TropicalRCE

This repository contains the analysis scripts and output for the TropicalRCE project, written in Julia. We aim to investigate how RCE simulations differ from the tropical atmosphere and why.

Created/Mantained By: Nathanael Wong ([email protected])
Other Collaborators: Zhiming Kuang ([email protected])

Introductory Text Here.

Progress

0. Motivation

RCE simulations are often taken as an approximation to the tropical atmosphere. However, from initial runs iof SAM in Radiative-Convective Equilibrium (RCE) Cloud-Resolving Mode (CRM), we find that cloud-resolving RCE runs in the System of Atmospheric Modelling (SAM) v6.10.6 have a net energy balance at the surface is much greater than the tropical average O(40) W/m2. The net energy balance in our runs at 302 K and equatorial insolation hovers at O(120) W/m2. These results are not unique. A recent study done by Wing et al. (2020) as part of RCEMIP shows that for small domains (O(100)km in both horizontal directions), the mean surface energy balance is O(100) W/m2.

In a way, these results are somewhat expected, given that atmospheric transport out of the tropics is not accounted for in these small-domain CRMs. However, even when accounting for large-scale vertical motion as a proxy for large-scale atmospheric motion and ascent in the tropics, the surface energy balance still hovers at O(120) W/m2. Therefore, in this project, we aim to explore and understand the differences between our results in small-domain RCE simulations, and reanalysis data (ERA5) that acts as a proxy to observations.

1. Datasets Used

A. Reanalysis Data

We used the following ERA5 Reanalysis data from the Climate Data Store:

Surface Fluxes: Net Solar, Net Longwave, Sensible and Latent
TOA Fluxes: Net Solar, Net Longwave
Cloud Cover: Total, High (0-450 hPa), Medium (450-800 hPa), High (800-1000 hPa)
Temperature: 2m Air Temperature, Sea Surface, Pressure Lvl
Total Column Water, Column Water Vapour
Land-Sea Mask
Vertical winds (by pressure level)

B. Observational Data

We used GPM IMERG precipitation data from the PMM website, and ETOPO1 grid data.

C. Model Data

I. System of Atmospheric Modelling (SAM)

In our project, we ran SAM v6.10.6 (w/ modifications by Dr. Peter Blossey) as a limited area CRM (64 x 64 x 64) with 2 km horizontal resolution in order to get the baseline RCE state for a variety of different combinations of:

Fixed sea-surface temperature
Insolation (averaged over domain latitudes)

Other notable configurations:

1/2-moment Microphysics (SAM1MOM and M2005 options in SAM)
Diurnal cycle on, perpetual spring equinox
RRTM Radiation Scheme

II. Super-Parameterized Community Atmosphere Model (SP-CAM)

Our group maintains a version of the Community Atmosphere Model coupled to a 2D-domain of SAM run in CRM mode for the super-parameterization of convection in CAM, known as SP-CAM. In order to see if the discrepancy in the our surface energy balance are due to the setup of our experiments, or artifact of SAM, we compare our results against runs in SP-CAM:

If the surface energy-balance in SP-CAM is similar to our results in pure RCE mode, then our results are likely an artifact of an error in SAM
If the surface energy balance is similar to that in reanalysis, then there is some aspect that is missing in our model setup that results in this energy imbalance.

Notable configurations:

Perpetual February (spunup from average February climatology, Feb 15 insolation)

2. Domains

We considered the following domains in our analysis:

DTP (Deep Tropics): 0-360ºE, 15ºS-15ºN
IPW (Indo-Pacific Warmpool): 90-180ºE, 15ºS-15ºN
WPW (West Pacific Warmpool): 135-180ºE, 10ºS-5ºN
DRY (Dry Pacific): 180-275ºE, 5ºS-5ºN

These domains were chosen based on a combination of sea-surface temperature and precipitation characteristics.

IPW and WPW domains have relatively high SST compared to the rest of the tropics (WPW more so and more localized than IPW)
DRY domain has a relatively low SST despite equatorial insolation and very low precipitation

3. Model Runs Table

We ran SAM with the following SST and Insolation Configurations:

Domain	Insol / W m**-2	SST Range / K	Avg SST / K	Microphysics
DTP	1345.6	299-303, step 0.5	300.7	SAM1MOM (1M)
IPW	1345.6	300.5-303, step 0.5	301.9	SAM1MOM (1M)
WPW	1355.8	301-303, step 0.5	302.4	SAM1MOM (1M)
DRY	1359.3	297-302, step 0.5	299.7	SAM1MOM (1M)
DTP	1345.6	299-303, step 0.5	300.7	M2005 (2M)
IPW	1345.6	300.5-303, step 0.5	301.9	M2005 (2M)
WPW	1355.8	301-303, step 0.5	302.4	M2005 (2M)
DRY	1359.3	297-302, step 0.5	299.7	M2005 (2M)

We conducted three batches of the above experiments:

RCE in 3D, 64x64 horizontal grid at 2 km resolution, 64 vertical levels (3D)
RCE in 2D, 64 horizontal points at 2 km resolution, 64 vertical levels (2D)
RCE in 2D, 512 horizontal points at 0.5 km resolution, 64 vertical levels (2DH)

4. Comparison of SAM Runs against Reanalysis and Observation Data

A. Surface Variables

Precipitation / mm/day

Domain	Insol	SST / K	GPM	ERA5	3D-1M	3D-2M	2D-1M	2D-2M	2DH-1M	2DH-2M
DTP	1345.6	300.7	4.70	4.70	3.05	2.64	3.17	2.67	2.97	2.60
IPW	1345.6	301.9	7.05	6.53	3.22	2.80	3.33	2.82	3.01	2.71
WPW	1355.8	302.4	8.90	7.98	3.26	2.86	3.34	2.83	3.11	2.81
DRY	1359.3	299.7	1.86	2.61	2.91	2.49	3.05	2.50	2.73	2.45

(Note: We plot the ERA5 precipitation below:)

It is interesting to see that the ERA5 domain mean precipitation over the deep tropics as a whole is almost exactly the same as that given by the GPM IMERG dataset. However, on a smaller domain scale, such as over the WPW or DRY domains, differences remain. We do note that the ERA5 precipitation in the DRY region is very similar to that from our RCE runs, though the significance of this is as of yet unclear (caa 28 Oct 2020).

Total Column Water / mm

Domain	Insol	SST / K	ERA5	3D-1M	3D-2M	2D-1M	2D-2M	2DH-1M	2DH-2M
DTP	1345.6	300.7	42.78	40.66	43.20	39.73	43.47	42.78	44.13
IPW	1345.6	301.9	49.05	45.30	47.80	44.50	47.87	49.14	49.52
WPW	1355.8	302.4	52.15	47.57	50.01	47.04	50.65	51.37	51.51
DRY	1359.3	299.7	40.69	37.15	39.95	35.91	39.87	40.64	40.40

Right off the bat, we see that 2-moment microphysics simulations have a higher overall precipitable water in the atmosphere compared to 1-moment microphysics. However, this does not correspond to higher rainfall. Indeed, precipitation actually falls in simulations with 2-moment microphysics despite there being higher overall column water. Furthermore, we see that at the same horizontal resolution, 2D simulations have higher rainfall than 3D simulations, but as the resolution of the simulations increases, rainfall rate decreases.

B. Atmospheric Temperature (Mean-State)

We compare the profiles for atmospheric temperature (see figures below, top for observations and comparison between domains, bottom for comparison between SAM and reanalysis.)

We see that the atmospheric profile for the tropical domain as a whole is generally cooler than the specific-domain profiles. This is likely for different reasons for different domains (note, I am only analyzing profiles over the ocean):

For the IPW profile, we see that temperature is greater than DTP below 800 hPa and above 600 hPa all the way to around the tropopause. This is likely because of the fact that at these levels, convective activity is higher on average at these pressure levels than in DTP.
For the WPW profile, we see a similar curve as with IPW, which indicates that convective activity is indeed partially responsible. However, there is an overall mean increase in temperature compared to the IPW profile, which means that the temperature increase could be due to overall higher insolation
For the DRY profile, we see that the mean profile average temperature anomaly is around the same as WPW, which indicates that higher insolation plays a role in the temperature difference compared to DTP. However, with that taken into account, we see that the profile seems to be reversed compared to WPW, with higher temperatures in 400-800 hPa compared to WPW. This indicates that convective inhibition and downwelling plays a role in increasing the temperature profiles here.

Now on to a comparison with our model results ...

(Note: Black line is reanalysis for domain, blue/red lines are for 1/2-moment microphysics respectively, solid/dashed/dotted lines are for 3D/2D/2DH spatial configurations respectively.)

Here, we see that while lower-tropospheric temperatures are higher than in observations, upper-tropospheric temperatures near the tropopause are lower. However, temperatures in the stratosphere are higher. We see that for 1-moment microphysics, high resolution simulations are needed in 2D in order to prevent high temperature biases in the stratosphere above the tropopause. However, there is otherwise not significant difference in the temperature profiles across all our simulations. We do note that 1-moment microphysics simulations have a colder upper-troposphere in general than 2-moment microphysics simulations.

C. Relative Humidity Profile

We compare the profiles for relative humidity between SAM and reanalysis. Since ERA5 calculates relative humidity profiles using both liquid and ice saturation vapour pressures, we calculate relative humidity profiles from SAM via the following

When the atmospheric temperature T < 0 (Note: T is taken instantaneously), then we calculate using ice saturation vapour pressure
When the atmospheric temperature T > 0 (Note: T is taken instantaneously), then we calculate using liquid saturation vapour pressure

We note that this is not directly equivalent to the method that ERA5 uses to calculate relative humidity, which can be found here

D. Cloud Fraction Profile

We compare the profiles for cloud cover between SAM and reanalysis.

We see that all SAM simulations are able to capture the presence of high clouds in the upper troposphere. However, 3D SAM conversely is unable to capture the presence of medium clouds, and significantly underestimates the presence of low clouds. However, in all our 2D simulations, we see that there is an improvement in the fraction of cloud cover in the mid-troposphere.

The 2-moment M2005 microphysics overestimates the presence of high clouds. We see that the high cloud cover in the M2005 microphysics simulation is closer to observations in regions that by default have significant large-scale vertical ascent, which is not representative of a stable small-domain RCE state. Indeed, it seems that for statistical purposes, 1-moment microphysics would be a better fit to observational data, especially since overall in the TRP region there is overall a small amount of large-scale ascent.

5. Surface Energy Balance

A. Radiative-Convective Equilibrium in SAM

Here, I display a summary of the surface energy balance for the model runs in SAM with averaged SST for each ERA5 domain. A full table containing all the experiments will be provided elsewhere.

Domain	Insol	SST / K	Config	Net SW	Net LW	Sensible	Latent	SFC Bal
DTP	1345.6	300.7	3D-1M	+280.89	-68.66	-10.66	-88.59	+112.98
IPW	1345.6	301.9	3D-1M	+279.96	-64.29	-10.41	-93.23	+112.03
WPW	1355.8	302.4	3D-1M	+281.43	-62.22	-10.20	-94.83	+114.18
DRY	1359.3	299.7	3D-1M	+285.66	-72.13	-10.75	-84.62	+118.17
DTP	1345.6	300.7	3D-2M	+271.18	-64.21	-6.67	-76.66	+123.65
IPW	1345.6	301.9	3D-2M	+270.28	-60.38	-6.57	-81.30	+122.03
WPW	1355.8	302.4	3D-2M	+272.64	-58.61	-6.47	-83.23	+124.33
DRY	1359.3	299.7	3D-2M	+274.36	-66.87	-6.62	-72.27	+128.60

We see overall that the energy balance of small-domain RCE simulations in SAM have an overall surface energy balance of ~O(120) W/m2, compared to typical values of about 0-40 W/m2 (see comparison with reanalysis data below), and this is largely due to the very high net shortwave radiation into the ocean, which is only partially compensated by slightly increased net longwave radiation upward.

Domain	Insol	SST / K	Config	Net SW	Net LW	Sensible	Latent	SFC Bal
DTP	1345.6	300.7	2D-1M	+274.42	-67.25	-10.33	-91.04	+105.81
IPW	1345.6	301.9	2D-1M	+274.81	-62.49	-10.17	-96.58	+105.57
WPW	1355.8	302.4	2D-1M	+278.70	-60.24	-9.89	-97.53	+111.04
DRY	1359.3	299.7	2D-1M	+279.28	-71.22	-10.72	-88.70	+108.64
DTP	1345.6	300.7	2D-2M	+257.07	-60.02	-6.04	-76.93	+114.09
IPW	1345.6	301.9	2D-2M	+258.70	-56.98	-5.98	-82.24	+113.50
WPW	1355.8	302.4	2D-2M	+257.60	-53.90	-5.72	-82.50	+115.48
DRY	1359.3	299.7	2D-2M	+260.03	-63.37	-6.06	-72.60	+118.00

Domain	Insol	SST / K	Config	Net SW	Net LW	Sensible	Latent	SFC Bal
DTP	1345.6	300.7	2DH-1M	+264.99	-63.64	-10.45	-87.70	+103.20
IPW	1345.6	301.9	2DH-1M	+262.44	-57.07	-9.50	-88.37	+107.50
WPW	1355.8	302.4	2DH-1M	+262.14	-55.56	-9.46	-90.57	+106.55
DRY	1359.3	299.7	2DH-1M	+263.22	-63.68	-10.13	-80.20	+109.21
DTP	1345.6	300.7	2DH-2M	+258.88	-59.62	-6.53	-75.37	+117.35
IPW	1345.6	301.9	2DH-2M	+256.48	-54.87	-6.34	-78.75	+116.52
WPW	1355.8	302.4	2DH-2M	+258.27	-53.63	-6.30	-81.87	+116.47
DRY	1359.3	299.7	2DH-2M	+260.99	-62.90	-6.58	-71.42	+120.10

A comparison between 3D and 2D simulations shows us that for 2D simulations, the surface energy imbalance is slightly lower, though much higher than observations, at ~O(110) W/m2 on average. There appears to be no significant difference between simulations of the original size, and simulations that are larger in area and higher in resolution.

Something that is consistent among all simulations is that when 1-moment microphysics is switched to 2-moment microphysics, the Shortwave fluxes decrease in magnitude. However, at the same time the Sensible and Latent Heat fluxes decrease in magnitude. Therefore, even though less shortwave reaches the surface, the net change in sensible and latent heat fluxes are such that the surface imbalance in 2-moment microphysics actually increases by about 10% compared to 1-moment microphysics.

B. Comparison with ERA5 Reanalysis

We find that the surface energy balance in ERA5 reanalysis is much lower than that in our SAM model runs.

Domain	Insol	SST / K	Net SW	Net LW	Sensible	Latent	SFC Bal
DTP	1345.6	300.7	+224.71	-52.36	-10.39	-124.63	+37.33
IPW	1345.6	301.9	+217.57	-50.97	-11.83	-124.98	+29.79
WPW	1355.8	302.4	+215.72	-49.68	-12.83	-117.66	+35.56
DRY	1359.3	299.7	+239.39	-51.72	-7.32	-103.93	+76.42

C. Comparison with SP-CAM

As mentioned in Section 2, we also compare our results to that from SP-CAM in order to determine if the anomalously high surface imbalances we see in our SAM model runs are an artifact of our experiments being RCE runs (and therefore other aspects that must be included for the energy imbalance to be more realistic), or if this is a problem with SAM.

Domain	Insol	SST / K	Net SW	Net LW	Sensible	Latent	SFC Bal
DTP	1345.6	300.7	+202.43	-48.83	-10.93	-137.01	+5.65
IPW	1345.6	301.9	+183.81	-45.70	-12.77	-149.11	-23.77
WPW	1355.8	302.4	+180.08	-40.93	-11.29	-132.38	-4.52
DRY	1359.3	299.7	+224.75	-44.15	-5.03	-90.69	+84.88

We see from the SP-CAM runs that the surface energy balance within the tropics is definitely not as high as in our SP-CAM runs. The magnitude of the surface energy balance in the tropics and the net shortwave downwards into the ground at ~O(190) W/m2 which is close to that of reanalysis, indicates that the very high net shortwave ~O(280) W/m2 observed in small-domain RCE simulations is due to the RCE setup itself, rather that it being an artifact of SAM.

It is notable however, that in our SP-CAM runs the surface energy balance of the tropical regions is negative. However, this is likely due to the fact that the SP-CAM model was run in a perpetual mid-February insolation, which is not equinoctal, and where there is more insolation in the southern hemisphere that would account for an overall loss in surface energy balance in many tropical regions. In fact, we do see a gradient in the surface energy balance from north to south.

Domain	Insol	SST / K	Net SW	Net LW	Sensible	Latent	SFC Bal
DTP	1345.6	300.7	+236.93	-53.37	-9.79	-122.98	+50.79
IPW	1345.6	301.9	+224.71	-50.70	-11.68	-128.49	+33.85
WPW	1355.8	302.4	+220.04	-48.15	-13.02	-125.64	+33.24
DRY	1359.3	299.7	+246.68	-49.33	-6.28	-88.23	+102.84

6. Finding the Underlying Reason for Large Surface Imbalances in RCE

Our next step involves trying to figure out what exactly is the cause of the large surface energy imbalance that we observe in our RCE simulations as opposed to the reanalysis data. We hypothesize that the energy imbalance is largely due to differences in the cloud cover between our RCE model and reanalysis data. We investigate several different possible mechanisms:

The large-scale circulation (parameterized by vertical ascent/descent)
Vertical wind-shear (which could be responsible for distributing cloud cover more widely over the domain)
The absence of low cloud cover in RCE simulations compared to reanalysis
Differences in the diurnal cycle of cloud cover

A. Large-scale Circulation

In our limited-area RCE simulations, the large-scale circulation in the Tropics (especially the deep tropics) is parameterized in the form of the large-scale vertical motion. We therefore first obtain vertical profiles of the large-scale vertical motion using the ERA5 reanalysis data.

We see that the mean vertical velocity of the atmospheric column in the deep tropical region (blue) increases with height up until around 250-300 hPa, before decreasing and reaching zero at about the tropopause. This is consistent with our understanding that the tropics are regions of atmospheric ascent. This ascent is more pronounced in the Indo-Pacific and West Pacific warmpools, which is also consistent with our understanding that these are regions of strong convective activity due to a warmer ocean surface temperature. In contrast, over the Dry Pacific region there is atmospheric descent, as the ocean surface is of a lower mean temperature than the tropical mean.

In SAM, the take these mean profiles and use them as sounding inputs into their respective model runs. So, for example, the DTP vertical profiles are sounding inputs for our DTP runs, the IPW vertical profiles for IPW runs, and so forth. The impact of adding this vertical profile of vertical velocity as an analogue to the large-scale circulation on the surface energy imbalance is shown in the table below.

Domain	Insol	SST / K	Config	Net SW	Net LW	Sensible	Latent	SFC Bal
DTP	1345.6	300.7	2D-1M	+253.39	-55.28	-9.40	-78.89	+109.83
IPW	1345.6	301.9	2D-1M	+226.90	-45.04	-8.99	-72.47	+100.40
WPW	1355.8	302.4	2D-1M	+221.41	-41.63	-8.76	-69.83	+101.19
DRY	1359.3	299.7	2D-1M	+292.76	-68.31	-8.53	-76.67	+139.24
DTP	1345.6	300.7	2D-2M	+247.41	-53.74	-5.34	-66.37	+121.96
IPW	1345.6	301.9	2D-2M	+218.67	-43.95	-4.98	-59.38	+110.36
WPW	1355.8	302.4	2D-2M	+204.21	-38.48	-4.84	-54.93	+105.96
DRY	1359.3	299.7	2D-2M	+295.80	-70.45	-6.65	-76.85	+141.85

We see that there is a noticeable decrease in the net shortwave flux at the surface when there is strong vertical ascent in the domain (see the IPW and WPW analogues). There is not much change in the DTP-analogue simulation, but that is because the vertical ascent decreases. However, this is counterbalanced by a noticeable decrease in the magnitude of the other fluxes, most notably the longwave and latent fluxes, such that the overall surface balance is still roughly the same as before.

B. Wind Shear

We next attempted to see if adding vertical wind shear would help reduce the surface energy imbalance. In a manner similar to Blossey et al. (2010), we nudged the average zonal winds to a sheared-wind profile, with zonal wind increasing linearly with height from 0 m/s at the surface to 5 m/s at 14 km height. We aim to see if this would allow us to spread cloud cover with height. Thus, instead of near-vertical cloud cover, the clouds would be sheared and therefore would be spread out more over the domain, which would block more shortwave radiation.

The impact of adding vertical wind shear on the surface fluxes is shown in the table below.

Domain	Insol	SST / K	Config	Net SW	Net LW	Sensible	Latent	SFC Bal
DTP	1345.6	300.7	2D-1M	+274.38	-67.39	-10.32	-85.91	+110.77
IPW	1345.6	301.9	2D-1M	+259.0	-57.83	-10.40	-85.44	+105.33
WPW	1355.8	302.4	2D-1M	+249.74	-51.11	-10.01	-80.51	+108.11
DRY	1359.3	299.7	2D-1M	+298.69	-71.96	-7.99	-75.60	+143.14
DTP	1345.6	300.7	2D-2M	+261.91	-61.98	-6.64	-74.50	+118.79
IPW	1345.6	301.9	2D-2M	+237.94	-50.58	-6.15	-69.41	+111.80
WPW	1355.8	302.4	2D-2M	+224.07	-44.00	-5.83	-63.58	+110.67
DRY	1359.3	299.7	2D-2M	+298.55	-73.45	-6.54	-76.27	+142.29

We instead see that adding vertical wind shear instead serves to significantly increase the net shortwave radiation instead of reducing it, which is partially mitigated by small increases in the magnitude of the longwave and latent fluxes.

C. Low/Middle Cloud Cover in RCE

Text

D. The importance of the Diurnal Cycle

Text

X. Imposing a Weak Temperature Gradient

Domain	Insol	SST / K	Config	Net SW	Net LW	Sensible	Latent	SFC Bal
IPW	1345.6	301.9	2D-1M	+227.56	-66.97	-21.29	-138.79	+0.51
WPW	1355.8	302.4	2D-1M	+210.57	-67.27	-26.55	-156.19	-39.44
DRY	1359.3	299.7	2D-1M	+330.47	-130.13	-7.36	-160.71	+32.27
IPW	1345.6	301.9	2D-2M	+215.23	-61.26	-10.56	-110.10	+33.31
WPW	1355.8	302.4	2D-2M	+166.07	-57.10	-18.58	-145.46	-55.07
DRY	1359.3	299.7	2D-2M	+331.72	-132.13	-7.65	-158.78	+33.16

Installation

To (locally) reproduce this project, do the following:

Download this code base. Notice that raw data are typically not included in the git-history and may need to be downloaded independently.

Open a Julia console and do:

julia> ] activate .
 Activating environment at `~/Projects/TropicalRCE/Project.toml`

(TropicalRCE) pkg> instantiate
(TropicalRCE) pkg> add GeoRegions#master SAMTools#master

This will install all necessary packages for you to be able to run the scripts and everything should work out of the box.

(Note: You need to install the #master versions of GeoRegions.jl and SAMTools.jl as of now.)

Other Acknowledgements

Project Repository Template generated using DrWatson.jl created by George Datseris.

natgeo-wong / tropicalrce Goto Github PK

tropicalrce's Introduction