How does the “box modifying” code work?

Within a 3-dimensional box, the temperature tendencies due to convection, cloud and radiation processes can be modified using a scaling factor. The box can be either static or move along a linear path. The duration of the scaling in time can be chosen arbitrarily and within this period the scaling factor is applied in every integration time step.

Consider the followings:

  1. When you set the box, you also have to set a zone around it where smoothing is applied between the inner part of the box and its "environment".
  2. When you increase (reduce) the original tendency values, i.e. the scaling factor is greater (lower) than 1, it does not mean necessarily heating (cooling). If the original temperature tendency was negative, it leads to increased (reduced) cooling, e.g., in case of temperature tendency due to radiation at night.

  3. The model time step was 900 s in the experiments. The solution of the radiative transfer equation to obtain the fluxes is very expensive, so computer time spent in the radiation calculations are saved by using a reduced radiation grid and calling the full radiation with a reduced time frequency (3 hours). Between the full radiation time steps, a reduced radiation configuration is applied to update the radiative fluxes at every grid point and every time step for the relevant instantaneous temperature profile and solar zenith angle.

  4. For visualization, temperature tendency is accumulated during the days, its unit is K/day. The humidity tendencies are also accumulated and scaled to K/day.

  5. The magnitude of temperature tendency due to radiation is much lower (a few K/day) than the one due to convection (reaching even 20-30 K/day in some regions).

OpenIFS users can request a copy of the "box modifying" code for the user group experiments. Please contact openifs-support@ecmwf.int.

On this page

User group experiments

Table: User group experiments.

Experiment

Scientific contentGroup
controlOpenIFS 43r3 experiment at TCo399L137 resolution with 900 s time step initialized from ERA5 at 00 UTC on 25 September 2019 (no scaling or scaling=1)All groups
rad_0_NA_topScaling to 0 for radiative tendency between 400 and 50 hPa over the North AtlanticJake, Leo
rad_2_NA_topDoubling the radiative tendency between 400 and 50 hPa over the North AtlanticJake, Leo
rad_0_KarlScaling to 0 for radiative tendency between 900 and 100 hPa over KarlArnaud, Lorenzo, Miriam, Mokhliss

all_0_Karl

Scaling to 0 for tendencies due to radiation, convection and cloud processes between 900 and 100 hPa over Karl

Arnaud, Lorenzo, Miriam, Mokhliss
con_0_Karl_48h

Scaling to 0 for convective tendency in a box moving with Karl in the first 48 hours

Irina, Lauri, Mika, Terhi + Joakim
con_2_Karl_48hDoubling the convective tendency in a box moving with Karl in the first 48 hoursIrina, Lauri, Mika, Terhi + Joakim
con_0.5_def_24h

Scaling to 0.5 for convective tendency in the “default” box between 900 and 200 hPa in the first 24 hours

Guokun, Jian-Feng, Jun, Ying + Victoria
con_1.5_def_24hScaling to 1.5 for convective tendency in the “default” box between 900 and 200 hPa in the first 24 hoursGuokun, Jian-Feng, Jun, Ying + Victoria
con_0.5_Karl_48h

Scaling to 0.5 for convective tendency between 900 and 300 hPa in a box moving with Karl in the first 48 hours

Bethan, Federico, Marvin
con_1.5_Karl_48hScaling to 1.5 for convective tendency between 900 and 300 hPa in a box moving with Karl in the first 48 hoursBethan, Federico, Marvin
cloud_0.5_Karl_48h

 Scaling to 0.5 for tendency due to cloud microphysics between 900 and 300 hPa in a box moving with Karl in the first 48 hours

Bethan, Federico, Marvin
cloud_0.5_Karl_48hScaling to 1.5 for tendency due to cloud microphysics between 900 and 300 hPa in a box moving with Karl in the first 48 hoursBethan, Federico, Marvin

Modifying the temperature tendency due to radiation

Main conclusions:

  1. Increasing tendency from radiation (heating at daytime, cooling at night) in the upper levels over the North Atlantic (experiment rad_2_NA_top in Table) leads to higher mean sea level pressure values over the North Atlantic and it is compensated with lower MSLP almost everywhere else (Fig. 1).
  2. Nullified tendency from radiation over the North Atlantic from 900 to 100 hPa (experiment rad_0_Karl) slightly deepens the cyclone (or shifts it eastwards). The output tendency from radiation remains below 1K/day after 48 hours (i.e. at 00 UTC; Fig. 2).

Explanation:

  • The magnitude of temperature tendency due to radiation is small, a few K/day. Its modification has significant impact only if it occurs over a large area and during a longer period.

Figure 1: Difference in mean sea level pressure (in hPa) between the experiment with no temperature tendency from radiation (rad_0_NA_top) and the experiment with doubled temperature tendency from radiation (rad_2_NA_top).Figure 2: Cross section for 48-hour forecast of temperature tendency from radiation (in K/day, with shading) and potential temperature (in K, with isolines) along the line highlighted in red in the top map. The graph shows the result of the experiment with zero temperature tendency from radiation (rad_0_Karl). The dotted rectangle represents the box where the tendency modification was applied.

Modifying the temperature tendency due to convection

Main conclusions:

  1. Nullifying and doubling the temperature tendency due to convection in a box covering (and following) Karl in the first 48 hours (experiments con_0_Karl_48h and con_2_Karl_48h) leads to increased and decreased mean sea level pressure on the first day, respectively. The direction of the change is clear, but its magnitude is not proportional with the size of the scaling factor (confirmed also by the experiment pairs of con_0.5_def_24h and con_1.5_def_24h, con_0.5_Karl_48h and con_1.5_Karl_48h applying different scaling factors). The structure of the potential vorticity also coincides with that.
  2. The temperature tendency due to convection can reach 3 K/day between 900 and 400 hPa in the 24- and 48-hour forecasts (Fig. 3), even if we remove (con_0_Karl_48h) or halve (con_0.5_Karl_48h) the tendency in a few time steps. It results in a small extra drying in the humidity tendency in the same layer (not shown).
  3. Increasing the convective temperature tendency by 50 % in the first 24 hours in a small area close to the mean sea level pressure minimum of Karl (con_1.5_def_24h) leads to slight changes in the daily accumulated convective tendency: small increase in the lower levels and drop above 900 hPa (Fig. 3). The scaling has straightforward effect on the precipitation especially if it is applied during a 48-hour period (con_1.5_Karl_48h; Fig. 4).

 

Figure 3: Vertical profile for 48-hour forecasts of temperature tendency from different sources (in K/day) over the area highlighted in red in the top right map. The top graph shows the result of the control experiment, the middle and the lower graphs show the results of the experiments where scaling factor of 0.5 and 1.5 was applied on the convective temperature tendency (con_0.5_Karl_48h and con_1.5_Karl_48h), respectively. Note the different scale of the x-axis.

Explanation:

  • If the box moves along the trajectory of Karl (as in experiments con_{0|0.5|1.5|2}_Karl_48h), the tendency modification is applied over a given grid point only for few time steps.
  • The convection scheme continuously acts in the model, so the temperature (humidity) tendency at step +24h is the sum of the tendencies resulted by the convection scheme during every 900 s (i.e. 15 minutes) between 00 UTC on 25 September and 00 UTC on 26 September 2016. Therefore, even if we remove the temperature tendency due to convection over a given area during a part of the day, the deep convection mechanisms considerably develop during the rest of the day.
  • More intense convection processes deepen the cyclone and the opposite happens with a reduced convection influencing also the precipitation amount.

 

Figure 4: 66-hour precipitation forecasts resulted by the experiments where scaling factor of 0.5 (left; con_0.5_Karl_48h) and 1.5 (right; con_1.5_Karl_48h) was applied on the convective temperature tendency.

Modifying the temperature tendency due to cloud processes

Main conclusions:

  1. Halving the temperature tendency due to cloud processes in a box following Karl in the first 48 hours (experiment cloud_0.5_Karl_48h) increases the accumulated near-surface heating from the cloud microphysics and it is compensated by enhanced low-level cooling from the dynamics. Shallow convection seems to be less active resulting in a reduced drying near the surface. Between 800 and 400 hPa the deep convection gets more intense accompanied by more excessive heating and drying in this layer (Fig. 5). At the same time, the dynamical processes add moisture here.
  2. Changing the temperature tendency due to cloud processes in a box covering Karl in the first 48 hours has significant impact on the precipitation over Scandinavia later (Fig. 6). Applying scaling factor of 1.5 (experiment cloud_1.5_Karl_48h) leads to an earlier precipitation maximum before 18 UTC 27 September 2016, while scaling factor of 0.5 (experiment cloud_0.5_Karl_48h) results in a delay in precipitation maximum.

Explanation:

  • The stability changes due to the condensation heating in upper levels and changes in the evaporation (cooling) in the lower levels. Convection largely compensates the changes in the cloud tendencies.
  • Dynamics also reacts to the changes in the heating, mainly balances the drying due to convection and cloud physics above the boundary layer. The heating generates lifting of air parcels and the lifting generates moistening.

  

Figure 5: Vertical profile for 48-hour forecasts of humidity tendency from different sources (in K/day) over the area highlighted in red in the top right map. The top graph shows the result of the control experiment, the lower graph shows the result of the experiment where scaling factor of 0.5 was applied on the temperature tendency from cloud processes (cloud_0.5_Karl_48h). Note the different scale of the x-axis.


 

Figure 6: 66-hour (top) and 72-hour (bottom) precipitation forecasts resulted by the experiments in which scaling factor of 0.5 (left; cloud_0.5_Karl_48h) and 1.5 (right; cloud_1.5_Karl_48h) was applied on the temperature tendency from cloud processes.

Acknowledgement

Grateful thanks to Peter Bechtold who helped in understanding the results and phrasing the explanations. We are also grateful to Richard Forbes who provided the box modifying code to the OpenIFS team.