Kengo Matsubayashi Japan Meteorological Agency The Met Office Outline Motivation Observation and evaluation method Evaluation result Summary In many convection schemes water substances are empirically modelled ID: 1026050
Download Presentation The PPT/PDF document "Evaluation of convective precipitation u..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
1. Evaluation of convective precipitation using space-borne radar observationKengo MatsubayashiJapan Meteorological AgencyThe Met Office
2. OutlineMotivationObservation and evaluation methodEvaluation resultSummary
3. In many convection schemes, water substances are empirically modelled.Evaluation of water substances in convection schemes is important because of its latent heat release, evaporation and water loading directly influence the global circulation. In this study, as a proxy of direct evaluation of vertical profile of convective precipitation in convection schemes, radar reflectivity is evaluated using satellite-based radar observation.Motivation
4. Observation and evaluation method
5. Two space-borne radarsIn this study, 2 kinds of space-born radar observation, CloudSat CPR and TRMM PR, are used.These 2 radars have different characteristics, so using two radars helps understanding what causes discrepancies between models and observation.CloudSat CPRTRMM PRResolution, Range1.4km * 1.7km1.4km swath widthVertical 500m5km*5km(4.3km*4.3km until 2001 Aug)245km swath widthVertical 250mFrequency94.0GHz13.8GHzMinimum detectable echoes-27.5dBZ18dBZ2B-GEOPROF P1_R052APR Version 06A
6. CloudSat CPRTRMM PRResolution, Range1.4km * 1.7km1.4km swath widthVertical 500m5km*5km(4.3km*4.3km until 2001 Aug)245km swath widthVertical 250mFrequency94.0GHz13.8GHzMinimum detectable echoes-27.5dBZ18dBZTwo space-borne radarsWhile CloudSat CPR has higher horizontal resolution, TRMM PR has wide swath and can observe the horizontal distribution of convection.
7. CloudSat CPRTRMM PRResolution, Range1.4km * 1.7km1.4km swath widthVertical 500m5km*5km(4.3km*4.3km until 2001 Aug)245km swath widthVertical 250mFrequency94.0GHz13.8GHzMinimum detectable echoes-27.5dBZ18dBZTwo space-borne radarsCPR strongly attenuates by large particles (>1mm) in heavy precipitation, whereas PR can observe heavy precipitation.TRMM PRCloudSat CPR
8. CloudSat CPRTRMM PRResolution, Range1.4km * 1.7km1.4km swath widthVertical 500m5km*5km(4.3km*4.3km until 2001 Aug)245km swath widthVertical 250mFrequency94.0GHz13.8GHzMinimum detectable echoes-27.5dBZ18dBZCPR has high sensitivity and can detect very weak echoes by cloud with small particles (e.g. cirrus), on the other hand PR cannot detect weak echoes.Two space-borne radars
9. Unified Model AMIP run~208km*139km@EquatorCOSP(Bodas-Salcedo et al.2011)Split the model grids into sub-columnsAssign large-scale condensation considering cloud overlapping in a consistent manner to radiation scheme.Estimate convective water substances from precipitation flux using particle size distribution(PSD) and terminal velocity, and assign to subcolumns.Calculate vertical profile of radar reflectivity based on PSDs in each sub-columns.How do we assign convective precipitation to subcolumns?We have to diagnose convective precipitation fraction(CPF)which is the area fraction convective rain precipitate in a model grid.Simulation of radar reflectivity
10. To develop a method to estimate convective precipitation fraction(CPF), we made use of TRMM PR as it can observe horizontal distribution of convection.TRMM PR observation is assigned to AMIP run model grids(208km*139km).Grids, where observation covers more than 95% of its area, are used in this study. We classified TRMM PR observation to convective column or not, based on its cloud base height estimated from radar reflectivity profile.Thus, we calculated convective precipitation fraction at the model resolution.Convective precipitation fraction(CPF) estimation
11. Using TRMM PR observation, the relation between CPF and grid-box mean convective precipitation is investigated.If CPF has correlation to grid-box mean convective precipitation, CPF can be estimated using grid-box mean convective precipitation which convective scheme predicts.The relation derived from TRMM PR looks to have correlation, but has large spread. So, we handled CPF stochastically depending on convective precipitation amount.We produced PDF of CPF using TRMM PR observation in advance, and simulated the probability of radar reflectivity occurrence using the PDF.Convective precipitation fraction(CPF) estimation
12. This method is based on the assumption that the convective precipitation amount predicted in the model is consistent to observation. Actually, due to the intermittency of the current convection scheme, the model predicts heavy precipitation too frequent, and weak precipitation too low frequently.Model1 time-step precipitationObservation(TRMM PR)Precipitation intensity histogram
13. This intermittency problem is going to be resolved in the new convection scheme (Michael Whitall’s talk on Monday). So putting that aside for now, 1-hour averaged precipitation, which well follows observation frequency, is used in this study.Model1 hour mean precipitationObservation(TRMM PR)Precipitation intensity histogram
14. Evaluation result
15. TRMM PR(>18dBZ)CloudSat CPR(>-27.5dBZ)ObsMdlSimulated radar detection well corresponds to observation, but there are several differences.In the tropics(N20-S20), simulated CPR and simulated PR over 5km has too infrequent detection compared to observation.Radar detection frequency
16. TRMM PR(>18dBZ)CloudSat CPR(>-27.5dBZ)ObsMdlSimulated radar detection well corresponds to observation, but there are several differences.In the tropics(N20-S20), simulated CPR and simulated PR over 5km has too infrequent detection compared to observation.Radar detection frequency (N40-S40)
17. Reflectivity-Altitude joint histogram(N20-S20)TRMM PRCloudSat CPRObsMdlThere are two large differences between the observations and simulated fields.Radar reflectivity over 5km, which is expected to be ice, is too weak in simulated.Radar reflectivity in lower layer is too infrequent compared to observations.
18. Reflectivity-Altitude joint histogram(N20-S20)TRMM PRCloudSat CPRObsMdlRadar reflectivity over 5km, which is expected to be ice, is too weak in simulated.Wrong ice category classification?In the radar reflectivity simulation, convective ice are assumed to be snow aggregates. If this discrepancy is caused by graupel/hail, which have large diameter, PR radar reflectivity should be stronger and CPR reflectivity should be strongly attenuated.
19. Reflectivity-Altitude joint histogram(N20-S20)TRMM PRCloudSat CPRObsMdlRadar reflectivity over 5km, which is expected to be ice, is too weak in simulated.Wrong particle size distribution?We tried other ice PSDs (e.g. Gunn and Marshall, Sekhon and Srivastava), there was not much difference.
20. Reflectivity-Altitude joint histogram(N20-S20)TRMM PRCloudSat CPRObsMdlRadar reflectivity over 5km, which is expected to be ice, is too weak in simulated.Too few ice?The current convection scheme in UM employs a constant threshold to determine the amount of water substance to fall out from updraft. However, actually it should depend on vertical velocity. Much more ice with large diameter may be carried to upper layer.
21. Reflectivity-Altitude joint histogram(N20-S20)TRMM PRCloudSat CPRObsMdlRadar reflectivity in lower layer is too infrequent compared to observations.Weak (and shallow) convection could be too infrequent in the model. The frequency of the convection scheme triggering is too low compared to observations.
22. Although the frequency of convection occurrence depending on resolution is expected to follow the observed distribution(red dots), the frequency of convection scheme triggering in the model at ~208km*139km resolution is too low (blue dot).Current convection scheme predicts weak precipitation which is less frequent compared to observations. It implies the number of weak convective events predicted in the model is insufficient.Frequency of convection triggering
23. ObservationTo see the convective precipitation in detail, extract cloud with cloud base height lower than 3km.Congestus can be seen in both observations and simulated reaching above ~5km.Although observed congestus reaches to ~7km, congestus predicted in the model is lower than observations and its height is too concentrated. This is due to the lack of overshooting in the convection scheme.CloudSat CPR CFADModel
24. ObservationTo see the convective precipitation in detail, extract cloud with cloud base height is lower than 3km.Congestus can be seen in both observations and simulated data above ~5km.Although observed congestus reaches ~7km, congestus predicted in the model is lower than observations and its height is too concentrated. This may be due to the lack of the representation of overshooting in the convection scheme.CloudSat CPR CFADModel
25. SummaryAs a proxy of direct evaluation of convective precipitation, radar reflectivity evaluation with stochastic convective precipitation fraction has been developed.Radar reflectivity evaluation using 2 radar observation revealed several discrepancies.Too weak reflectivity over 5kmLess frequent convective precipitation in low layerToo low congestusAs a next step, we are going to evaluate the new convection scheme called CoMorph which is being developed at the Met Office.
26. Questions?kengo.matsubayashi@metoffice.gov.uk