Following the Flow

February of this year brought record-breaking rainfalls to California when two back-to-back atmospheric rivers rolled in from the Pacific Ocean, delivering half a year’s worth of rain in just a few days. A state of emergency was declared in multiple counties as hurricane-force winds, flash floods, mudslides, and power outages wreaked havoc up and down the coast. Nine people died.

In 2017, Hurricane Harvey dumped as much as 60 inches of rain on coastal Texas over the course of four days, killing more than a hundred people. Harvey tied 2005's Hurricane Katrina, which killed more than 1800 people, for the costliest tropical cyclone on record with an estimated 125 billion dollars in damages.

Weather-related catastrophes like these are common and getting more so. Coastal urban areas sit at the precarious confluence of ceaseless urbanization and accelerating climate change. Urbanization changes how water moves through the landscape—paved streets and city blocks reduce permeation and increase runoff. Meanwhile, climate change is driving water availability to dangerous new extremes—too much brings floods and landslides, while too little brings drought and wildfire. Nearly half of the global population lives within 100 kilometers of a coast and 90 percent of the world’s international trade relies on coastal infrastructure. How can coastal population centers best prepare for rising seas and intensifying storms? Better modeling of the world’s water is a key starting point. 

Los Alamos modeler and Earth-system scientist Rich Fiorella specializes in the hydrologic cycle, the continual circulation of water through land and air by way of alternating precipitation and evaporation. He is building a new computer model to track water as it moves around the globe, with the aim of improving how coastal urban hydrology is modeled.

Coastal urban areas present specific challenges for representing hydrologic processes within Earth-system models (ESMs), which drive the computer simulations that forecast weather and climate change. As computational power has grown, ESMs have increased in complexity and resolution, but gaps remain. For example, the way water movement is generally modeled in ESMs only includes mass and direction, so the details of these masses and their origins can be forgotten. A large fraction of precipitation over land originates from evapotranspiration over land somewhere else. These interactions and how they change with climate are critical to understanding future risks to food and water security or from extreme climate events. Fiorella is building a new water-accounting capability for ESMs that will fill some of these gaps.

“Climate change is altering the water cycle, affecting everything from coastal infrastructure to food and water security, to the health and resilience of the ecosystems that we all depend on,” explains Fiorella. “My work looks at how water moves through the Earth system to understand how we can best manage this critical resource.”

Coastal urban areas sit at the precarious confluence of urbanization and climate change

Water works

The origins of ESMs lie in earlier generations of climate system models: general circulation models (GCMs). GCMs were built on mathematical representations of the physical processes that move mass and energy around the planet. ESMs go beyond GCMs by also seeking to represent the biological and chemical processes that shape our planet, thereby providing a tool to understand how human choices influence or are best guided by environmental variability and change.

The Energy Exascale Earth System Model, or E3SM, is a specific supercomputer ESM that is funded by the Department of Energy for the investigation and simulation of energy-relevant processes around the globe. Many Los Alamos scientists, including Fiorella, are contributing to this effort. 

The E3SM is made up of various submodels, each of which models a specific piece of the picture—called a domain—like the atmosphere, land, rivers, or the ocean. Hydrologic processes in the various domains, from cloud formation microphysics in the atmosphere to how permafrost in the Arctic melts, are described by a complex set of interacting equations. The mathematical relationships must be fairly good approximations of the real-world relationships, otherwise it’s not a good model.

One of the key elements of model building is parameterization. This is essentially the inclusion of processes which, for one reason or another, aren’t explicitly resolved by the model itself. The value of each parameter is assigned by model developers, based on, ideally, high-quality data. 

A common reason for parameterizing something is that the real-world thing is too small to be resolved by the model. E3SM submodels typically break the world into a grid composed of square cells on the order of 100 kilometers on a side, so anything smaller than that—like many urban areas and hydrological processes—won’t be very well resolved. (Bringing the grid-cell size down is an active area of work within the E3SM team, but parameterization will always be an inherent part of ESMs.)

“If a cloud is forming in the atmosphere model, for example,” explains Fiorella, “we have to make assumptions about the processes that form cloud droplets because these are too small to be resolved by the model. Because many of these processes are occurring at much smaller scales than the model grid cell, we must develop simplified representations called parameterizations. But since the observational or experimental data used to develop parameterizations are based on just a small set of the possible conditions on Earth, we must find ways to evaluate how well these parameterizations work across different environments.”

Fiorella wants to make parameterization within E3SM more robust to changing conditions, and one way he thinks he can do it is by turning convention on its head.

To understand how water is currently tracked in most ESMs, imagine a bucket out in a rainstorm. The models can tell you how much water is in the bucket, but they won’t tell you much about where each drop in the bucket originated or how it is connected to other Earth-system processes. One approach that is commonly used to examine how water moves through the hydrological cycle in ESMs is to use location-based water tracers, a numerical tool that assigns traceable tags to quantities of water so the quantities can be tracked as they move about within the model, between domains (atmosphere, land, rivers, etc.) and grid cells.

The location-based approach to water tracing allows a researcher to say, “Ten percent of the water in the bucket came from this region, twenty percent came from that region.” But if there are a lot of regions, then there are a lot of tags, and the model becomes slow and expensive to run. Also, the researcher must know in advance which areas contributed water to the bucket. Fiorella and his collaborators are moving away from these location-based tags by instead resolving how water moves through the model based on particular processes—like how rain from an individual storm moves through the land surface: Does it enter a river or is it used by a plant? And does it end up contributing to an urban flood, being discharged into the ocean, or evaporating into the atmosphere? 

"Location-based tags are good for looking at what is changing, while process-based tags are much better at looking at why."

The advantage of process-oriented tracking over location-oriented tracking is that it’s more flexible. For location-based tracking, the model either must include all locations, which slows it down, or the user must know in advance which areas to focus on. With process-based tracking, Fiorella thinks he can get much of the same information with a fraction of the input, making it faster, and he can also ask the model different kinds of questions.

"Location-based tags are really good for looking at what is changing," he explains, "while the process-based tags are much better at analyzing why the water cycle is changing."

Informed by isotopes

Different physical or chemical qualities can be used as water tracers, the numerical tags that track water through ESMs, but Fiorella’s project uses the natural phenomenon of stable isotopes. Isotopes of an element have the same number of protons but different masses due to differing numbers of neutrons. Take oxygen, for example: the dominant isotope of oxygen is oxygen-16 (¹⁶O), which has 8 protons and 8 neutrons in the nucleus of each atom, for an atomic mass of 16. But the isotope oxygen-18 (¹⁸O) has two additional neutrons, and therefore has an atomic mass of 18. Some isotopes are unstable and decay at predictable rates into other isotopes or even other elements, but both ¹⁶O and ¹⁸O are stable, meaning they don’t change. The proportions of isotopes present in a sample are signatures that scientists can use to answer questions about the origin and history of that sample.

A famous example is the case of Ötzi the Iceman, a 5300-year-old natural mummy found in 1991 in the Alps near the Italy-Austria border. Stable isotope analysis helped scientists learn about Ötzi’s life. A water molecule contains a single oxygen atom, which can be either ¹⁸O or ¹⁶O (¹⁷O is also a stable isotope but it is less often used in isotope analyses). During Ötzi’s childhood, oxygen atoms from the water he drank were assimilated into his tooth enamel and preserved there. The ratio of ¹⁸O to ¹⁶O, still present today in his mummified teeth, was compared with ratios in freshwater sources around the region where he was found, and the best match came from the province of South Tyrol, in northern Italy. It was a signature that scientists used, along with other isotope signatures, to conclude that’s where Ötzi most likely came from.

Here’s how that works: The ratio of ¹⁸O to ¹⁶O in a water sample, denoted in relation to a standard isotope ratio as δ¹⁸O, is formed by a process called water isotope fractionation, which is the result of basic physics. The difference in weight between ¹⁸O and ¹⁶O, just two neutrons’ worth, affects the energy input needed for a phase change. Because H₂¹⁶O molecules are lighter, they require less energy to go from liquid to vapor, so during evaporation H₂¹⁶O molecules are the first to take off, leaving more molecules of the heavier H₂¹⁸O behind. This partitioning is strongly related to temperature and other environmental factors, so, δ¹⁸O measurements can be used to back-calculate the conditions under which the most recent phase change occurred. 

Water isotope fractionation creates strong patterns across space and time. When it rains, H₂¹⁸O molecules are the first to go from vapor back to liquid water, so coastal and low-elevation areas, the places where rain clouds rain first, get more H₂¹⁸O, while areas farther inland, upland, or at higher latitudes, receive more H₂¹⁶O because much of the water has already been wrung out. The physics dictating water isotope fractionation are the same now as they were in Ötzi’s days, so his drinking water sources should have the same δ¹⁸O signature today as they did when he was alive.

Hydrogen, the other element in water, also exists in nature as various isotopes, and interest in using water isotopes as climate tracers goes back to the 1950s. Atmospheric tests of thermonuclear devices during that time caused a spike in environmental tritium (a hydrogen isotope with two neutrons, or 3H). The International Atomic Energy Agency and the World Meteorological Organization realized that they could both make use of variations in water isotope ratios—for test monitoring and climate science, respectively—so together they created the Global Network for Isotopes in Precipitation in 1961. Today it has the world’s largest and most comprehensive collection of water isotope data for precipitation.

The isotope ratios present in a sample are signatures that scientists can use to answer questions about its origin and history

However, despite more than sixty years of data collection and well-understood fractionation physics, the interpretation of water isotope ratio (IR) data isn’t always straightforward. The main challenge is connecting IR data unequivocally to specific real-world processes, and while it’s not always clear which process or which combination of processes produced which IR, it also presents an opportunity. If the model’s IR output matches the observational data, chances are good that the model is representing the entirety of the water cycle accurately, or close to it. By using the model to simulate IR data—input parameter values, apply prescribed fractionation physics, see what the IR output is—Fiorella can test hypotheses about which processes produce which IRs. 

“Isotope ratios allow us to test how closely our water cycle model is representing real world processes,” he says. “Because water IRs reflect many different processes influencing evaporation, transport, and condensation, the model has to get a lot right before the modeled IRs will match observed values. At the same time, individual IR observations can be consistent with multiple interpretations. The new tracer framework we are developing can help us determine which interpretation is correct.” 

Fresh ideas

Fiorella used to travel around collecting water samples and determining their IRs, but today he is primarily a modeler. He aggregates IR observational data, from a variety of sources including satellite-based collections and the Global Network for Isotopes in Precipitation, and uses these data to evaluate his process-based water tracing code.

The details of building the water tracing code presently occupy most of his time. For hours on end, one keystroke at a time, with ambient music in the background, Fiorella incrementally inches toward something that will make the already impressive E3SM better. 

“Model development can be a lot of trial and error,” he says. “But it is really rewarding when you see a new capability come to life. E3SM already has several amazing capabilities and was recently recognized with a Gordon Bell Prize for achievements in high-performance computing. My contribution through this project will be to add a new capability that allows us to understand the water cycle in a more detailed, process-driven way.” 

Variations in water isotope ratios are useful for both nuclear test monitoring and climate science.

The capability he is building will go through three phases. First, the ability to simulate water IRs will provide a new perspective on how well the model represents the hydrological cycle, as the model will need to get the underlying hydrologic processes mostly correct before it will match observational IR data. To the extent that modeled IRs match observations, scientists can trust that E3SM is capturing how water moves through natural systems. Second, an extension of the water IR code will allow him to look at hydrologic processes in greater detail and better understand how water moves through the E3SM. The process-based code will also include better connections between the atmosphere, land, river, and ocean domains. Finally, Fiorella and his team will apply these capabilities to ask more nuanced questions about how the hydrologic cycle is changing and how these changes might affect urban and coastal systems and communities.

Questions like, “What if southern California has more frequent back-to-back atmospheric river events?” Or, “What if Hurricane Harvey had not stalled over Houston, and instead continued to travel steadily inland?” Or, more broadly, “How can the human world be adjusted so that water doesn’t cause us so much trouble?”

"How can the human world be adjusted so that water doesn’t cause us so much trouble?"

The benefit is not limited to E3SM. The tools Fiorella is developing can also be applied in other models, to look at smaller scales. “The same concepts can be applied to specific watersheds,” he explains, “to see how much precipitation is entering streams compared to how much is evaporating from soils or through plants, for example, and how these relationships change across seasons.” 

ESMs are humanity’s best tools for preparing for catastrophic weather events. We’ve built our cities near water, working to harness it and redirect its power from paths of destruction to channels of industry. Now, as climate change and urbanization continue full tilt, so too will the problems they cause. Scientists are building models to help decision makers understand the risk to, and resilience of, coastal cities, and they need now more than ever to understand the world's water. LDRD