Graham Feingold and other theoreticians use the new tool to test model parameterizations

LASSO case studies currently bundle data on low clouds, such as these, based on measurements at the ARM Southern Great Plains atmospheric observatory.
LASSO case studies currently bundle data on low clouds, such as these, based on measurements at the ARM Southern Great Plains atmospheric observatory.

To some of us, “lasso” refers to a stiff loop of rope handy for throwing. To statisticians, it is a way of doing regression analysis.

For atmospheric scientists who develop and test models, however, LASSO is a new way to throw a noose around what they call “the cloud problem” and haul it in for a closer look.

The cloud problem is that global and regional models systematically misrepresent the distribution of clouds, which are the main source of uncertainty in simulations of the future earth system.

Properly, LASSO stands for the Large-Eddy Simulation (LES) ARM Symbiotic Simulation and Observation workflow, a way of combining LES modeling with daily observations of shallow clouds made by the U.S. Department of Energy's Atmospheric Radiation Measurement (ARM) user facility. Numerous Atmospheric System Research scientists are already using LASSO in their research, and more are expressing interest.

For decades, LES has been used by atmospheric scientists for the numerical simulation of the turbulent and often cloudy boundary layer over areas 15 to 25 kilometers wide. As computing power improves, LES is more common and easier to do than even a decade ago. However, LES is still often used for idealized simulations rather than real-world conditions.

Within the LASSO framework, LES is combined with ARM observational data on atmospheric conditions and land-atmosphere exchange processes and becomes part of a hybrid simulation of reality that samples a wide range of observed conditions.

LASSO is a proxy for conditions on a high-resolution meteorological scale. It also simplifies data discovery and analysis. Its case studies—bundled into packages of data—ease metadata searches, present parameters on a common grid, and include evaluative skill scores and diagnostics.

Currently, LASSO case studies only relate to low clouds and only represent the behavior of the atmosphere surrounding ARM’s longtime Southern Great Plains (SGP) atmospheric observatory. But after its initial pilot phase, the project will expand to other phenomena and to other ARM sites.

The goal is to someday soon have a deep archive of LASSO case studies useful to observationalists and modelers, as well as to theoreticians in search of streamlined “simple models” that represent the fundamentals of how atmospheric processes emerge and interact.

“LASSO is different things to different people. That’s the beauty of it,” says modeler and theoretician Graham Feingold, a research scientist at the Earth System Research Laboratory in Boulder, Colorado, an arm of the National Oceanic and Atmospheric Administration.

LASSO, which he helped launch, “provides a stream of routine cases everybody can run with their models,” says Feingold. His team is using LASSO forcing ensembles to provide scenarios intended to improve how scientists understand aerosol-cloud interactions.

“Having these soundings and forcing data sets made on a routine basis is exceptionally valuable for us,” he says.

Creating LASSO case studies requires collecting SGP data, initializing the models, and then packaging the data with specific observations made on days when shallow clouds are evident.

There will soon be an archive of about 50 such test cases in hand, with plans for hundreds more. As the library develops “it will make it easier to test model behavior,” says LASSO principal investigator William Gustafson, an atmospheric scientist at Pacific Northwest National Laboratory (PNNL).

Thin clouds are seen over the Kuparuk River in Alaska’s North Slope, near an ARM observatory. LASSO case studies will eventually represent different cloud regimes in other regions.
Thin clouds are seen over the Kuparuk River in Alaska’s North Slope, near an ARM observatory. LASSO case studies will eventually represent different cloud regimes in other regions.

By automating forcing data sets, “Bill (Gustafson) and company have done a lot of work for us,” says Feingold. He adds that by himself “I would only be able to do a few every year.”

Shallow Clouds and Beyond

Shallow cumulus clouds may seem relatively simple, but as with other clouds they are elusive, ephemeral, and hard to estimate at field scales—and are therefore a worthy first target for LASSO. These puffy, rain-free clouds critically affect the Earth’s energy budget, chiefly through modifying albedo, but their small size makes them difficult for global models to represent realistically.

Shallow clouds are a good target for the first phase of LASSO. However, the eventual expansion of LASSO to other cloud regimes is critically important, says Feingold, since a number of common cloud regimes, as represented in models, reflect systematic regional biases (offsets from observations) in cloud radiative forcing.

In a 2009 review article in Nature, Feingold and co-author Bjorn Stevens anticipated the need for something like LASSO. They also identified the cloud regimes “worthy of closer scrutiny:”

  • Shallow maritime clouds are numerous, climate-critical, hard to measure, and a possible pathway for regulating planetary albedo. (ARM also has an atmospheric observatory in the Azores that could someday enable LASSO case studies of maritime clouds.)
  • Polar stratiform cloud regimes mediate interactions in areas particularly sensitive to changing earth system conditions. ARM has an observatory in Utqiaġvik (formerly Barrow), Alaska. In 2019, there are plans to install a mobile observatory on an icebreaker that will drift in the Arctic Ocean ice. Both could be opportunities to extend routine LES to this important cloud regime.
  • Deep convection systems over land are hydrologically critical. This type of cloud regime could be targeted at the SGP as a way of expanding LASSO cases studies.

Proposed in 2013 and summarized in a report last November, LASSO’s vetted simulations will appeal to several audiences.

Observationalists can use them to test retrieval algorithms and measurement strategies. Examples are few so far, but a 2016 paper led by Mariko Oue of Stony Brook University in New York used virtual radars within the LASSO simulations to develop an algorithm that outlines how to deploy scanning radars.

Modelers can use LASSO case studies as starting points for tailored simulations. And theoreticians such as Feingold can use them to test sensitivity to meteorological condition, aerosol, and to assumptions used in model parameterizations.

Bridging Two Views of Complexity

A figure from Feingold’s 2009 Nature paper. The study anticipated LASSO by calling for “comprehensive data sets capable of documenting the behavior of cloud regimes on timescales of days to seasons.”
A figure from Feingold’s 2009 Nature paper. The study anticipated LASSO by calling for “comprehensive data sets capable of documenting the behavior of cloud regimes on timescales of days to seasons.”

In the Nature article nearly 10 years ago, Feingold and Stevens hinted at the need for something like LASSO, a tool that clarifies the interplay of aerosols, clouds, and precipitation.

To do that, says Feingold today, LASSO can be used as a bridge between the precision and small scale of LES and the grander, longer-term predictive power of regional and earth system models that integrate far more processes but at poorer resolution.

His present work involves using LASSO case studies for investigating what he calls “aerosol events” and the co-variability of aerosols and meteorology, both of which influence cloud formation, radiation reflected to space, and precipitation.

“Clouds are driven by meteorology,” he says. Combined with aerosols, the effects can create a stronger effect or be “buffered” to create a weaker effect.

In the Nature article, he and his co-author explained how buffered systems work, and how aerosol effects on clouds and precipitation are related.

In the shallow-cumulus regime mid-continental regions typified by the SGP, if you “hit a cloud system hard” with aerosols pouring in from the industrialized north “sometimes they work together to create brighter clouds,” says Feingold, “and in some cases the opposite.”

On the microphysical scale, Feingold says, “we have a reasonably good understanding of how aerosol particles influence droplet formation and growth. But these effects should be considered on the scale of a cloud field, where aerosols, depending on meteorological conditions, might produce brighter or less reflective clouds—or in the case of precipitating clouds, more rain or less.”

Meanwhile, he credits LASSO for “giving us real-life cases that are going to help us understand whether these systems are buffered or not.”

Newton, Darwin, and a New Synergy

In a 2016 paper in the Proceedings of the National Academy of Sciences of the United States of America (PNAS), Feingold and his co-authors even more explicitly called for something like LASSO.

They urged “routine LES driven by observed simultaneously varying meteorological and aerosol conditions to clarify the relationship between co-variability in aerosol and meteorology.”

Working at the level of both models and observation, they wrote, “will provide further confidence in the fidelity of simulations.”

Given the nearly overwhelming complexity of the cloud problem, such fidelity is welcome. Aerosols alone are complex. Clouds are sensitive to both aerosols and to the vagaries of meteorology. And the co-variability of cloud-aerosol interactions is vexingly complex too.

To address these layers of complexity, Feingold and his co-authors described the philosophical leap necessary to create a model-observation hybrid such as LASSO, which combines two views of complex systems.

The Newtonian view addresses complex systems by seeking the simplicity of universal laws, causality, patterns, and simple models.

The Darwinian view focuses on the complexity of a system, and tends to break the system up into subcomponents, all of which exhibit their own complexity. This view is reluctant to search for underlying simplicity.

As a theoretician, Feingold is drawn to the purity and simplicity of the Newtonian view. As a modeler, he also recognizes the corrective qualities of observations to quantify subcomponents of the system, which reflects a reality that sometimes seems chaotic.

As a source of those observations, he says, “ARM data has been a real boon to our field.”

Hence the synergistic, hybrid power of LASSO, a tool that represents what Feingold and his co-authors called for in the PNAS paper: “something intermediate in character.”

# # #

This work was supported by the U.S. Department of Energy's Office of Science, Office of Biological and Environmental Research as part of the Atmospheric System Research Program.