1. Widening of Tropical Belt in a Changing Climate: Model Simulations versus Observations
Qiang Fu
Dept. of Atmos. Sciences
University of Washington

2. Enhanced midlatitude warming in satellite measurements
Fu, Johanson, Wallace and Reichler (2006, Science)
Temperature trends
3. We first inferred a widening of the Hadley cell by looking at changes in the meridional temperature gradients from the MSU data. Temperature trends from the MSU show tropospheric warming from X, with bands of enhanced warming in the midlatitudes, near the position of the subtropical jets. The pattern of warming suggests a change in the meridional temperature gradients with reduced gradient on the equatorward side of the jet, and an enhanced temperature gradient on the poleward side of the jet. This means weaker westerlies equatorward of the jet and stronger westerlies poleward of the jet, resulting in a poleward displacement of the jet. This analysis suggested a total widening of the Hadley cell by about 2 degrees since 1979.

3. Changing meridional temperature gradients in the troposphere
Temperature trends
Fu, Johanson, Wallace and Reichler (2006, Science)
Meridional temperature gradients
reduced (equatorward), weaker westerlies
increased (poleward), stronger westerlies
Total widening of Hadley cell ~2°

4. Enhanced midlatitude warming in satellite measurements
Temperature trends (JJA)
3. We first inferred a widening of the Hadley cell by looking at changes in the meridional temperature gradients from the MSU data. Temperature trends from the MSU show tropospheric warming from X, with bands of enhanced warming in the midlatitudes, near the position of the subtropical jets. The pattern of warming suggests a change in the meridional temperature gradients with reduced gradient on the equatorward side of the jet, and an enhanced temperature gradient on the poleward side of the jet. This means weaker westerlies equatorward of the jet and stronger westerlies poleward of the jet, resulting in a poleward displacement of the jet. This analysis suggested a total widening of the Hadley cell by about 2 degrees since 1979.
Fu, Johanson, Wallace and Reichler (2006, Science)

5. Implications of Hadley cell widening
Poleward movement of Hadley cell edges ~ poleward displacement of the subtropical dry zone boundaries
Changes in rainfall patterns at subtropical dry zone margins
May contribute to desertification of marginal lands

6. Hadley cell
The Hadley cell is a large scale meridional overturning circulation driven by differential heating of the Earth's atmosphere. Moist warm air rises near the equator, moves poleward in the upper tropospheric branch of the Hadley cell and descends near 30 degrees.
Meridional overturning circulation driven by differential heating of atmosphere
Transports energy and momentum poleward
Rising, moist air near Equator
Large scale sinking of dry, warm air ~30NS

7. Locating Hadley cell edges
Stream function
Evaporation – Precipitation (E-P)
Hadley cells
OLR contours
Upper tropospheric jets
2a. There are a number of ways to locate the edges of the Hadley cell. We can look at the stream function (the annual mean stream function is shown in this panel...the blue colors show clockwise circulation and the yellow colors show counter clockwise circulation. The Hadley cells are the two cells near the equator). The edges of the Hadley cells are found where the stream function becomes zero, which shows where the poleward transport becomes zero.
2b. Looking at the field of zonal mean wind gives us the position of the subtropical jets, which are located at the edges of the Hadley cells.
2c. From a top down perspective, we can infer the edges of the Hadley cell by finding the edges of the subtropical dry zones. The air in the sinking branch of the Hadley cell is warm and dry, and the large scale subsidence suppresses convection. This figure shows the map of Evaporation minus Precipitation (E-P). E-P is negative in the equatorial regions where precipitation is very large, and then decreases poleward as precipitation decreases. E-P is positive is the subtropical dry zones (shown in yellow), corresponding to the descending branch of the Hadley cell. We can therefore infer the edge of the Hadley cell by locating the edges of they dry zones.
2d. We may also look at outgoing longwave radiation (OLR) contours. OLR is large in the subtropical dry zones because of the general lack of clouds and aridity of the area. It decreases poleward as clouds increase and temperatures drop. One way to infer a Hadley cell boundaries from OLR is to identify the latitude where OLR is 250 W/m2, (shown in white on this plot).

8. Locating Hadley cell edges
Total ozone (e.g., TOMS total ozone, 3/11/1990
Days/year with height > 15 km
Upper tropospheric jets
2a. There are a number of ways to locate the edges of the Hadley cell. We can look at the stream function (the annual mean stream function is shown in this panel...the blue colors show clockwise circulation and the yellow colors show counter clockwise circulation. The Hadley cells are the two cells near the equator). The edges of the Hadley cells are found where the stream function becomes zero, which shows where the poleward transport becomes zero.
2b. Looking at the field of zonal mean wind gives us the position of the subtropical jets, which are located at the edges of the Hadley cells.
2c. From a top down perspective, we can infer the edges of the Hadley cell by finding the edges of the subtropical dry zones. The air in the sinking branch of the Hadley cell is warm and dry, and the large scale subsidence suppresses convection. This figure shows the map of Evaporation minus Precipitation (E-P). E-P is negative in the equatorial regions where precipitation is very large, and then decreases poleward as precipitation decreases. E-P is positive is the subtropical dry zones (shown in yellow), corresponding to the descending branch of the Hadley cell. We can therefore infer the edge of the Hadley cell by locating the edges of they dry zones.
2d. We may also look at outgoing longwave radiation (OLR) contours. OLR is large in the subtropical dry zones because of the general lack of clouds and aridity of the area. It decreases poleward as clouds increase and temperatures drop. One way to infer a Hadley cell boundaries from OLR is to identify the latitude where OLR is 250 W/m2, (shown in white on this plot).

9. Observed Hadley cell widening
Seidel, Fu, Randel, and Reichler (2008, Nature Geoscience)
Also see Hudson et al. (2006), Hu and Fu (2007), Seidel and Randel (2007), Archer and Caldeira (2008)
4a. A number of studies have confirmed this widening. This figure shows the timeseries of Hadley cell width from a number of different observing platforms. A general upward trend in the width is evident in each of these timeseries. The consensus from observations is that the Hadley cell has widened by about 2-5 degrees since 1979.
4b. It is important that we understand the recent widening because it has implications for the global hydrological cycle. As the Hadley cell widens, the edges of the subtropical dry zones move poleward. A poleward displacement is associated with changes in rainfall patterns near the subtropical dry zone boundaries, which may reduce water availability and affect agricultural practices. It may also contribute to desertification of marginal lands.
The cause of this widening is still unknown.
Consensus: Hadley cell has widened by ~3° since 1979

10. Mechanisms for Hadley cell widening
Global warming: wider Hadley cell in warmer world
Held (2000):
Aquaplanet GCMs (Frierson et al. 2007)
Simulations of 21st century climate (Lu et al. 2007)
Warming of the Indo-West Pacific ocean (Lau et al. 2008)
Strengthening of polar stratospheric vortex (Polvani and Kushner 2002)
Tropospheric warming/stratospheric cooling
Altered wind shear across tropopause, increasing phase speed of baroclinic eddies (Chen and Held 2007)
5. Possible mechanisms to explain the recent widening include global warming, and stratospheric cooling. In studies with aquaplanet GCMs and in simulations of the 21st century under global warming scenarios, it was found that the Hadley cell widens with increasing global mean temperature.
The Hadley cell is also wider when the polar stratospheric vortex is stronger, which occurs when the polar stratosphere cools. Low latitude stratospheric warming that occurs during times of solar maximum also results in a wider Hadley cell.
The spatial pattern of warming in the Indo-West Pacific ocean may be an important factor, as well as the change in the zonal mean tropospheric winds, which may increase the phase speed of baroclinic eddies and push the Hadley cell polewards.
No one has compared the observed recent widening of the Hadley cell to model simulations forced with historical changes in forcings, however. The approach of these studies has been to use idealized models, or simulations of future climate.

11. Objective: Comparison of tropical widening between model simulations and observations
Tools
6. Our goal was to perform a direct comparison between the observed widening and model simulations. We wanted to know 3 things: first, is the recent widening due to natural variability. To answer this question, we compared the observed widening to model simulations run with no external forcings. We found that natural variability could not explain the widening. Secondly, we wanted to know if models that were run with historical changes in forcings spanning the same time period as the observations could explain the widening. Surprisingly, none of the model simulations of 20th century climate could explain the widening. Finally, we wanted to compare the observed widening to widening projected over this century. The observed recent widening was found to be much larger than projected future widening.

12. Description of models used to investigate Hadley cell widening
CMIP3 Coupled Model Intercomparison Project
AMIP2 Atmospheric Model Intercomparison Project
Forcings included in 20C simulations
7. We use models submitted to the IPCCAR4 archive. They are the CMIP3 models from the Coupled Model Intercomparison Project. The CMIP3 models provide over 9000 years of pre-industrial control (or PIC) simulations, simulations of 21st century climate based on 3 scenerios: A2 with high radiative forcing, A1B with medium forcing and B1 with low radiative forcing. We also use the 20th century simulations from 24 models with various combinations of external forcings. This table shows the forcings included in the models. For example, they all contain increases in greenhouse gases, about half of them also include stratospheric ozone depletion, and then some of them also contain various aerosol, land use, volcanic, and solar forcings. The other set of model simulations for 20th century climate is from the AMIP2 project. These are atmospheric models that are forced with observed changes in sea surface temperatures.

13. Description of observations used to investigate Hadley cell widening
Latitude where stream function = 0 at 500 hPa from reanalysis
NCEP/NCAR
NCEP/DOE
ERA-40
Latitude where OLR = 250 W m-2 from satellite
HIRS
ISCCP
GEWEX
Hu and Fu (2007)
8. For the observations, we chose to use stream function measurements from 3 reanalysis datasets. This figure shows the zonal mean evolution of the stream function in the Northern Hemisphere. Here, the stream function changes sign between the blue and yellow contours, indicating the poleward extent of the Hadley cell. By tracing that edge in each hemisphere, we obtain a measure of Hadley cell width.
Secondly, we use OLR contours. This figure shows the zonal mean evolution of OLR in the Northern Hemisphere. The Hadley cell edge is taken as the latitude where OLR is about 250 W/m2, which is found between the blue and yellow contours in this figure.

14. Results: time series of Hadley cell width
Johanson and Fu (2008, submitted)
Stream function
OLR
9. This figure shows the timeseries of Hadley cell widths from observations and models. The black lines show the multi-dataset mean timeseries from the observations, based on streamfunction in the left panel and OLR contours in the right panel. The upward trend in the HC widths from observations is much larger than the trend from the multi-model ensemble mean from the 20th century simulations over the same time period (shown in blue—the dark blue is the multi-model ensemble mean and the light blue is a time series from one of the models). The models do show a widening of the Hadley cell, but it occurs over the next 100 years and the rate is slower than what has been observed over the past few decades.
Interannual timescales: variability similar in observations and individual models
Hadley cell width
Observations: increase
20th century simulations: no change
Hadley cell width
A1B: increase
By 2099, Hadley cell much wider than pre-industrial control

15. Observed widening not explained by natural variability
Mean Hadley cell widening
Mean widening
Multi-dataset ensemble mean trends from observations
Ensemble mean trends from samples of 25-yr time periods in PIC
Range of widening
Trend distribution from observational datasets
Trend distribution from samples of 25- yr time periods in PIC
Observed
3° on average
Range °
Pre-industrial control (PIC)
No tenancy to shift Hadley cell towards wider or narrower state
Widening occurs as often as narrowing
Hadley cell width variation within ±1.5°
Distributions of Hadley cell widening

16. 20th century simulations do not reproduce observations
Mean widening
Multi-dataset ensemble mean trends from observations and 20C-A1B from
Range of widening
Trend distributions from observational datasets
Trend distributions from 20C-A1B ensemble members
Widening in 20C-A1B from
Less than 1/5th observed values
At least 97.5% of the 20C-A1B trends from ensemble members show less widening than observed
Mean Hadley cell widening
Distributions of Hadley cell widening

17. Observed widening not explained by pattern of SST warming
Mean widening
Multi-dataset ensemble mean trends from observations and AMIP2 from
Range of widening
Trend distributions from observational datasets
Trend distributions from AMIP2 ensemble members from model end year
Widening in AMIP2
less than 1/4 observed values
Only 2 of 57 realizations had widening larger than 3º
Next largest trend less than 2º
Mean Hadley cell widening
Distributions of Hadley cell widening

18. Observed widening not explained by global warming
1. Hu and Fu (2007)
2. Johanson and Fu (2008)
3. Lu et al. (2007)
4. Frierson et al. (2007)
Rate of widening per degree of global warming is ~10 times larger in observations than models
Global warming alone can't explain recent widening as simulated by GCMs
12. While models do show a widening of the Hadley cell in response to global warming, the rate of widening per degree warming is much smaller in model simulations than in the observations. This figure shows the HC widening per degree of global warming from observations, model simulations of the 20th century, 21st century, and from idealized models. These results show that global warming as simulated by GCMs can't explain the recent widening.

19. Observed widening not explained by stratospheric ozone depletion
Strengthening of polar stratospheric jet → wider Hadley cell (Polvani and Kushner )
Strengthening of S. Hem jet in recent decades largely attributable to ozone depletion (Thompson and Solomon , Science; Gillett and Thompson , Science)
No consensus among model simulations of 20th century
13. Strengthening of the stratospheric polar vortex in simple models results in a poleward displacement of the jet. We might expect that the recent strengthening of the southern hemisphere polar vortex attributed largely to ozone depletion to have an effect on the Hadley cell width. However, there isn't a consensus among the models on the importance of ozone depletion and widening. This figure shows the widening from the models that include stratospheric ozone depletion and from those that do not include it. (A little more than half of the models include it). Using stream function or OLR as a measure of Hadley cell width, the models that include ozone depletion show about twice as much widening as those without. But when we use P-E as a measure of Hadley cell width, the models without ozone depletion show more widening.

20. GCM results with interactive chemistry and good representation of stratospheric dynamics
Lamarque & Solomon (2008)

21. Observed widening of about 3º since 1979
Conclusions
Observed widening of about 3º since 1979
Not due to natural variability
Cannot be explained by historical changes in forcings as simulated by GCMs
Cannot be explained by pattern of SST warming
Occurs at a faster rate than in projections of future widening
Further investigation required to resolve discrepancy between observations and GCM simulations