by Judith Curry
There is growing interest in the scientific, operational and applications communities in developing forecasts that fill the gap between medium range weather forecasts (up to two weeks) and seasonal forecasts (3-6 months).
Weather forecasts on timescales of days to season can be used to reduce vulnerability to weather variability as well as capitalize on opportunities. Substantial progress has been made in recent years on the development and applications of medium-range weather forecasts and seasonal climate predictions. However, forecasting on the subseasonal time scale (two weeks to two months) has received much less attention, in part because this time horizon has been considered a ‘predictability desert’. From the perspective of business, weather forecasts on the sub-seasonal time scale provides an opportunity because it lies between the well-established application of daily weather forecasts and the increasing use of seasonal forecasts. Many decisions fall into the intervening two-weekly to two-monthly time scale, so the application of subseasonal forecasts provides the potential to augment actionable forecast information.
Many numerical weather prediction centers now use coupled ocean-atmosphere models to produce ensemble forecasts on the subseasonal time scale. There is now a significant opportunity to develop methods that use subseasonal forecasts to provide actionable information. Probabilistic forecasts can be used to develop decision rules and hedging strategies, identify risk of exceeding critical thresholds, and support cost/loss scenarios and analysis. Actually realizing the potential value of such information, however, depends on the sensitivity to particular weather events, their capacity to act to avoid losses or enhance benefits, and the ability of probabilistic predictive information to influence their decisions.
Under the auspices of the World Meteorological Organization, there is a new initiative Subseasonal to Seasonal Prediction Research. A reader’s digest version can be found [here].
From the Executive Summary of the WMO doc:
The subseasonal to seasonal timescale provides a unique opportunity to capitalise on the expertise of the weather and climate research communities, and to bring them together to improve predictions on a timescale of particular relevance to the Global Framework for Climate Services (GFCS).
For NWP forecasts, model error is not usually so dominant that a reforecast set is needed but for the subseasonal to seasonal range model error is too large to be ignored. Therefore an extensive reforecast set spanning several years is needed to calculate model bias, which in some cases can also be used to evaluate skill. Careful calibration and judicious combination of ensembles of forecasts from different models into a larger ensemble can give higher skill than that from any single model. Comparing, verifying and testing multi-model combinations from these forecasts, quantifying their uncertainty as well as the handling of such a massive dataset will nevertheless be challenging. An important aspect will be to promote use of these forecasts and their uncertainty estimates by the applications community.
Needs and Applications
From the WMO doc:
Weather and climate events continue to exact a toll on society despite the tremendous success and investment in prediction science and operational forecasting over the past century. While many end-users have benefited by applying weather and climate forecasts in their decision-making, there remains ample evidence to suggest that such information is underutilized across a wide range of economic sectors. This may be explained partly by the presence of ‘gaps’ in forecasting capabilities, for example at the subseasonal scale of prediction, and partly by a lack of understanding and appreciation of the complex processes and numerous facets involved in decision making.
In the context of humanitarian aid and disaster preparedness, the Red Cross Climate Centre/IRI have proposed a “Ready-Set-Go” concept for making use of forecasts from weather to seasonal, in which seasonal forecasts are used to begin monitoring of subseasonal and short-range forecasts, update contingency plans, train volunteers, and enable early warning systems (“Ready”); sub-monthly forecasts are used to alert volunteers, warn communities (“Set’); and, weather forecasts are then used to activate volunteers, distribute instructions to communities, and evacuate if needed (“Go”). This paradigm could be useful in other sectors as well, as a means to frame the contribution of subseasonal forecasts to climate service development within GFCS.
Examples of possible applications/users include: warnings of the likelihood of severe high impact weather (droughts, flooding, tropical and extratropical cyclones etc.) to help protect life and property; humanitarian planning and response to disasters; agriculture and disease planning/control (e.g., malaria and meningitis), particularly in developing countries; river-flow and river-discharge for flood prediction, hydroelectric power generation and reservoir management; landslides; coastal inundation; transport; power generation; insurance.
JC note: From the perspective of my company Climate Forecast Applications Network, I am particularly interested in business applications of the sub seasonal forecasts. Here are some sectors that we are investigating:
Energy sector. The nation’s energy companies comprise the primary sector engaging the private sector meteorology industry. Weather is a primary driver for commodity prices in energy, having an impact on both energy production and consumption. Improved forecasts on subseasonal timescales would support hedging for anticipated energy demand, managing and protecting distribution and transmission infrastructure, and weather related energy trading opportunities and risks. The growth of renewable energy is providing new challenges and opportunities for applications of weather and climate forecasts. As of 2013, renewable energy made up 12.9% of the U.S. domestically produced electricity. However, goals of 80% clean energy production for the U.S. by 2035 imply substantial increases in hydropower, wind and solar energy production. On subseasonal timescales, probabilistic predictions of wind, solar and hydropower generation can help stabilize energy costs and supply by improving scheduling and trading, maintenance scheduling, reducing curtailments and imbalance penalties, improving decisions about reserve energy sources, maximizing grid integration, and planning capacity commitments. Specific groups from the commercial sector that would benefit from subseasonal forecasts include energy trading firms, regional power generators/suppliers, and investors.
Agricultural sector. Weather forecasts support operational decision making on the timing of cultivating, irrigating, spraying, harvesting. Seasonal forecasts support strategic decisions regarding crop cultivar selection and intended acreage for planting. Subseasonal weather forecasts present a specific opportunity to bridge the gap in these two time frames. Viable forecast information beyond the traditional 10 day window can extend the time horizon for agricultural commodity price analysis and forecasting, and so support farmers’ decisions about production, storage and marketing, as well as logistical decisions in dealing with regional shortfalls and excess product availability. For commodities with futures markets, subseasonal weather forecasts can support hedging strategies. Futures, forward contracts and hedging are a prevalent practice with agricultural commodities. A subseasonal decision support system has the potential to help users better navigate what is often a volatile agricultural commodity marketplace and reign in risk exposure faced by agricultural producers and suppliers. It also has the potential to help various participants in the agriculture cycle more intelligently join in the appropriate risk management markets via the extension of reliable outlooks beyond the current limited time scales.
Retail sector. I have given this sector less thought, but I would imagine there would be applications for stocking retail stores, particularly stores like Home Depot who need to stock for hurricanes, snow fall, plus timing of seasonal changes. There’s a company Planalytics that focuses on this sector, some text from their web page:
The weather is affecting your business more than you realize and continues to do so each and every day. However, attention to the weather almost always ends up being reactionary and temporary. Removing weather volatility from last year’s sales typically returns 10-40 basis points of total topline revenue (through the reduction of inventory carrying costs and lost sales alone). This means a $10 billion department store chain could capture $25 million in annual savings; a $900 million specialty retailer can easily save $3 million each year; and so on. Last year’s weather is “baked into” last year’s sales data and it’s costing you money. It’s an unintentional error that leads to unwanted surprises. When you look at your business as a whole, the weather only repeats itself year-to-year about 15-20% of the time. If the market-by-market, week-by-week weather volatility is left untouched when building demand forecasts, you are essentially expecting last year to happen again. It rarely does. With Planalytics’ Weather-Driven Demand (WDD) calculations you can “deweatherize” your sales history, knowing exactly when, where and how much weather impacts sales before planning for the selling season.
Sources of predictability
From the WMO doc:
Short to medium-range weather prediction is considered to be mainly an atmospheric initial value problem. The estimated limit for making skilful forecasts of mid-latitude weather systems is about two weeks, largely due to the sensitivity of forecasts to the atmospheric initial conditions. Subseasonal predictions, on the other hand, benefit from both atmospheric initial conditions and factors external to the atmosphere, such as the state of the ocean, land, and cryosphere.
Processes internal to the atmosphere including the Madden-Julian Oscillation (MJO) and low-frequency atmospheric patterns of variability also contribute significantly to the predictability. Furthermore, in a subseasonal forecast, some kind of time average (e.g. weekly or pentad mean) is usually used, which removes part of the weather noise. Therefore, it is reasonable to expect subseasonal forecasts i.e. beyond the traditional weather forecast limit of two weeks, to have useful skill. At this time range the forecasts have to be probabilistic.
Sources of subseasonal predictability come from various processes in the atmosphere, ocean and land, although they are not yet fully understood. A few examples of such processes are:
1) The MJO: as the dominant mode of intraseasonal variability in the tropics that couples with organized convective activity, the MJO has a considerable impact not only in the tropics, but also in the middle and high latitudes, and is considered as a major source of global predictability on the subseasonal time scale ;
2) Soil moisture: memory in soil moisture can last several weeks which can influence the atmosphere through changes in evaporation and surface energy budget and can affect the forecast of air temperature and precipitation in certain areas during certain times of the year on intraseasonal time scales ;
3) Snow cover: The radiative and thermal properties of widespread snow cover anomalies have the potential to modulate local and remote climate over monthly to seasonal time scales
4) Stratosphere-troposphere interaction: signals of changes in the polar vortex and the Northern Annular Mode/Arctic Oscillation (NAM/AO) are often seen to come from the stratosphere, with the anomalous tropospheric flow lasting up to about two months
5) Ocean conditions: anomalies in SST lead to changes in air-sea heat flux and convection which affect atmospheric circulation. The tropical intraseasonal variability (ISV) forecast skill is found to be improved when a coupled model is used.
Teleconnections- Forecasts of opportunity
From the WMO doc:
Extratropical weather is frequently influenced by recurring circulation patterns, usually referred to as flow regimes or modes of variability. Examples of such circulation patterns include the Pacific-North American pattern (PNA), the North Atlantic Oscillation (NAO)/Arctic Oscillation (AO), the East Atlantic (EA), the West Pacific (WP), and the tropical/Northern Hemisphere (TNH). The circulation patterns are usually associated with global teleconnections as in many cases propagation of Rossby wave trains is involved and the atmospheric variability in one place is related to a forcing in another. Because of their large scale and low-frequency nature, the circulation patterns contribute greatly to the atmospheric predictability on the subseasonal time scale.
The strength of planetary-scale teleconnections with both ENSO and the MJO and other sources of subseasonal and seasonal predictability raise the possibility of important windows of opportunity for skilful subseasonal to seasonal forecasts when and where these teleconnections are active and interacting. Such targeted “forecasts of opportunity” would represent a departure from the usual practice in seasonal forecasting where skill levels are averaged across all reforecasts for a particular season and start date, and might spawn a substantial research effort needed to properly represent and convey the conditional skill of such forecasts, perhaps in terms of spread-skill relationships.
JC note: to use a poker analogy, the secret to winning is to know when to ‘hold’ vs when to ‘fold’ (i.e. potentially actionable forecast information versus no useful forecast for that particular period).
JC comments
For some background on ensemble weather forecast, see my previous post How should we interpret an ensemble of models. Part I: Weather models. Also, Wendy Parker has written a very nice non technical article that explains ensemble weather and climate forecasts [link]. If you are looking for a more technical explanation, here is an overview paper by Tim Palmer [link].
I think that focusing on the sub seasonal time scale is important for several reasons:
- there is untapped predictability, with potential for substantial socioeconomic benefits
- to better understand and predict at the timescales of climate change, we need to work up ladder of the timescales, and figure out how to predict at the sub seasonal and seasonal time scales.
The question then becomes how the scientific and funding priorities should be rebalance to to bring a focus to this time scale. I wrote a previous post on this: Climate versus weather prediction: should we rebalance?
My motivation for this post at this time is that I am in the process of writing a proposal about applications of subseasonal weather forecasts. I would appreciate hearing about any ideas you have regarding applications.
Filed under: climate models