Scheduled Downtime
On Friday 21 April 2023 @ 5pm MT, this website will be down for maintenance and expected to return online the morning of 24 April 2023 at the latest

RE: Mismatch between Observed vs WRF Diurnal Cycle

This post was from a previous version of the WRF&MPAS-A Support Forum. New replies have been disabled and if you have follow up questions related to this post, then please start a new thread from the forum home page.

K

kunaldayal

Member
Hi,

I am using WRF to simulate the winds for a small island state in the tropics and I have compared simulated and observed results via diurnal cycle for a month and for some stations there is a huge mismatch, whereas for some locations WRF works pretty well.

For the mismatch locations, WRF is unable to capture the sea/land breezes.

What can be the possible causes for the huge mismatch?

What can I do to improve my results at the locations where there is a huge mismatch?

Appreciate your assistance and advice.

Regards
Kunal
 
Kunal,

Are you running WRF to integrate continuously for one month? Or you run multiple cases within one-month period and each run only covers a few days?
In the former case, you need to turn on sst_update so that the land-sea thermal contrast can be better represented by the model. In the latter case, you don't need to set sst_update and the mismatch could be due to other reasons. How complex the island terrain is in your case? The terrain effect might be interwovean with land-sea effect, leading to problems in the simulation. Have you tried the option topo_wind (&physics), which is expected to improve surface wind simulation. Note that this option only works with YSU PBL.
 
Hi,

Yes, I am running WRF to integrate continuously for one month with a spin-up time of 3 days.

SST_update is turned on for my simulations.

The island terrain is mixed from simple to complex and has a plateau of 1300 m in between the windward and leeward sides of the island (in the middle of the island).

The island has a diameter of approximately 130 km and the stations that are not able to capture the diurnal cycle are on the leeward side with elevations of 10 - 50 m on average over a radius of 3 km and are 1.3 - 10.5 km away from the shore.

I am using WRF-ARW v3.9.1.1 with 3 two-way nested domains of 20 km, 4 km, and 1.33 km.

I am using YSU PBL, so I can try using the option topo_wind = ? for all the domains? or just the two-smaller domains?

Appreciate your assistance and advice.

Regards
Kunal
 
Please turn on the topo_wind option for all domains.
Note that land-sea-breeze heavily relies on thermal contrast between the ocean and the land. You may also need to check the simulation of temperature in lower levels and examine the quality of your SST data.
 
Hi,

I have tried using topo_wind = 1/2 + ysu_topdown_pblmix = 1 options for my simulations and high-resolution sea surface temperature from RTG-SST_HR available at 0.083 deg x 0.083 deg resolution.

Some Observations:
- The topo_wind options does not improve the simulation results for the diurnal cycle on the leeward side.
- While for the windward side the amplitude of the simulation decreases in comparison with the measured data.
- The diurnal cycle improves for two near shore stations located on either ends of the island and on the margin of the leeward and windward side.
- There is still a huge overestimation on the leeward side of the island in terms of the wind speed and the diurnal cycle as well.

Is there anything else I can do or try to improve my simulations?

Appreciate your assistance and advice.

Regards
Kunal
 
I am suspicious that the leeside overestimation is probably due to the combined effects of land-sea breeze and downslope wind. Can you run a few sensitivity tests to further identify the possible reasons? For example, remove the topography, reduce the thermal-contract between land and ocean, etc.? WRF performance is highly case-dependent. One needs to explore the specific reasons for less realistic results.
 
Okay, I take note.

Well, I have run a number of sensitivity tests as follows, but the observations are almost similar as highlighted earlier:
- Topo_wind = 0, FNL-SST, diff_opt = 1 (Simple Diffusion), diff_6th_opt = 0 (6th Order Horizontal Diffusion), km_opt = 4 (2d Deformation - K Option).
- Topo_wind = 1, FNL-SST, diff_opt = 1 (Simple Diffusion), diff_6th_opt = 0 (6th Order Horizontal Diffusion), km_opt = 4 (2d Deformation - K Option).
- Topo_wind = 2, FNL-SST, diff_opt = 1 (Simple Diffusion), diff_6th_opt = 0 (6th Order Horizontal Diffusion), km_opt = 4 (2d Deformation - K Option).
- Topo_wind = 1, RTG-SST_HR, diff_opt = 1 (Simple Diffusion), diff_6th_opt = 0 (6th Order Horizontal Diffusion), km_opt = 4 (2d Deformation - K Option).
- Topo_wind = 2, RTG-SST_HR, diff_opt = 1 (Simple Diffusion), diff_6th_opt = 0 (6th Order Horizontal Diffusion), km_opt = 4 (2d Deformation - K Option).
- Topo_wind = 1, RTG-SST_HR, PBLMIX = 1, diff_opt = 1 (Simple Diffusion), diff_6th_opt = 0 (6th Order Horizontal Diffusion), km_opt = 4 (2d Deformation - K Option).
- Topo_wind = 2, RTG-SST_HR, PBLMIX = 1, diff_opt = 1 (Simple Diffusion), diff_6th_opt = 0 (6th Order Horizontal Diffusion), km_opt = 4 (2d Deformation - K Option).
- Topo_wind = 1, RTG-SST_HR, diff_opt = 2 (Full Diffusion), diff_6th_opt = 2 (6th Order Horizontal Diffusion = Positive Definite), km_opt = 4 (2d Deformation - K Option).
- Topo_wind = 2, RTG-SST_HR, diff_opt = 2 (Full Diffusion), diff_6th_opt = 2 (6th Order Horizontal Diffusion = Positive Definite), km_opt = 4 (2d Deformation - K Option).
- Topo_wind = 1, RTG-SST_HR, PBLMIX = 1, diff_opt = 2 (Full Diffusion), diff_6th_opt = 2 (6th Order Horizontal Diffusion = Positive Definite), km_opt = 4 (2d Deformation - K Option).
- Topo_wind = 2, RTG-SST_HR, PBLMIX = 1, diff_opt = 2 (Full Diffusion), diff_6th_opt = 2 (6th Order Horizontal Diffusion = Positive Definite), km_opt = 4 (2d Deformation - K Option).
- Topo_wind = 0, RTG-SST_HR, diff_opt = 1 (Simple Diffusion), diff_6th_opt = 0 (6th Order Horizontal Diffusion), km_opt = 4 (2d Deformation - K Option) + mp_physics = 3, cu_physics = 1, ra_lw_physics = 1, ra_sw_physics = 1, surface_input_source = 3, isfflx = 1, ifsnow = 1, h_mom_adv_order = 5, h_sca_adv_order = 5, v_mom_adv_order = 3, v_sca_adv_order = 3.
- Topo_wind = 2, RTG-SST_HR, PBLMIX =1, diff_opt = 1 (Simple Diffusion), diff_6th_opt = 0 (6th Order Horizontal Diffusion), km_opt = 4 (2d Deformation - K Option) + sf_urban_physics = 1, surface_input_source = 3, isfflx = 1, ifsnow = 1, h_mom_adv_order = 5, h_sca_adv_order = 5, v_mom_adv_order = 3, v_sca_adv_order = 3.

I have used the tropical physics suite for all options except for the ones in which I have specified the physics options.

Removing the topography is making topo_wind = 0.
How can I reduce the thermal-contrast between the land and the ocean? any specific settings in physics??

I take note and the surprising thing is that on the windward side WRF is performing well.

Appreciate your assistance and advice.

Regards
Kunal
 
Hi,

Despite running a few more sensitivity tests, the results on the leeward side did not improve.

I have a few questions:

- Since the input topography data is at a resolution of approximately 1 km x 1 km and my grid size for the smaller domain is 1.33 km x 1.33 km, is it possible that the model is missing important surrounding topography such as valley and hills on the leeward side which could slow down the wind?

- Is the model not picking up low speed Katabatic winds at night that may be generated by the mountain and valleys on the leeward side surrounding the stations?

- Would it be scientifically justified to denote that the overnight over-prediction on the leeward side is caused by systematic errors that are associated with the model due to the various errors while simplifying atmospheric processes via various physics and dynamics parameterizations schemes and inaccurate topographical representation?

Appreciate your assistance and advise.

Regards
Kunal
 
Please see my answers below to your questions:
(1) It is possible that the fine structure of topography is not resolved by the model, which may lead to biases in the results
(2) PBL scheme works better in the daytime than in the nighttime. This is a known issue and the less realistic performance of PBL scheme may affect surface variables, such as 10-m winds and 2-m temperature.
(3) I have no answer to this question, --- it needs a large number of case studies to prove that 'systematic error' exists in the model simulations. We don't have such information.
 
Hi,

I have plotted the 2 m time series and the diurnal cycle of the simulated temperature and to my understanding, on the leeward side stations, there is not much thermal contrast in the WRF simulated temperature during the day and night which may also be the reason for the poor simulation of the diurnal cycle of wind speed.

Is there any way in which the model temperature can be improved? Probably some physics settings?

Appreciate your assistance and advice.

Regards
Kunal
 
Hi,

I have found out that the soil moisture data input into WRF is at a coarse resolution of 1 deg x 1 deg and for a tropical small island country like Fiji, all the islands have the same volumetric soil moisture content and also there is not much difference in soil temperature at the 4 levels of 10 cm, 40 cm, 100 cm and 200 cm.

This maybe the reason for the poor simulation of temperature on the leeward side of the islands which affects the wind simulation as well.

Is there anything that I can do to fix this problem??

Is there any physics settings which can use to update probably the soil temperature and which may probably improve the simulation of 2 m temperature and the 10 m surface winds??

Appreciate your assistance and advice.

Regards
Kunal
 
You must log in or register to reply here.
Top