Q & A with Bruno Sansó

What problem are you working on now? What problem are you trying to solve?

My work in statistical climatology focuses on quantifying the uncertainty involved in estimating different quantities that are important for understanding the climate and its evolution

I have worked on estimating the properties of the climate system, such as how the system is able to absorb heat, or how it would react to doubling the amount of CO2 in the atmosphere, for example. These are properties we can’t observe directly, so we need to infer them from historical observations and from simulations of the system. It’s important to quantify how much uncertainty we attach to our best guesses. I’ve also worked on the problem of blending information from different climate models, and comparing their simulations with the observed record.

Even direct observations are subject to uncertainty. For example, the temperature of the ocean is measured by different types of devices. This results in a heterogeneous body of observations that have different accuracy, and are usually irregularly located. I develop methods that blend all the disparate sources of information to produce a smooth surface, as well as a measure of the uncertainty attached to the estimates. To achieve those goals, we have to deal with the challenge of building probabilistic models that make scientific sense, and then learn the components of those models using large amounts of data.

I’m currently working on two projects with my graduate students and some collaborators in the National Oceanic and Atmospheric Administration (NOAA) and the National Aeronautics and Space Administration Jet Propulsion Laboratory (NASA-JPL).

One of the problems I’m working on is related to the estimation of the amount of rain that is brought to California by atmospheric rivers (AR). ARs are elongated regions of water vapor present in the atmosphere that play an important role in the California water cycle. Once an AR hits the mountains of the California coast, the water vapor cools down and induces precipitation. These extreme weather events are responsible for the heavy rain that is observed during the California winter. ARshave a direct impact on the water supply of the state, so it’s very important to have accurate predictions of any possible changes that might occur in their number and intensity as a result of climate change. We’re exploring this with the help of statistical models that focus on extreme events.

The other problem I’m working on is focused on measuring albedo, which is the fraction of incoming solar radiation reflected by the land surface. Albedo can be a sensitive indicator of environmental changes. An international effort involving the USA (National Oceanic and Atmospheric Administration), the EU (European Organisation for the Exploitation of Meteorological Satellites), and Japan (Japan Meteorological Agency) is acquiring data from five different satellites. The current scope of the project is to cover the period 1995–2017, and the dataset already contains about 12 million files and 270 TB. In addition, there is a network of 2,000 ground-based sites whose data can be used to validate the satellite retrievals.

This problem illustrates the compelling need for statistical approaches that can produce predictions and inferences about massively large data on heterogeneous spatial fields, produced by different sources of information.

Why is this important?

In addition to tackling important problems in climatology, my research also has an impact on developing methods and algorithms to handle large spatial and spatio-temporal data, and on making software packages available for their implementation.

I have recently submitted a couple of technical papers that deal with the problem of interpolating large data fields using irregularly scattered observations. One of the papers considers a fast and efficient approach that accounts for spatially heterogeneous fields using multi-resolution models. The other considers a divide-and-conquer strategy to accelerate the computations needed for spatial interpolation, leveraging a multiprocessor environment.

What are (or could be) the impacts and implications of these advances?

We are about to release general purpose open source packages for spatial interpolation using the methods I just mentioned. We’re working on adapting and extending them to tackle the multi-satellite albedo problem and the description of ARs.

Are there any specific impacts for California?

Providing a high quality unified product for albedo measurements will be important for the climate community in general, not only in California. Understanding the drivers of atmospheric rivers and being able to predict future occurrences are very important for the water supply in California.

I understand you are a member of a new IPCC subgroup. Can you tell us about what this group is doing, and what your role in it is?

I’m a member of the working group on Emerging Data Science Tools for the US Climate Variability and Predictability Program (US CLIVAR). US CLIVAR is a federally-funded research program that contributes to the US Global Change Research Program as well as to the International CLIVAR Program of the World Climate Research Program under United Nations auspices.

The main aim of the working group is to help foster an understanding, adoption, and further development of modern data science tools for the analysis of large-to-massive climate data sets. In other words, the group hopes to provide guidance for and demonstrate the use of state-of-the-art data science methods to improve our understanding of the climate system and its evolution.

How can engineering help mitigate climate change? Should engineers “take on” climate change mitigation, or is climate change a governmental/political and personal (human) problem to solve?

I haven’t worked much in the area of climate change mitigation. My focus has been on using data-based tools to describe, understand, and make predictions related to changes in the climate system. There are massive bodies of data collected by remote sensing devices, as well as from observations collected directly on the ground, oceans, and atmosphere.

The study of those data benefits from having powerful systems to process, communicate, store, and access such data. It also benefits from methods of data visualization, data mining, and learning. These are the areas where modern engineers can have an impact in climate change research.

Video