A note on horizontal resolution dependence for monsoon rainfall simulations

Transcription

1 Meteorol. Atmos. Phys. 74, 11±17 (2000) 1 Department of Meteorology, Florida State University, Tallahassee, Florida 2 T. J. Watson Laboratory IBM, Yorktown, New York A note on horizontal resolution dependence for monsoon rainfall simulations B. Jha 1, T. N. Krishnamurti 1, and Z. Christides 2 With 4 Figures Received August 16, 1999 Revised October 14, 1999 Summary The motivation for this study came from recent results of an Atmospheric Model Inter-comparison Project (AMIP) coordinated by the Lawrence Livermore Laboratory at Livermore, California. That project included a review of seasonal monsoon simulations from 13 different atmospheric models over the world. Most of the models used a horizontal resolution of roughly 300 km. The seasonal monsoon simulations from these models varied signi cantly. The poor performance by these models stems in part from the use of the coarse resolution. The purpose of this note is to show that by using the same model physics and lower boundary conditions, such as snow/ice cover and sea surface temperatures, the use of the higher horizontal resolution does have a stronger positive impact on the skill of monthly rainfall when compared to a lower horizontal resolution. In this note we present the results of such a comparison between the horizontal resolutions of T42 and T170. These studies are carried out for the prescribed lower boundary speci cation of sea surface temperatures and snow/ice cover with the help of an Atmospheric General Circulation Model. 1. Introduction In recent years precipitation climatology has been improved from the use of rainfall estimates from microwave radiometers and radar from satellites such as the Defense Meteorological Satellite Program (DMSP) and Tropical Rainfall Measurement Missions (TRMM). Outgoing long-wave radiation (OLR) has also been used to estimate rain rates. These types of data sets have been used to validate seasonal or monthly rainfall climatology of climate models. This short contribution is intended to describe a comparison of monthly mean monsoon rainfall simulation at two different horizontal resolutions from an Atmospheric General Circulation Model (AGCM). The results of seasonal monsoon simulations for the Atmospheric Model Intercomparison Project (AMIP) were reported by Sperber and Palmer (1996), and Gadgil and Sajani (1998). Most of the models used in this intercomparison employed horizontal resolutions of roughly 300 kilometers. Seasonal monsoon simulations have been a very dif cult issue for modelers. The slow varying forcing from such surface parameter as sea surface temperatures (SST) and snow/ice cover are some of the most important factors for seasonal monsoon simulations. We have developed a method called physical initialization for the incorporation of ``observed'' rain rates to improve the initial states of weather and seasonal simulation in global models (Krishnamurti et al., 1991, 1999). This physical initialization is invoked within the data assimilation phase of the model forecast, and by using reverse algorithms, it entails the restructuring of the vertical distribution of humidity for given estimates of the observed rainfall rates. A reverse similarity theory restructures the humidity in the surface layer (the constant ux layer)

2 12 B. Jha et al. such that the surface evaporation and the prescribed precipitation are in close balance within the assimilation. This component provides a balance among the vertically integrated evaporation, precipitation and the substantial derivative of the speci c humidity following Yanai et al. (1973) de nition of the apparent moisture sink. In addition to these, a restructuring of the speci c humidity of the upper troposphere provides a matching of the forecast model-based OLR with that obtained from satellite estimates. All of the aforementioned components of physical initialization are contained within the model's data assimilation. The purpose of this note is to illustrate that a higher horizontal resolution model provides a better de nition of the initial rainfall and that this information helps the one-month lead simulation of rainfall especially over the monsoon environment. The scope of this study is however fairly limited; it shows a comparison of monthly monsoon rainfall simulations for two horizontal resolutions at triangular truncations 42 and 170 waves for a global spectral model. These correspond to roughly 300 km and 70 km, respectively. integration for the period June 1 to June 30, 1993, at the respective horizontal resolutions. 3. Comparison of the T170 and T42 rainfall Here we shall show correlations between the observed and the simulated monthly rainfall 2. Design of model run and data assimilation The results shown here are based on a multi-level Florida State University (FSU) global spectral model, which is described in Krishnamurti et al. (1991). The FSU spectral model is outlined in the Appendix of this note. The start date of the experiment is June 1, 1993, for both the T42 and T170 experiments. The initial data for June 1, 1993, 00UTC are obtained from ECMWF analyses. The initial data sets also include a 10 day average SST for the period June 1, 1993 through June 30, The orographies for the respective horizontal resolutions are obtained from the U.S. Navy's global 10-minute tabulation. The snow and ice cover for the initial states is also obtained from ECMWF daily analyses. Furthermore, physical initialization following Krishnamurti et al. (1991, 1993) is invoked within a data assimilation phase of the model forecast. This provides a better coupling between the large-scale circulation and the meso-convective systems at the horizontal resolution T170 compared to that at T42. This is followed by Fig. 1. Monthly mean rainfall (mm/day) for June 1993, a Observed over monsoon domain; b Model T170 over monsoon domain; c Model T42 over monsoon domain

3 A note on horizontal resolution dependence for monsoon rainfall simulations 13 totals at the horizontal resolutions of T170 and T42. FSU has an ongoing program on precipitation analysis, following Gairola and Krishnamurti (1992). This detailed procedure makes use of OLR-based rain rates, following Arkin et al. (1994) and thus a rst guess eld is obtained. The rain gauge data sets (archived at National Center for Atmospheric Research) and the Special Sensor Microwave Imager (SSM/I) radiance for the U.S. Air Force DMSP Satellite F10 through F14 (as available) are used for an objective analysis of six-hourly rainfall totals. Here we translate the SSM/I radiance at 85GHz, 37GHz, 22GHz and 19GHz to rain rates using the Olson et al. (1996) algorithms. Six-hourly elds are composited to obtain the monthly rain rates used in this study. We have noted that the corrrelations between the ``observed'' monthly mean satellite/gauge based rainfall totals and the model forecast monthly mean rainfall at horizontal resolution T170 and T42 over the monsoon are the order of 0.45 and 0.26, respectively. Figure 1 (a,b,c) illustrates the results of the monthly mean observed rainfall for June 1993 and the monthly mean simulated rainfall for June 1993 at the horizontal resolutions T170 and T42, respectively from the global model. The maximum monthly rainfall amounts from the T42 simulations are roughly three times larger than the T170 simulations, which are closer to the observed monthly distribution of rainfall over the monsoon belt. The monsoon belt is de ned here between 30E to 150E and 30S to 40N. Our emphasis is on the impact of resolution and physical initialization on the skill of onemonth rainfall simulations. Stephenson et al. (1988) showed the phenomenological impact of resolution on monsoon circulation, and Krishnamurti (1990) also studied this in the same context. The streamlines at 850 hpa and the daily rainfall for the respective horizontal resolutions at T170 and T42 of the global model for selected dates 1, 15 and 30 June 1993 are illustrated in Fig. 2. The impact of horizontal resolution is clearly apparent from the details of the daily circulation features and the intensity of the rainfall. One of the differences in the daily rainfall distribution is the more meridionally oriented leg of convective elements at the horizontal resolution T170. At the lower horizontal resolution the precipitating elements exhibit a more zonal extension. This compares well with a stronger meridional wind component for the monsoon simulation at the higher horizontal resolution. We did note considerable differences in subseasonal variability at these two horizontal resolutions. In this note we have only addressed the mean state, i.e., the monthly precipitation totals which differ considerably over the tropics where the convective precipitation is dominant. It may be worth noting that Sperber et al. (1994) found that increasing horizontal resolution did not change the simulation of the mean state, but rather was bene cial for the simulation of subseasonal variability. The daily rainfall and the winds at 850 hpa for the observed and for the model simulations at T170 and T42 are averaged over 50E±70E and EQ±20N over monsoon zone. This daily time series of rainfall (mm/day) for the month of June 1993 is shown in Fig. 3. The T170 rainfall distribution is much closer to the observed as compared to the T42 model simulated rainfall distributions. A similar illustration is prepared for the magnitude of 850 hpa winds and is shown in Fig. 4. The magnitude of the 850 hpa winds is of the order of 2 to 8 m/s in the analyzed elds. The T170 simulation is able to replicate similar magnitudes while the T42 forecast in general underestimates that feature. 4. Concluding remarks Seasonal monsoon simulations are a very dif cult issue because of a number of modeling problems. The current thrust in the seasonal simulation problem is to move towards global coupled atmosphere-ocean modeling. Dif culties in monsoon simulations arise largely from the horizontal resolution issues. A coarse resolution such as a 2.5-degree latitude/longitude mesh is not able to resolve the monsoon convection adequately. Krishnamurti et al. (1998) have recently shown that the mature monsoon carries with it an organization of convection which is somewhat similar to that seen in hurricanes except for the scales and geometry of the organization. In that study, it was shown that for monsoon predictions on the order of three to four days, a low horizontal resolution model very quickly contaminates the organization of con-

4 14 B. Jha et al. Fig hour rainfall (mm/day) and streamlines at 850 hpa; a For June T170; b For June T170; c For June T170; d For June T42; e For June T42; f For June T42

5 A note on horizontal resolution dependence for monsoon rainfall simulations 15 Fig. 3. Daily rainfall (mm/day) averaged over 50E±70E and 0±20N for the month of June 1993; a Observed; b T170; c T42 vection along the low-level ow. It was also shown that a higher horizontal resolution model (resolution T255) subjected to rain rate initialization (called physical initialization) appears to preserve the organization of convection and the forecast skill over 3 to 4 days. The organization of convection furthermore appears to have some large implication for the overall monsoon energetics. Most previous studies on monsoon simulations summarized by Gadgil and Sajani (1998) were carried out at the horizontal resolution of T42. Although those AMIP studies did not address the horizontal resolution issue, there was at least one model that was run at two different horizontal resolutions: Rhomboidal 40 (R40) with a resolution of roughly 170 km at the equator and Triangular 80 (T80) with a resolution of roughly 140 km at the equator. The above were the versions of National Center for Environmental Prediction (NCEP) operational models. Unfortunately, in addition to the change of horizontal resolution, a large number of model changes were also incorporated by the time T80 was developed. Hence, a comparison of results from changes in the horizontal resolution alone could not be made for that model. However, it was noted that the seasonal monsoon forecasts from the T80 were vastly superior to those obtained from R40. Here, we have performed a number of AGCM experiments at two horizontal resolutios (T42 and T170), and we nd that if the initial data are subjected to physical initialization, then a measurable improvement of skill for a one

6 16 B. Jha et al. Fig. 4. Same as Fig. 3 but for wind (m/sec) at 850 hpa month rainfall simulation over the monsoon region is possible at T170 when compared to T42. A coupled ocean-atmosphere model at high horizontal resolution needs to be examined for the seasonal monsoon simulation issues. This example of the AGCM, presented in this paper suggests that a higher horizontal resolution climate model may be needed for such studies. Acknowledgement This research was supported by the following grants to the Florida State University: NOAA Grant No. NA77WA0571, NASA Grant No. NAGS-4729, NSF Grant No. ATM ± , and ONR Grant No. NO Appendix An outline of the FSU Global Spectral Model The global model used in this study is identical in all respects to that used in Krishnamurti et al. (1991). The following is an outline of the global model: (a) Independent variables: (x, y,, t). (b) Dependent variables: vorticity, divergence, surface pressure, vertical velocity, temperature, and humidity. (c) Horizontal resolution: Triangular 42 and 170 waves. (d) Vertical resolution: 15 layers between roughly 50 and 1000 mb. (e) Semi-implicit time-differencing scheme. (f) Envelope orography (Wallace et al., 1983). (g) Centered differences in the vertical for all variables except humidity, which is handled by an upstream differencing scheme.

7 A note on horizontal resolution dependence for monsoon rainfall simulations 17 (h) Fourth-order horizontal diffusion (Kanamitsu et al., 1983). (i) Kuo-type cumulus parameterization (Krishmanurti et al., 1983). (j) Shallow convection (Tiedke, 1984). (k) Dry convective adjustment. (l) Large scale condensation. (m) Surface uxes via similarity theory (Businger et al., 1971) (n) Vertical distributions of uxes utilizing diffusive formulation where the exchange coef cients are functions of the Richardon number (Louis, 1979). (o) Long- and shortwave radiative uxes based on band model (Harshvardan and Corsetti, 1984; Lacis and Hansen, 1974). (p) Diurnal cycle (q) Parameterization of low, middle and high clouds based on threshold relative humidity for radiative transfer calculations. (r) Surface energy balance couple to the similarity theory (Krishnamurti et al., 1991). (s) Nonlinear normal mode initialization ± ve vertical modes. (t) Physical initialization (Krishnamurti et al., 1991). References Arkin PA, Joyce R, Janowiak J (1994) The estimation of global monthly mean rainfall using infrared satellite data: The GOES precipitation index (GPI). Remote Sens Rev 11: 107±124 Businger JA, Wyngard JC, Izumi Y, Bradley EF (1971) Flux pro le relationship in the atmospheric surface layer J Atmos Sci 28: 181±189 Gadgil S, Sajani S (1998) Monsoon Precipitation in the AMIP runs. Climate Dynamics 14: 659±689 Gairola RK, Krishnamurti TN (1992) Rain rates based on SSM/I, OLR and rainguage data sets. Meteorol Atmos Phys 50: 165±174 Harhvardan, Coresetti TG (1984) Long-wave parameterization for the UCLA/GLAS GCM. NASA Tech Memo 86072, 52 pp Kanamitsu M (1975) On the numerical prediction over a global tropical belt. Rep. 75±1, Dept of Meteorology, The Florida State University, 1±282 Kanamitsu M, Tada K, Kudo K, Sato N, Isa S (1983) Description of the JMA operational spectral Model. J Meteor Soc, Japan 61: 812±828 Krishnamurti TN, Bedi HS, Wei Han (1998) Organization of convection and monsoon forecasts. Meteorol Atmos Phys 67: 117±134 Krishnamurti TN, Bedi HS, Ingles K (1993) Physical Initialization using SSM/I rain rates. Tellus 45A: 247±269 Krishnamurti TN, Xue J, Bedi HS, Ingkles K, Oosterhof D (1991) Physical Initialization for numerical weather prediction over the tropics. Tellus 43AB: 53±81 Krishnamurti TN (1990) Monsoon prediction at different resolutions with a global spectrl model. Mausam 41: 234± 240 Krishnamurti TN, Low-Nam S, Pasch R (1983) Cumulus parameterization and rainfall rates II. Mon Wea Rev 111: 816±828 Lacis AA, Hansen JE (1974) A parameterization of the absorption of solar radiation in the earth's atmosphere. J Atmos Sci 31: 118±133 Louis JF (1979) A parametric model of vertical eddy uxes in the atmosphere. Bound-Layer Meteor 17: 187±202 Olson WS, Kummerow C, Heyms eld GM, Giglio L (1996) A method for combined passive-active microwave retrievals of cloud and precipitation pro les. J Appl Meteor 35: 1763±1789 Sperber KR, Sultan Hameed, Potter GL, Boyle JS (1994) Simulation of the northern summer monsoon in the ECMWF model: Sensitivity to horizontal resolution. Mon Wea Rev 122: 2461±2481 Sperber KR, Palmer TN (1996) Interannual tropical rainfall variability in general circulation model simulations associated with the AMIP. J Climate 11: 2727±2750 Stephenson DB, Chauvin F, Royer J-F (1998) Simulation of the Asian Summer Monsoon and its dependence on model horizontal resolution. J Metero Soc, Japan 76: 237±265 Tiedke M (1984) The sensitivity of the time-mean large-scale ow to cumulus convection in the ECMWF model. Large- Scale Workshop on the Convection in the Numerical Models. Reading, United Kingdom, ECMWF, 297±316 Wallace JM, Tibaldi S, Simmons AJ (1983) Reduction of systematic forecast errors in the ECMWF model through the introduction of envelope orography. Quart J Roy Meteor Soc 109: 683±718 Yanai M, Elbensen S, Chu JH (1973) Determination of bulk properties of tropical cloud clusters from larger scale heat and moisture budgets. J Atmos Sci 30: 611±627 Authors' addresses: B. Jha and T. N. Krishnamurti, Department of Meteorology, Florida State University, Tallahassee, Florida , USA; Z. Christides, T. J. Watson Laboratory IBM, Yorktown, New York, USA.