biases in CGCMs. By definition this type of activity would reach across the entire CLIVAR community thus requiring a working group or task force approach.
Enable multi-scale interaction research: The distinction across timescales from weather to climate prediction is becoming more blurred. The incorporation of chemical, hydrological and biological processes into weather and climate models will allow a much broader range of environmental parameters to be forecast, including air quality, flooding, sand and dust storms, and changes in vegetation. Many of the applications and impacts of weather and climate share a common underlying scientific basis. Based on these considerations, CLIVAR needs to work to develop a unified approach to multidisciplinary weather, climate, water, and environmental prediction research. The first steps in this regard are increase collaborations between the TIGGE and CHFP projects.
Climate-system Historical Forecast Project: The impact of the different components of the climate system on seasonal prediction quality remains an area in need of active research, both in terms of initialisation and in terms of model development (e.g. resolution of stratospheric processes, stratosphere-troposphere coupling). The CLIVAR Working Group on Seasonal Prediction (WGSIP) encourages the seasonal prediction community to participate in the Climate-system Historical Forecast Project (CHFP), which is an experimental test bed to assess the untapped potential seasonal predictive skill to be gained by resolving interactions between components of the fully coupled physical climate system.