IRI@AGU: Bridging the Climate-Weather Gap

Andrew Robertson, Senior Research Scientist at the IRI

This post is the last in a series of five Q&As with scientists from the International Research Institute for Climate and Society who will be presenting their work at the annual meeting of the American Geophysical Union in San Francisco December 9 to 13.

Not all climate forecasts are created (and researched) equally, something that climate scientists like Andrew Robertson of the IRI are trying to change. Forecasting abilities and characteristics vary greatly depending on the time and spatial scales being considered. Earth system models are used to project long-term average changes, while numerical weather prediction models can predict the next day’s weather with reasonable accuracy. In between these limits, however, forecasts can be organized around many other time horizons, including medium range weather, subseasonal, seasonal, and interannual to decadal. Forecasts at these scales can be useful for a range of decision-making applications, such as improved agriculture and water management and disaster risk reduction. Robertson and his colleagues are working together to create forecast products that integrate forecasts across these multiple time scales in ways that make them more useful for applications.

Let’s begin with the session you’re convening at AGU entitled, “Subseasonal to Seasonal Prediction: Bridging the Gap between Weather and Climate.” Can you specify what timescales you’ll be exploring here? Why is there a gap between weather and climate forecasts?

Here the goal is to improve forecasts between about 2 weeks, up to which we talk about weather, and 2 months, beyond which we are in the realm of climate forecasting where we are limited to broad advice on the upcoming season. We call this interval the subseasonal-to-seasonal range, or S2S for short. The weather-climate gap is partly for historical reasons; weather was tackled first, and then climate researchers discovered sources of seasonal predictability. But the gap also exists because until recently our models have not been very good at the S2S timescales.

Why is prediction needed at this level and how can it be applied?

Early warning of high-impact weather events like a superstorm or typhoon can, for example, help humanitarian aid agencies be more proactive in their planning for disasters. A seasonal forecast could give an early heads up of increased odds of a flooding event happening somewhere over a region and at some point during the season. A subseasonal forecast could provide weekly updates with more specificity about when and where an event may happen. The hope is that mechanisms can be developed between the global centers that produce these forecasts and a range of different end-user groups to deliver a much broader range of weather and climate forecast information that is more targeted and useful than was previously possible.

What are the some of the most promising recent findings in this area?

Today’s models do a much better job of predicting the tropical Madden-Julian Oscillation (MJO), and this is leading to better forecasts beyond 2 weeks. Although the MJO is initiated in tropical Pacific, its effects cascade globally, so forecasts are improving even over North America and Europe.

One of your own posters at AGU is about predicting precipitation at the sub-monthly time scale. How do phenomena such as El Niño Southern Oscillation (ENSO) and Madden-Julian Oscillation (MJO) allow researchers to attempt prediction beyond the typical 7-10 day weather forecast?

These are the best-understood and most predictable climate phenomena on S2S timescales. The hope is that we can improve our forecasts selectively: by looking for the times and places where ENSO and MJO are both exerting a large influence on the local climate, it may be possible to give skillful “forecasts of opportunity.”

What regions do you think might have the best opportunity for predictability at this timescale?

In the poster I’m presenting, we found that the island of Borneo in SE Asia is a promising case. This could be of potential value in some of our fire-risk management work there with colleagues from Bogor Agricultural University in Indonesia.

Your other poster at AGU jumps up a few timescales to the study of the Bhakra reservoir in northwestern India and how climate variability may affect water storage at the interannual to interdecadal scale. For this research, you combine multiple kinds of data, from observations to paleo records to global climate model projections. How do your results at this timescale compare to your results at the submonthly timescale? (e.g., prediction vs. simulation)

Multi-year storage reservoirs like the Bhakra have to be managed in the face of weather and climate variability across time scales ranging from daily weather to interannual climate. So the S2S forecasts could be relevant here too someday. But at the longer end, the decadal timescale starts to overlap with projections of near-term climate change (up to 2050) made with earth system models. Our focus here is to scrutinize the decadal variability and change in these models and compare them with historical data and tree-ring records. Our goal is to combine together these different sources of climate information for reservoir managers in a way that is useable and consistent with the large uncertainties in each of them.

How do the different formats of results change the way that decision makers can use the information?

In the case of the Bhakra dam, with Columbia Water Center we are developing a reservoir management optimization tool that will bring together all of these sources of information in a single format. This tool could be used demonstrate to reservoir managers the potential economic value of climate information across timescales.