How Climate Models Predict Global Warming and Drought

During El Niño winters, the probability of having a rainy winter in Southern California increases. However, this does not guarantee a rainy winter.

The probability of a rainy winter is substantially enhanced during a strong El Niño, especially in Northern California. Conversely, the probability of a rainy winter in Northern California remains unaffected during La Niña events, while it decreases during La Niña years for Southern California.

Climate Models and Construction

Constructing a Climate Model

How are climate models constructed?

A typical atmospheric GCM (General Circulation Model) grid features the following characteristics:

• Each grid cell has a single value for each variable (temperature, moisture, wind, and pressure).
• The vertical coordinate follows Earth’s surface topography (bending over terrain and mountains), meaning vertical spacing between layers varies.
• Each grid cell exchanges mass, energy, and moisture (fluxes) with its neighboring cells.
• Physical equations are applied to each grid cell.
• The model moves forward (simulates) one time step at a time (e.g., a 100-year simulation at a 15-minute time step requires 4 million steps).

Finer, more detailed resolution provides better details but comes with a higher computational cost.

Computational Time Formula:
Computational time = (computer time per operation) × (operations per equation) × (# equations per grid-box) × (number of grid boxes) × (# of time steps per simulation)

Increasing resolution increases the number of grid boxes and decreases the time step. For example, halving the horizontal grid size requires halving the time step to accurately capture time evolution. Doubling the resolution in x, y, and z increases the grid cells by 2 × 2 × 2 = 8, and doubles the time steps, increasing the total computational cost by a factor of 16.

How well do current climate models perform?

High resolution in climate models involves tradeoffs: it offers more realistic regional details but demands much higher computational costs.

Consider these two conditions:

1. Estimating global surface temperature by the end of the 21st century: Run as many simulations as possible using a low-resolution climate model to counter uncertainties associated with future greenhouse gas (GHG) emissions, natural climate variability, and cloud feedbacks. This forms an ensemble to provide confidence in the estimate.
2. Prioritizing detail over uncertainty: For regional impacts, details are crucial. For example, snowpack in the Sierra Nevada depends strongly on regional topography, making regional details more important.

Downscaling Climate Change for Regional Impacts

Using nested domains, a model starts with a large, low-resolution outer region and then zooms into a smaller, high-resolution region. The regional model requires boundary data to know what air, moisture, or energy enters and exits the domain.

Parameterizing Small-Scale Processes

Mixing and Surface Fluxes:

• Evaporation fluxes depend on the difference between the saturation humidity at the surface and the actual humidity (q) in the lowest atmospheric level.
• Vertical mixing (net moisture flux across the boundary between grid boxes) is proportional to the difference in q between level k and level k-1.

Climate models cannot directly resolve many small-scale processes like clouds, turbulence, and precipitation (sub-grid scale processes). Parameterization is used to estimate their average effects within each grid.

Moist convection is parameterized by determining a rising air parcel’s buoyancy:

• Unstable: The air parcel is warmer than the surrounding temperature → positive buoyancy → stronger moist convection (upward cloud motion, more condensation, more vertical mixing).
• Stable: The air parcel is colder than the surrounding temperature → weaker moist convection.

Sea Ice Model Processes:

• Energy conservation (determines ice temperature and the amount of ice frozen or melted).
• Mass balance (tracks the amount of ice versus freshwater lost or gained from the ocean).
• Parameterized sub-grid scale processes affect these, such as melt ponds affecting albedo and lead fractions affecting heat exchange.

Land Surface Models: Soil Moisture and Evapotranspiration

Precipitation leads to:

1. Interception: Water directly captured by leaves and branches.
2. Infiltration: Water soaking into the soil.
3. Runoff: Excess water flowing over the land.

Soil Capacity: The capacity of soil to hold water, which varies by soil type.

Leaf Area Index (LAI): A larger LAI indicates more leaves, which increases interception, transpiration, ground shade, and the exchange of water and energy with the atmosphere.

Evapotranspiration: The sum of evaporation and transpiration.

Climate Simulations and Drift

• Spin-up: The adjustment period of a climate model before it reaches an equilibrated, stable climate state. The deep ocean takes centuries to adjust. After spin-up, researchers average the latter part of the simulation to define the model’s normal climate state.
• Climate Drift: The difference between the model’s average climate and the real observed climate (model biases).
• Climate Change Experiments: These focus on the change from the model’s own baseline (future minus baseline). Biases can affect estimated changes; for example, if soil moisture is biased low, the reduction will be underestimated.

Evaluating Present-Day Simulations

Models are based on physical principles and can start from extremely simple initial conditions (e.g., no winds, currents, or precipitation). However, a small global error (such as shifts or extensions) in a climate feature can lead to large regional errors. Therefore, models must be validated against observations for specific climate variables, seasons, and regions.

Water Vapor and Temperature Rise

Increased water vapor is critical for precipitation because warmer air holds more moisture—roughly 7% more per 1°C temperature rise, assuming relative humidity remains stable. This increases evaporation and evapotranspiration, adding more water vapor to the atmosphere. Atmospheric circulation then transports this extra moisture into regions of convergence and rising motion, such as the tropics, where it condenses and produces heavier rainfall. This explains the "rich-get-richer" pattern: wet regions tend to get wetter, while dry subtropical regions may become drier because they continue to export moisture and experience sinking motion.

Projections of global warming responses differ from ENSO predictions due to ongoing radiative forcing by greenhouse gases (GHGs) and aerosols. Slightly different initial conditions yield different time dependencies for natural variability, and varying climate sensitivities to forcing lead to a wide range of projections.

Climate Scenarios for Global Warming

Greenhouse Gases and Climate Forcings

Aerosols: Small particles (notably sulfate aerosols) that exert a net cooling effect by reflecting sunlight. They have short atmospheric residence times compared to long-lived GHGs.

Spatial Patterns of Radiative Forcing:

• Greenhouse Gases: Global, positive radiative forcing (warming).
• Sulfates: Regional, negative radiative forcing (cooling).
• In future scenarios, long-lived greenhouse gases tend to dominate over aerosol cooling, which currently offsets some warming.

Scenarios for future radiative forcing depend heavily on current and future societal choices. A range of scenarios is used to capture these plausible choices, differing slightly across phases of the Coupled Model Intercomparison Project (CMIP).

• Physical Climate Models: Run with predetermined estimates of concentrations.
• Biogeochemical Models: Include cycles like the carbon cycle, running with emissions to simulate the detailed evolution of concentrations.

Commonly Used Climate Scenarios

1. CMIP3 Scenarios:

• A1 Scenario Family: Low population growth, rapid economic growth, and reduced regional income differences. Includes A1FI (fossil-fuel intensive), A1T (green technology), and A1B (balanced energy mix).
• A2: Uneven regional economic growth and high reliance on non-fossil fuels.
• B1: Focus on regional sustainability and environmental solutions ("greenest" path).

2. CMIP5 Scenarios (Representative Concentration Pathways - RCPs):

RCPs are named according to their target radiative forcing level in the year 2100:

• RCP 8.5: High emissions (8.5 W/m² in 2100, ~1370 ppm CO² equivalent).
• RCP 6.0: Stabilization without overshoot (~850 ppm CO² equivalent).
• RCP 4.5: Mid-range stabilization without overshoot (~650 ppm CO² equivalent).
• RCP 3-PD: Peak radiative forcing at 3 W/m² before 2100, followed by a decline (~490 ppm CO² equivalent peak, then declining).

3. CMIP6 Scenarios (Shared Socioeconomic Pathways - SSPs):

• SSP5: Fossil-fueled development with peaking population.
• SSP3: Resurgent nationalism, low international priority for environmental concerns, and slower development.
• SSP2: Continuation of historical social, economic, and technological trends; declining energy intensity despite uneven development; moderate population growth. SSP2-4.5 represents a medium-range pathway comparable to RCP4.5.
• SSP1: A gradual shift toward global sustainability (low-end forcing pathway).

If emissions are not reduced quickly, CO² levels will overshoot stabilization targets, requiring negative emissions technologies (e.g., carbon capture). The faster emissions reach net-zero, the lower the long-term atmospheric concentrations.

Spatial Patterns of Warming Responses

A 30-year average is typically used in time-dependent warming scenarios to smooth out short-term climate variability (like El Niño), making the long-term warming signal clearer and easier to track.

Poleward Amplification of Warming:

1. Snow/Ice Feedback: Operates strongly in polar regions, making regional impacts larger than the global average.
2. Sea Ice Reduction: Leads to greater heat transfer from the ocean to the atmosphere during autumn and winter.
3. Lapse Rate Feedback: The rate of temperature decrease with height (lapse rate) is larger at high latitudes. Consequently, atmospheric warming leads to a larger increase in surface temperature at high latitudes than at low latitudes.

Other Key Spatial Patterns:

• Continents generally warm faster than oceans.
• Warming is seasonally dependent, with winter warming in high latitudes being greater than in summer.
• Regional-scale predictions carry higher uncertainty and show greater variation between models.

Measuring Extremes:

Extremes can be measured by the percentage of days when the maximum daily temperature (TX) or minimum daily temperature (TN) exceeds the 90th percentile of the base period.

The "Rich-Get-Richer" Effect on Precipitation:

In the current climate, moisture evaporated in the subtropics is transported to convective regions in the tropics and storm tracks in the mid-latitudes. Warmer temperatures increase atmospheric moisture, enhancing this transport and leading to higher moisture convergence and heavier precipitation in wet regions. Conversely, warming increases evapotranspiration, which can deplete soil moisture even in regions without significant precipitation changes.

If the standard deviation of daily temperatures remains constant while the mean temperature rises, extreme events will occur much more frequently.

A smaller standard deviation combined with an increased mean results in an even more dramatic shift in extreme frequency. This applies to frost days (which decrease as mean temperatures rise) and precipitation. Higher atmospheric moisture increases the probability of heavy rainfall events, while dry periods between rain events may extend to compensate. Key indices include maximum 5-day precipitation, flood risk, and daily intensity.

Understanding the California Drought

Drought severity and its impacts depend on factors beyond simple precipitation. Warmer temperatures increase evapotranspiration, depleting soil moisture. Key metrics include:

• Potential Evapotranspiration (PET): The amount of evapotranspiration that would occur if there were unlimited soil moisture.
• Palmer Drought Severity Index (PDSI): An empirical measure of soil moisture based on precipitation, temperature, and other factors affecting PET. It is highly sensitive to natural variability, which can lead to unprecedented drought conditions.

The amplitude of natural variations depends heavily on the spatial and temporal averages analyzed. Weather and temperature variability caused by heat transport anomalies tend to cancel out over large regional averages. In contrast, the anthropogenic warming trend has a large spatial scale and becomes much clearer when viewed against the "noise" of large-scale averages.

Natural vs. Anthropogenic Forcing

Detection and attribution studies ask whether observed warming is consistent with natural variability or human-induced forcing. Results show that observed warming is only consistent with models that account for both natural and anthropogenic forcings. The observed temperature increase far exceeds the range of natural variability simulated by climate models.