Factors complicating climate forecasts

The challenges of climate change modeling: "The Earth is an unfathomably complex place, a nesting doll of systems within systems. Feedback loops among temperature, land, air, and water are made even more complicated by the fact that every place on Earth is a little different. Natural variability and human-driven warming further alter the rules that govern each of those fundamental interactions. On every continent except Antarctica, certain regions showed up as mysterious hot spots, suffering repeated heat waves worse than what any model could predict or explain. Across places where a third of humanity lives, actual daily temperature records are outpacing model predictions. And a global jump in temperature that lasted from mid-2023 to this past June remains largely unexplained. Per one researcher: “We have to approximate cloud formation because we don’t have the small scales necessary to resolve individual water droplets coming together." "Similarly, models approximate topography, because the scale at which mountain ranges undulate is smaller than the resolution of global climate models, which tend to represent Earth in, at best, 100-square-kilometer pixels. That resolution is good for understanding phenomena such as Arctic warming over decades. But “you can’t resolve a tornado worth anything.” "Models simply can’t function on the scale at which people live, because assessing the impact of current emissions on the future world requires hundreds of years of simulations. Some variables are missing from climate models entirely. Trees and land have been considered major sinks for carbon emissions. But it is changing: Trees and land absorbed much less carbon than normal in 2023. In Finland, forests have stopped absorbing the majority of the carbon they once did, and recently became a net source of emissions, which swamped all gains the country has made in cutting emissions from all other sectors since the early 1990s. The interactions of the ice sheets with the oceans are also largely missing from models. Changing ocean-temperature patterns are currently making climate modelers at NOAA rethink their models of El Niño and La Niña; the agency initially predicted that La Niña’s cooling powers would kick in much sooner than it now appears they will. "The models may be underestimating future climate risks across several regions because of a yet-unclear limitation. And underestimating risk is far more dangerous than overestimating it. Excerpts from The Atlantic article: Climate Models Can’t Explain What’s Happening to Earth Global warming is moving faster than the best models can keep a handle on. By Zoë Schlanger

Not the day after tomorrow: Why we can't predict the timing of climate tipping points. Uncertainties are currently too large to accurately predict exact tipping times for critical Earth system components like the Atlantic Meridional Overturning Circulation (AMOC), polar ice sheets, or tropical rainforests. These tipping events, which might unfold in response to human-caused global warming, are characterized by rapid, irreversible climate changes with potentially catastrophic consequences. However, predicting when these events will occur is more difficult than previously thought. Climate scientists have identified three primary sources of uncertainty. -First, predictions rely on assumptions regarding underlying physical mechanisms, as well as regarding future human actions to extrapolate past data into the future. These assumptions can be overly simplistic and lead to significant errors. -Second, long-term, direct observations of the climate system are rare and Earth system components in question may not be suitably represented by the data. -Third, historical climate data is incomplete. Huge data gaps, especially for the longer past, and methods used to fill these gaps can introduce errors in statistics used to predict possible tipping times. To illustrate their findings, researchers examined AMOC, a crucial ocean current system. Previous predictions from historical data suggested a collapse could occur between 2025-2095. However, a new study revealed that uncertainties are so large these predictions are not reliable. Using different fingerprints=data sets, predicted tipping times for AMOC ranged from 2050-8065 even if underlying mechanistic assumptions were true. Knowing that AMOC might tip somewhere within a 6000-yr window isn't practically useful, and this large range highlights complexity/uncertainty involved in such predictions. Researchers conclude that while idea of predicting climate tipping points is appealing, reality is fraught with uncertainties. Current methods and data are not up to the task. Research is both a wake-up call and a cautionary tale. There are things we still can't predict, and we need to invest in better data and a more in-depth understanding of systems in question. The stakes are too high to rely on shaky predictions. While study shows we can't reliably predict tipping events, possibility of such events can't be ruled out either. Statistical methods are still very good at telling us which parts of the climate have become more unstable. This incl not only AMOC, but also Amazon rainforest and ice sheets. The large uncertainties imply we need to be even more cautious than if we were able to precisely estimate a tipping time. We still need to do everything we can to reduce our impact on the climate, first and foremost by cutting greenhouse gas emissions and greed of people! Even if we can't predict tipping times, probability for key Earth system components to tip still increases with every tenth of a degree of warming!

Climate whiplash effects due to rapidly intensifying El Niño cycles Our new study, just published in the journal Nature Communications, reveals that the El Niño-Southern Oscillation (ENSO), a key driver of global climate variability, is projected to undergo a dramatic transformation due to greenhouse warming. Using high-resolution climate models, our team of researchers from South Korea, the USA, Germany, and Ireland found that ENSO could intensify rapidly over the coming decades (see Figure below) and synchronize with other major climate phenomena, such as the North Atlantic Oscillation, reshaping global temperature and rainfall patterns by the end of the 21st century. Our study projects an abrupt shift within the next 30-40 years from irregular El Niño-La Niña cycles to highly regular oscillations, characterized by amplified fluctuations in sea surface temperature. In a warmer world, the tropical Pacific can undergo a type of climate bifurcation, switching from stable to unstable oscillatory behavior. This is the first time this type of transition has been identified unequivocally in a complex climate model. The shift in ENSO and other climate modes can be explained by enhanced air-sea coupling in a warming climate, combined with more variable weather in the tropics, which leads to a transition in amplitude and regularity. You can download the full paper here: https://lnkd.in/d6E3UDQ3

In 2006, two climate scientists at Princeton, Isaac Held and Brian Soden, thought they had cracked the code on how global warming would change rainfall patterns around the world. Their discovery was encapsulated in a simple slogan "wet gets wetter, dry gets drier." The basic idea was straightforward. When the atmosphere warms up, it can hold more water vapor, about 7% more for every degree of warming, thanks to something called the Clausius-Clapeyron equation. Think of the atmosphere like a giant conveyor belt that picks up moisture from dry places and dumps it on wet places. Make that conveyor belt bigger and stronger, and it moves even more water from the have-nots to the haves. The Amazon rainforest, already soaking wet, would get even more rain. The Sahara Desert, already bone dry, would become even more parched. The tropics, where it's already hot and humid, would turn more watery. Places like Southeast Asia, the Congo Basin, and northern South America would see their monsoons and rainy seasons become larger. Meanwhile, the subtropics, those regions where deserts naturally form, like the American Southwest, the Mediterranean, southern Africa, and much of Australia, would dry out even more. Mid-latitude areas would see their weather patterns intensify, with some regions getting hammered by stronger storms while others faced longer droughts. Even the Arctic would get wetter as warming allowed moisture to travel further north than ever before. For a few years, this framework seemed to explain everything. Climate scientists had their roadmap for the future, and it was both scientifically sound and easy to communicate. But then something interesting happened when researchers started comparing the theory to actual weather data from around the world. The "wet gets wetter, dry gets drier" rule worked great over the oceans, but over land, things got complicated fast. The "wet-get-wetter, dry-get-drier" scaling does not hold over land. The reason has everything to do with how soil, plants, and the atmosphere interact in ways that the original theory didn't account for. The problem turned out to be that land doesn't just sit there passively like the ocean does. It feeds back water to create rainfall, in whats called the small water cycle. It turned out that when soil dries out, it starts a chain reaction. Dry soils raise the sensible heat flux, which produces a warmer and drier low-level atmosphere and increases the potential evapotranspiration. Dry land actually makes the air above it even hotter and drier, creating a feedback loop that the simple "wet gets wetter" rule couldn't predict. Plants respond by closing their pores to save water, soil reflects more heat back into the atmosphere, and suddenly you have a complex system where the land itself is actively changing the weather patterns above it. The real world out turned out to be more complex than the highly influential models of Held and Soden suggested. They didn't include land-use.

The most sophisticated flood model is only as reliable as its weakest link: the climate signal that drives it. In the world of climate risk, it's easy to be impressed by a high-resolution flood map. But precision is not the same as accuracy. Many models rely on a common shortcut—the "delta change" method—to account for future climate impacts. This approach is built on a dangerous assumption: that the future will be a simple, linear extension of the past. It provides a false sense of security while systematically underestimating the non-linear, extreme events where true financial risk resides. A high-resolution map of a flawed future is a dangerous illusion. We believe a more scientifically robust, stochastic approach is essential. Instead of a single, deterministic forecast, we must model the full probability distribution of outcomes to capture the true, complex nature of the climate signal. In our new white paper, we break down why the scientific integrity of the climate signal—not the resolution of the map—is the most critical component for accurate financial flood risk assessment. Read the full paper here: https://lnkd.in/gSapa4Fm #ClimateRisk #FloodRisk #RiskManagement #FinancialRisk #ESG #StochasticModeling

Functional Generative Networks (FGNs) Weather prediction is challenging because the atmosphere is inherently uncertain, and this uncertainty must be reflected in our models. Forecast uncertainty arises from two sources. 🔴 Epistemic uncertainty stems from incomplete knowledge and the limitations of the model itself. 🔴 Aleatoric uncertainty arises from the intrinsic stochasticity of the weather system. Most machine learning models produce a single deterministic output, but the true objective in weather forecasting is to predict a distribution of possible future states. Classical numerical methods address this by perturbing internal model parameters to generate slightly different physical trajectories. In machine learning, we can achieve an analogous effect by injecting noise into the model through shared conditional normalization, which produces structured, coherent variability rather than unstructured noise in inputs or outputs. This effectively perturbs the function itself, corresponding to sampling from a distribution of neural network weights. Additionally, training an ensemble of independently initialized models leads each to converge to a different local minimum, forming a collection of partially knowledgeable but skillful experts, thereby capturing epistemic uncertainty.

Why was Hurricane Helene such a catastrophic surprise in North Carolina? As Marshall Shepherd wrote in Forbes, rainfall forecasts were accurate but other factors conspired to tragedy. Some of the factors were human: Floodplain development, nearly unavoidable in narrow Appalachian valleys and tolerated, if not encouraged, by our national flood insurance policy. Misplaced confidence that nothing like the 1916 Asheville flood would happen again. Ineffective communication about the danger of the storm as it moved inland. The inability of some people to evacuate because they could not afford the expense or risk losing their jobs. Then there is the North Carolina legislature, which passed laws limiting the state's ability to account for climate change and moderate development on steep slopes. Pressured by real estate interests, in 2011 the legislature eliminated funding for a North Carolina Geological Survey landslide mapping program that could have informed residents, homebuyers, and emergency managers about existing and potential landslide problems in their communities. To their credit, some local jurisdictions and geo-consultants stepped in to help fill the gap. And, the NCGS is mapping landslides again. There were also non-human factors. It's not just getting the rainfall right that's important. It's knowing that antecedent rainfall and soil moisture from events days and weeks before a storm control the way a landscape—in this case flood- and landslide-prone terrain that concentrates water and people in the same places—responds to rain. It's a problem that motivated my PhD research on Ohio Valley landslides during the 1980s and continues to be an important element of my consulting work. In some cases, water pressures in slopes will be sensitive to small amounts of rain. In other cases, not so much. Similar logic applies to flooding. Western North Carolina was primed for disaster. Understanding the role of antecedent rain and soil moisture after the fact isn't difficult. Useful short-term forecasts are conceivable if we know the antecedent conditions before a rainfall event. But, knowing what conditions will be like before a big event 25 or 50 years down the road for long-term planning? That's a challenge. Probabilistic models can help quantify the uncertainty if we're willing to accept likelihoods rather than certainties of occurrence, which we already do with daily weather forecasts. Appalachia will see similar events in the future, maybe not in North Carolina next time, but Tennessee or Kentucky or Virginia or West Virginia or Pennsylvania. Everything we know about climate change tells us the ability of hurricanes and tropical storms to deliver inland rain will increase over time. We know enough about flood and landslide initiation to understand the implications. I hope we will be prepared, not surprised, next time. https://lnkd.in/gt34R_M2

“Limitations of future hydrologic forecasts”   As forced climatic shifting becomes increasingly accepted by a once skeptical water resources industry, the various products being generated should be carefully assessed to see what they actually reveal.   Often, what the developers of these forecasts claim is not really what their results show.  Let me explain.   Since the 80s, climatic modeling has improved.  Improvements have occurred at various spatial and temporal scales, various pre- and post-processing steps, within continually improved process-based coding, and involving various bias corrections, spatial downscaling methods etc.     Unfortunately, many researchers do not stop there. Many take the next logical step and extend their findings to represent a projection of likely water AVAILABILITY in the future.  In some instances, researchers go well into the next century.   As water practitioners know (or should know), water availability is based on many factors; hydroclimatology being only ONE.    Water availability for human and societal use relies on a complex system of water infrastructure, water operations, water laws and regulations, and third-party water agreements, transfers, wheeling contracts, etc.  Without these factors, it doesn’t really matter how much precipitation falls or doesn’t fall.     To date, however, NO ONE has developed a credible forecast of future water infrastructure, water operations, water laws and regulations, and water agreements among the myriad of potential third-party water agreements.    Without accounting for potential future water infrastructure (e.g., dams, canals, tunnels, cross-channel diversions, flood bypasses, MAR storage programs, etc.), water operations (e.g., flow ramping, flood releases, water quality releases, etc.), water laws and regulations (e.g., a new CWA, ESA, Flood Control Act, CVPIA, Reclamation Act, Warren Acts, etc.), and accounting for the many intra- and inter-basin water transfers, sales, storage agreements, etc., that could emerge over the next 80-years, one would be MISSING a significant element of the forecasting calculation.   No modeling team, ANYWHERE, has developed an accurate portrayal of these vital water management requisite elements; nor can it be expected that anyone could … We advise large water companies how to utilize current climate studies, their strengths, weaknesses, and how to best plan for an uncertain future.

🌧️ Can We Really Predict the Monsoon Months—or Even a Year—Ahead? The Reality Behind Climate Models Every year, governments, farmers, water managers, and disaster authorities ask the same question: “Can we predict the monsoon early enough to prepare for floods, droughts, and emergencies?” As someone working closely with water and climate systems, I get this question often—and the answer is both fascinating and complex. 🔍 What Climate Models Can Predict — and What They Can’t Modern climate systems like ECMWF Seasonal Forecasts, NMME, CFSv2, and CMIP6 models allow us to understand large-scale patterns months in advance. They help us anticipate whether the upcoming monsoon might be stronger, weaker, or near normal. But here’s the reality: 👉 We cannot accurately predict the exact rainfall amount, exact timing, or location of monsoon rains 6 months or a year ahead. Why? Because the monsoon is influenced by a chaotic mix of global and regional factors: ENSO (El Niño/La Niña) Indian Ocean Dipole Pre-monsoon heating over the subcontinent Jet stream shifts Himalayan snow cover Madden–Julian Oscillation (predictable only 2–4 weeks ahead) These systems interact in ways that are nonlinear and constantly evolving — making long-range precision extremely challenging. 🌡️ How Far Ahead Can We Predict? Different time horizons offer different levels of certainty: ✔ 2 Weeks Ahead → High Accuracy We can predict low-pressure systems, heavy rainfall spells, and flood risks with strong confidence. ✔ 1–3 Months Ahead → Moderate Accuracy Seasonal forecasts allow us to judge whether monsoon rainfall will be above normal, normal, or below normal. ✔ 6 Months Ahead → Low Accuracy We can only predict broad climate tendencies, mostly tied to ENSO and sea-surface temperature patterns. ✔ 1 Year Ahead → Very Low Accuracy At this range, we’re looking at climate “signals,” not operational forecasts. 🌊 Floods vs. Monsoon: Two Very Different Predictive Systems People often confuse two things: 1️⃣ Flood Return Periods (e.g., 10-year flood) These are statistical probabilities, not forecasts. They help engineers and planners—but they cannot predict when a specific flood will occur. 2️⃣ Monsoon Rainfall Forecasts These are dynamic predictions, influenced by real-time climate systems. Much harder, much more uncertain, especially long in advance. 🎯 So How Accurate Are Our Current Models? Seasonal monsoon category prediction (wet/dry/normal): 40–60% accurate at 4–6 months lead. District or basin-level rainfall prediction: Not reliable beyond 2–4 weeks. Onset, breaks, and active phases: Best predicted using subseasonal (S2S) models 10–20 days ahead. In short: Long-range monsoon forecasting is improving, but precise accuracy is still limited by climate chaos. #NDMA #MoCC #GCISC #Waterresouces #Floods #Droughts #Monsoon

Rapid Surge In Global Warming Mainly Due To Reduced Planetary Albedo -- https://lnkd.in/gSNNuN83 <-- shared technical article -- https://lnkd.in/gDb4a_HF <-- shared paper -- “2023 set a number of alarming new records. The global mean temperature also rose to nearly 1½°C above the preindustrial level, another record. Seeking to identify the causes of this sudden rise has proven a challenge for researchers. After all, factoring in the effects of anthropogenic influences like the accumulation of greenhouse gases in the atmosphere, of the weather phenomenon El Niño, and of natural events like volcanic eruptions, can account for a major portion of the warming. But doing so still leaves a gap of roughly ¼°C, which has never been satisfactorily explained. A team puts forward a possible explanation for the rise in global mean temperature - our planet has become less reflective because certain types of clouds have declined [paper link above.] "In addition to the influence of El Niño and the expected long-term warming from anthropogenic greenhouse gases, several other factors have already been discussed that could have contributed to the surprisingly high global mean temperatures since 2023," says Dr. Helge Goessling, including increased solar activity, large amounts of water vapor from a volcanic eruption, or fewer aerosol particles in the atmosphere. But if all these factors are combined, there is still 0.2°C of warming with no readily apparent cause. "The 0.2°C 'explanation gap' for 2023 is currently one of the most intensely discussed questions in climate research," says [the primary author’] In an effort to close that gap, climate modelers from the AWI and the European Center for Medium-Range Weather Forecasts (ECMWF) took a closer look at satellite data from NASA, as well as the ECMWF's own reanalysis data, in which a range of observational data is combined with a complex weather model. In some cases, the data goes back to 1940, permitting a detailed analysis of how the global energy budget and cloud cover at different altitudes have evolved. "What caught our eye was that, in both the NASA and ECMWF datasets, 2023 stood out as the year with the lowest planetary albedo," says co-author Dr. Thomas Rackow from the ECMWF. Planetary albedo describes the percentage of incoming solar radiation that is reflected back into space after all interactions with the atmosphere and the surface of the Earth. "We had already observed a slight decline in recent years. The data indicates that in 2023, the planetary albedo may have been at its lowest since at least 1940." This would worsen global warming and could explain the "missing" 0.2°C. But what caused this near-record drop in planetary albedo?...” #GIS #spatial #mapping #globe #global #climatechange #warming #remotesensing #PlanetaryAlbedo #Planetary #Albedo #model #modeling #AWI #ECMWF #NASA #weather #energybudget #causation