• Anthes, R. A., 1983: Regional model of the atmosphere in middle latitudes. Mon. Wea. Rev., 111 , 13061335.

  • Barnes, S. L., 1964: A technique for maximizing details in numerical weather map analysis. J. Appl. Meteor., 3 , 396409.

  • Benoit, R., , J. Cote, , and J. Mailhot, 1989: Inclusion of a TKE boundary layer parameterization in the Canadian regional finite-element model. Mon. Wea. Rev., 117 , 17261750.

    • Search Google Scholar
    • Export Citation
  • Benoit, R., , M. Desgagne, , P. Pellerin, , S. Pellerin, , Y. Chartier, , and S. Desjardins, 1997: The Canadian MC2: A semi-implicit semi-Lagrangian wide-band atmospheric model suited for fine-scale process studies and simulation. Mon. Wea. Rev., 125 , 23822415.

    • Search Google Scholar
    • Export Citation
  • Benoit, R., , N. Kouwen, , W. Yu, , S. Chamberland, , and P. Pellerin, 2003: Hydrometeorological aspects of the real-time ultrafinescale forecast support during the special observing period of the MAP. Hydrol. Earth Syst. Sci., 7 , 877889.

    • Search Google Scholar
    • Export Citation
  • Bosart, L. F., , and F. Sanders, 1991: An early-season coastal storm: Conceptual success and model failure. Mon. Wea. Rev., 119 , 28312851.

    • Search Google Scholar
    • Export Citation
  • Davies, C. A., 1992: A potential vorticity diagnosis of the importance of the initial structure and condensational heating in observed extratropical cyclogenesis. Mon. Wea. Rev., 120 , 24092428.

    • Search Google Scholar
    • Export Citation
  • Davies, C. A., , and K. A. Emanuel, 1991: Potential vorticity diagnostics of cyclogenesis. Mon. Wea. Rev., 119 , 19291953.

  • Davolio, S., , and A. Buzzi, 2004: A nudging scheme for the assimilation of precipitation data into a mesoscale model. Wea. Forecasting, 19 , 855871.

    • Search Google Scholar
    • Export Citation
  • Ertel, H., 1942: Ein nueur hydrodynanischer wirbelsatz. Meteor. Z., 59 , 277281.

  • Fillion, L., , and R. Errico, 1997: Variational assimilation of precipitation data using moist convective parameterization schemes. Mon. Wea. Rev., 125 , 29172942.

    • Search Google Scholar
    • Export Citation
  • Fillion, L., , and S. Belair, 2004: Tangent linear aspects of the Kain–Fritsch moist convective parameterization scheme. Mon. Wea. Rev., 132 , 24772494.

    • Search Google Scholar
    • Export Citation
  • Fritsch, J. M., , and C. F. Chappell, 1980: Numerical prediction of convectively driven mesoscale pressure systems. Part I: Convective parameterization. J. Atmos. Sci., 37 , 17221733.

    • Search Google Scholar
    • Export Citation
  • Fritsch, J. M., and Coauthors, 1998: Quantitative precipitation forecasting: Report of the eighth prospectus development team, U.S. Weather Research Program. Bull. Amer. Meteor. Soc., 79 , 285299.

    • Search Google Scholar
    • Export Citation
  • Garand, L., 1983: Some improvements and complements to the infrared emissivity algorithm including a parameterization of the absorption in the continuum region. J. Atmos. Sci., 40 , 230244.

    • Search Google Scholar
    • Export Citation
  • Garand, L., , and C. Grassotti, 1995: Toward an objective analysis of rainfall rate combining observations and short-term forecast model estimates. J. Appl. Meteor., 34 , 19621977.

    • Search Google Scholar
    • Export Citation
  • Grassotti, C., , R. N. Hoffman, , E. R. Vivoni, , and D. Entekhabi, 2003: Multiple-timescale intercomparison of two radar products and rain gauge observations over the Arkansas–Red River basin. Wea. Forecasting, 18 , 12071229.

    • Search Google Scholar
    • Export Citation
  • Guo, Y-R., , Y-H. Kuo, , J. Dudhia, , D. Parsons, , and C. Rocken, 2000: Four-dimensional variational data assimilation of heterogeneous mesoscale observations for a strong convective case. Mon. Wea. Rev., 128 , 619643.

    • Search Google Scholar
    • Export Citation
  • Hamill, T. S., 1999: Hypothesis tests for evaluating numerical precipitation forecasts. Wea. Forecasting, 14 , 155167.

  • Hammer, G. R., , and P. M. Steurer, 1997: Data set documentation for hourly precipitation data. NOAA/NCDC Tech. Doc. 3240, Documentation Series, Asheville, NC, 1–18.

  • Hoskins, B. J., , M. E. McIntyre, , and R. W. Robertson, 1985: On the use and significance of isentropic potential vorticity maps. Quart. J. Roy. Meteor. Soc., 111 , 877946.

    • Search Google Scholar
    • Export Citation
  • Huo, Z., , D-L. Zhang, , and J. Gyakum, 1998: An application of potential vorticity inversion to improving the numerical prediction of the March 1993 superstorm. Mon. Wea. Rev., 126 , 424436.

    • Search Google Scholar
    • Export Citation
  • Huo, Z., , D-L. Zhang, , and J. Gyakum, 1999: Interaction of potential vorticity anomalies in extratropical cyclogenesis. Part I: Static piecewise inversion. Mon. Wea. Rev., 127 , 25462561.

    • Search Google Scholar
    • Export Citation
  • Kain, J. S., , and J. M. Fritsch, 1990: A one-dimensional entraining/detraining plume model and its application in convective parameterization. J. Atmos. Sci., 47 , 27842802.

    • Search Google Scholar
    • Export Citation
  • Kasahara, A., , A. P. Mizzi, , and L. J. Donner, 1992: Impact of cumulus initialization on the spinup of precipitation forecasts in the Tropics. Mon. Wea. Rev., 120 , 13601380.

    • Search Google Scholar
    • Export Citation
  • Kong, F. Y., , and M. K. Yau, 1997: An explicit approach to microphysics in MC2. Atmos.–Ocean, 35 , 257291.

  • Krishnamurti, T. N., , H. S. Bedi, , and K. Ingles, 1993: Physical initialization using the SSM/I rain rate. Tellus, 45A , 247269.

  • Kuo, H. L., 1974: Further studies of the parameterization of the influence of cumulus convection on large scale flow. J. Atmos. Sci., 31 , 12321240.

    • Search Google Scholar
    • Export Citation
  • Lin, C. A., , L. Wen, , M. Beland, , and D. Chaumont, 2002: A coupled atmospheric-hydrological modeling study of the 1996 Ha! Ha! River basin flash flood in Quebec, Canada. Geophys. Res. Lett., 29 .1026, doi:10.1029/2001GL013827.

    • Search Google Scholar
    • Export Citation
  • Lin, Y., , and K. E. Mitchell, 2005: The NCEP stage II/IV hourly precipitation analyses. Preprints, 19th Conf. on Hydrology, San Diego, CA, Amer. Meteor. Soc., 72–73.

  • Marecal, V., , and J-F. Mahfouf, 2002: Four-dimensional variational assimilation of total column water vapor in rainy areas. Mon. Wea. Rev., 130 , 4358.

    • Search Google Scholar
    • Export Citation
  • Mekis, E., , and W. D. Hogg, 1999: Rehabilitation and analysis of Canadian daily precipitation time series. Atmos.–Ocean, 37 , 5385.

  • Milbrandt, J. A., , and M. K. Yau, 2001: A mesoscale modeling study of the 1996 Saguenay flood. Mon. Wea. Rev., 129 , 14191440.

  • Molinari, J., , and M. Dudek, 1992: Parameterization of convective precipitation in mesoscale numerical models: A critical review. Mon. Wea. Rev., 120 , 326344.

    • Search Google Scholar
    • Export Citation
  • Nagarajan, B., , M. K. Yau, , and D-L. Zhang, 2001: A numerical study of a mesoscale convective system during TOGA COARE. Part I: Model description and verification. Mon. Wea. Rev., 129 , 25012520.

    • Search Google Scholar
    • Export Citation
  • Olson, D. A., , N. W. Junker, , and B. Korty, 1995: Evaluation of 33 years of quantitative precipitation forecasting at the NMC. Wea. Forecasting, 10 , 498511.

    • Search Google Scholar
    • Export Citation
  • Rossby, C. G., 1940: Planetary flow patterns in the atmosphere. Quart. J. Roy. Meteor. Soc., 66 , 6887.

  • Stensrud, D. J., , and J. M. Fritsch, 1994: Mesoscale convective systems in weakly forced large-scale environments. Part II: Generation of a mesoscale initial condition. Mon. Wea. Rev., 122 , 20682083.

    • Search Google Scholar
    • Export Citation
  • Treadon, R. E., , H-L. Pan, , W-S. Wu, , Y. Lin, , W. S. Olson, , and R. J. Kuligowski, 2003: Global and regional moisture analyses at NCEP. Proc. ECMWF/GEWEX Workshop on Humidity Analysis, Reading, United Kingdom, ECMWF, 33–48.

  • Tsuyuki, T., , K. Koizumi, , and Y. Ishikawa, 2003: The JMA mesoscale 4D-Var system and assimilation of precipitation and moisture data. Proc. ECMWF/GEWEX Workshop on Humidity Analysis, Reading, United Kingdom, ECMWF, 59–68.

  • Verret, R., , L. Lefaivre, , J-G. Desmarais, , and T. Robinson, 1996: Intense precipitation in Quebec, July 19–20, 1996. Can. Meteor. Center Rev., 3 , 129.

    • Search Google Scholar
    • Export Citation
  • Zhang, D-L., , and J. M. Fritsch, 1986: Numerical simulation of the meso-β-scale structure and evolution of the 1977 Johnstown flood. Part I: Model description and verification. J. Atmos. Sci., 43 , 19131943.

    • Search Google Scholar
    • Export Citation
  • Zou, X., , and Y-H. Kuo, 1996: Rainfall assimilation through an optimal control of initial and boundary conditions in a limited-area mesoscale model. Mon. Wea. Rev., 124 , 28592882.

    • Search Google Scholar
    • Export Citation
  • View in gallery

    The NCEP reanalyses of 850-hPa temperature (dashed contours every 2°C) superposed with SLP (solid contours every 4 hPa) for (a) 0000 UTC 19 Jul 1996 and (c) 1200 UTC 20 Jul 1996. (b), (d) The corresponding 500-hPa geopotential height (solid contours every 6 decameters) superposed with absolute vorticity (dashed contours every 2 × 10−5 s−1). Subjectively drawn front and trough lines are shown in (a), (c), and (b), respectively. The AVHRR satellite brightness temperature (shaded) with U.S. radar reflectivity composite (solid contours every 10 dBZ) are shown for (e) 0000 UTC and (f) 1200 UTC 19 Jul 1996. Blank regions in (e) and (f) depict regions with missing satellite data.

  • View in gallery

    The 48-h objectively analyzed rain gauge precipitation data for the period between 1300 UTC 19 Jul and 1300 UTC 21 Jul (shaded and contoured) is shown along with central SLP tracks from NCEP reanalysis (thin solid), the CTL (dashed–dotted), and NCEP (thick solid) simulation data. The three precipitation maxima are shown in mm. The central SLP locations are depicted every 6 h between 0000 UTC 19 Jul and and 1200 UTC 21 Jul. Maniwaki, Caribou, and Sept-Iles are the upper-air radiosonde stations. R1, R2, and R3 depict the location of three rain gauges whose measurements are used in evaluating the temporal variation of precipitation.

  • View in gallery

    (a) The coarse (A) and fine mesh domain (B) is shown with spatial distribution of the rain gauges. (b) The objectively analyzed rain gauge data show the precipitation rate (mm h−1) computed from the hourly rainfall accumulation ending 0000 UTC 19 Jul 1996. The 0.1, 1.0, 5.0, 10.0, 25.0, and 50.0 mm h−1 contours are plotted with the shaded area depicting regions with a rain rate greater than 0.1 mm h−1. The boxes P1, P2, P3, and P4 in (b) represent the areas occupied by the four distinct precipitation systems.

  • View in gallery

    (a) The difference in CAPE in J kg−1 between the CTL and NCEP runs (negative dashed and positive shaded) is shown at 0000 UTC 19 Jul 1996 over the regions P1–P4. (b) The shaded area (solid line) shows the corresponding difference moisture (temperature) in g kg−1 (°C). The 0.5° and 1.0°C temperature contours are shown. The CAPE is computed by mixing the lowest 125-hPa air layer and the moisture and temperature difference field represents a layer average over the lowest 125 hPa.

  • View in gallery

    A skew T–logp plot of the area-averaged temperature and dewpoint profiles over P4 (Fig. 3b) at 0000 UTC 19 Jul 1996. Thick solid (dotted) lines indicate temperature and dewpoint from CTL (NCEP).

  • View in gallery

    The spatial distribution of the 48-h precipitation accumulation (mm) ending 1200 UTC 21 Jul 1996 is shown for (a) CTL and (b) NCEP. (c) The surface cyclone location between 1800 UTC 19 Jul and 1200 UTC 21 Jul from CTL (C) and NCEP (N) is plotted every 3 h. (d) The TS (with bias equalized to CTL) and BS (unequalized) is solid (dashed) for the CTL (NCEP).

  • View in gallery

    A time series of the total accumulated precipitation (mm), plotted every 3 h, from rain gauge (OBS), CTL, and NCEP is shown for (a) R1, (b) R2, and (c) R3. The rain gauge locations R1, R2, and R3 are shown in Fig. 2. Rain gauge observations are absent before 9 h.

  • View in gallery

    A skew T–logp plot of the observed (solid) and model sounding (dashed) from CTL at 1200 UTC 21 Jul 1996 for (a) Caribou, (b) Sept-Iles, and (c) Maniwaki (see Fig. 2). The observed wind is plotted at the rightmost end of (a)–(c). A full barb is 5 m s−1. (d) The central SLP (hPa) from CTL, NCEP, and NCEP reanalysis (NCEP-REANALYSIS) is shown as a function of time.

  • View in gallery

    The 12-h accumulated precipitation (mm) ending at 1200 UTC 19 Jul superposed with the 500-hPa geopotential height (decameters) from (a) CTL and (b) NCEP. The 1200 UTC 19 Jul northern and southern troughs at 500 hPa are shown by the thick dotted line.

  • View in gallery

    The CTL-minus-NCEP layer-averaged potential vorticity anomaly (PVU) (a) Qm and (b) Qd at 1200 UTC 19 Jul. The Qd (Qm) is averaged between 500 and 200 hPa (900 and 500 hPa). The contours −0.2, −0.1, 0.1, and 0.2 (−0.5, −0.1, 0.1, and 0.5) PVU are shown for Qd (Qm). Shaded regions indicate values greater (less) than 0.1 (−0.1) PVU for Qd (Qm), where 1 PVU = 10−6 m2 K s−1 kg−1.

  • View in gallery

    The geopotential height (decameters) superposed with the CTL-minus-NCEP zonal wind component (m s−1) at 300 hPa associated with the southern trough (Figs. 9a,b) PV anomaly Qd at 1200 UTC 19 Jul. The northern and southern troughs are shown by the thick solid lines. Although the southern trough is not evident, it is subjectively depicted based on its location at 500 hPa (Fig. 9a).

  • View in gallery

    The PR48 TS (with bias equalized to CTL) and BS (unequalized) from CTL and two sensitivity experiments performed by setting the standard deviation error for observed precipitation rate to 5% (OBS ERR 5%) and 50% (OBS ERR 50%) of the background field. The standard deviation error in CTL is assumed to be 25%.

  • View in gallery

    A time series of area-averaged accumulated precipitation (mm) and the 500-hPa vertical motion in p coordinates (Pa s−1) from (a) CTL and (b) 50P are shown every 3 h. The area average is performed over the region enclosed by the 150-mm precipitation contour in Fig. 6a.

  • View in gallery

    The PR48 TS (with bias equalized to CTL) and BS (unequalized) from CTL and four sensitivity experiments: No-P1, No-P2, No-P3, and No-P4. The sensitivity experiment No-P1 refers to the absence of the temperature and moisture profiles used in CTL obtained through the 1DVAR scheme over P1. The other sensitivity experiments are defined similarly.

  • View in gallery

    (a) The 48-h accumulated convective and (b) grid-resolved precipitation (mm) ending 1200 UTC 21 Jul 1996 from experiment No-P4. (c) The 48-h accumulated precipitation (convective and grid resolved) from No-P2.

All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 11 11 7
PDF Downloads 2 2 0

A Numerical Study of the 1996 Saguenay Flood Cyclone: Effect of Assimilation of Precipitation Data on Quantitative Precipitation Forecasts

View More View Less
  • 1 Department of Atmospheric and Oceanic Sciences, McGill University, Montreal, Quebec, Canada
  • 2 Meteorological Service of Canada, Dorval, Quebec, Canada
© Get Permissions
Full access

Abstract

A one-dimensional variational (1DVAR) technique is applied to assimilate rain gauge precipitation data to extend the predictability of the Saguenay flood cyclone associated with a trough-merger event on 19–21 July 1996 in the Saguenay-Lac-Saint-Jean region of Quebec, Canada.

Two 60-h simulations initialized at 0000 UTC 19 July were performed with the Canadian Mesoscale Compressible Community (MC2) model. The control (CTL) and NCEP simulations were initialized with the enhanced temperature and moisture profiles obtained from the 1DVAR scheme and the NCEP reanalysis data, respectively.

Compared to observations, the CTL simulation reasonably reproduced the observed mass and wind fields and showed a marked improvement in the threat scores for heavy precipitation. The CTL run captured the observed spatial and temporal distribution of precipitation but overpredicted the area of precipitation. Sensitivity experiments showed that the threat (bias) scores are less (somewhat) sensitive to the specification of the observation error of the precipitation data. Of the four precipitation systems present at model initial time, the systems in the vicinity of the southern trough had the biggest impact on the threat score.

Potential vorticity diagnostics of the CTL simulation suggested that the initial temperature and moisture field near the southern trough decreased the condensational heating relative to NCEP. This resulted in a stronger zonal wind component in the upper levels associated with the southern trough in CTL that retarded the eastward propagation of the northern trough, resulting in a correct placement of the surface precipitation and an improvement in the threat scores relative to NCEP.

Corresponding author address: Dr. Badrinath Nagarajan, Department of Atmospheric and Oceanic Sciences, McGill University, 805 Sherbrooke St. West, Montreal, PQ H3A 2K6, Canada. Email: badrinath.nagarajan@elf.mcgill.ca

Abstract

A one-dimensional variational (1DVAR) technique is applied to assimilate rain gauge precipitation data to extend the predictability of the Saguenay flood cyclone associated with a trough-merger event on 19–21 July 1996 in the Saguenay-Lac-Saint-Jean region of Quebec, Canada.

Two 60-h simulations initialized at 0000 UTC 19 July were performed with the Canadian Mesoscale Compressible Community (MC2) model. The control (CTL) and NCEP simulations were initialized with the enhanced temperature and moisture profiles obtained from the 1DVAR scheme and the NCEP reanalysis data, respectively.

Compared to observations, the CTL simulation reasonably reproduced the observed mass and wind fields and showed a marked improvement in the threat scores for heavy precipitation. The CTL run captured the observed spatial and temporal distribution of precipitation but overpredicted the area of precipitation. Sensitivity experiments showed that the threat (bias) scores are less (somewhat) sensitive to the specification of the observation error of the precipitation data. Of the four precipitation systems present at model initial time, the systems in the vicinity of the southern trough had the biggest impact on the threat score.

Potential vorticity diagnostics of the CTL simulation suggested that the initial temperature and moisture field near the southern trough decreased the condensational heating relative to NCEP. This resulted in a stronger zonal wind component in the upper levels associated with the southern trough in CTL that retarded the eastward propagation of the northern trough, resulting in a correct placement of the surface precipitation and an improvement in the threat scores relative to NCEP.

Corresponding author address: Dr. Badrinath Nagarajan, Department of Atmospheric and Oceanic Sciences, McGill University, 805 Sherbrooke St. West, Montreal, PQ H3A 2K6, Canada. Email: badrinath.nagarajan@elf.mcgill.ca

1. Introduction

Quantitative precipitation forecasting (QPF), an important element of the daily weather forecast, is often quantified in terms of the threat score (Anthes 1983). Over the continental United States, Olson et al. (1995) showed that the threat scores are the lowest during the summer season when the most significant precipitation events occurred. Fritsch et al. (1998) attributed the poor QPF to deficiencies in four areas: basic understanding of the physics of rainstorms, effects of synoptic waves and surface boundaries, QPF techniques, and precipitation estimates for forecast validation. They suggested that one way to improve the QPF was to initialize entire precipitation systems, particularly for the subsynoptic scales, in model initial conditions through the assimilation of precipitation data.

Various techniques have been used to improve the large-scale model initial conditions, including physical initialization (Krishnamurti et al. 1993) and diabatic initialization (Kasahara et al. 1992). Traditionally, subsynoptic initial conditions have been improved, based on the presence of precipitation systems observed by satellites, by modifying the initial temperature and moisture fields (Stensrud and Fritsch 1994; Zhang and Fritsch 1986; Nagarajan et al. 2001). Guo et al. (2000) and Zou and Kuo (1996) employed a four-dimensional variational (4DVAR) technique to assimilate precipitation data to improve mesoscale initial conditions and showed an improvement in the QPF associated with a prefrontal rainband and a mesoscale convective system (MCS), respectively. Davolio and Buzzi (2004) used a nudging procedure with a mesoscale model to assimilate rainfall data associated with the passage of a frontal system over the Alps and an orographic precipitation event. Marecal and Mahfouf (2002) introduced the one-dimensional plus four-dimensional variational approach to assimilate precipitation data at the European Centre for Medium-Range Weather Forecasts (ECMWF). The Japan Meteorological Agency employs a 4DVAR system to assimilate radar-derived precipitation data (Tsuyuki et al. 2003). Satellite-derived precipitation data is assimilated at the National Centers for Environmental Prediction (NCEP) using a 3DVAR system (Treadon et al. 2003).

We examine the case of heavy rainfall that led to severe flooding in the Saguenay region of Quebec, Canada, between 19 and 21 July 1996 (Milbrandt and Yau 2001). The Canadian Meteorological Center (CMC) forecasts initialized at 0000 UTC 19 July 1996 poorly predicted the location and intensity of the precipitation (Milbrandt and Yau 2001; Verret et al. 1996). However the forecast was more accurate when the model was initialized 12 h later, suggesting that errors in the initial conditions may be responsible for the poor QPF in the 0000 UTC initialized run. This lack of skill is not surprising as several mesoscale precipitation systems were present over the Great Lakes region (Fig. 1e) at 0000 UTC. These systems may modify the initial temperature and moisture fields to affect the predictability of the Saguenay flood storm.

The objective of this study is to improve the prediction of the heavy rainfall associated with the Saguenay flood cyclone by improving model initial conditions through the application of a 1DVAR rainfall data assimilation scheme (Fillion and Errico 1997; Fillion and Belair 2004). The paper is organized as follows. A brief synoptic overview with a description of the mesoscale precipitation systems is presented in section 2. Section 3 contains model description and the generation of initial conditions using the 1DVAR scheme. The control (CTL) experiment is verified against observations in section 4. Section 5 discusses the improved QPF of the Saguenay flood cyclone in terms of the surface cyclone track using potential vorticity inversion techniques. Section 6 presents the sensitivity of the results to parameters in the 1DVAR scheme and to the systematic removal of various mesoscale precipitation systems in the initial conditions. A summary and concluding remarks are given in the final section.

2. Case overview

The Saguenay flood cyclone originated as a low pressure system over southern Manitoba, Canada. The 0000 UTC 19 July NCEP reanalysis depicts a low pressure system with a sea level pressure (SLP) of 1003 hPa located just south of a baroclinic zone between Lake Superior and Lake Michigan (Fig. 1a). The low at the surface is situated directly under a 500-hPa trough but downstream of another trough lying over the Hudson Bay (Fig. 1b). The cyclone explosively deepened (14 hPa in 12 h) between 1800 UTC 19 July and 0600 UTC 20 July (Milbrandt and Yau 2001).

Between 0000 UTC 19 July and 1200 UTC 20 July, the cyclone tracked eastward (Fig. 2). At 1200 UTC 20 July, the storm attained the deepest central SLP of 980 hPa (Fig. 1c). At the same time, strong thermal advection occurred at the low levels with the thermal field exhibiting an S-shaped pattern (Fig. 1c). At 500 hPa, the two troughs present initially merged and formed a closed circulation (Fig. 1d) that extended up to 250 hPa (not shown). From Figs. 1c,d, it is clear that the system is equivalent barotropic. The cyclone gradually filled over the next 24 h and remained quasi-stationary over the Gaspé Peninsula.

Since errors in the initial conditions may be responsible for a poor QPF when the CMC operational model is initialized at 0000 UTC 19 July, we describe the subsynoptic systems present at model initial time (Figs. 1e,f) making use of the U.S. composite radar reflectivity and the Advanced Very High Resolution Radiometer (AVHRR) data. The radar composite at this time depicts four precipitating systems: one located over the Great Lakes, the second in the region extending between Pennsylvania and the Atlantic seaboard, the third in the vicinity of Lake Ontario, and the fourth over Louisiana (Fig. 1e). These systems are associated with low infrared brightness temperatures (234–208 K), indicative of high (cold) cloud tops. The occurrence of a fifth precipitating system over Quebec is suggested by the presence of cold clouds (234–228 K; Fig. 1f). Figures 1e (1a) indicate that the first (second) precipitating system is associated with the surface low pressure system (surface trough).

Over the next 12 h, the first and fourth precipitating systems dissipated while the third system intensified (Fig. 1f). The part of the second system situated over Pennsylvania (the Atlantic Seaboard) also undergoes intensification (dissipation).

The most intense rainfall occurred during the explosive deepening phase of the cyclone. The 48-h accumulated precipitation depicts three maxima and the rainfall was confined to central-eastern Quebec north of the track of the surface cyclone (Fig. 2). Milbrandt and Yau (2001) showed that the northernmost precipitation maximum resulted from a sustained saturated ascent forced by a confluent flow at the low levels. The southernmost maximum was attributed to an increase in the vertical motion and moisture convergence due to the presence of the Saguenay valley and a mountain southeast (SE) (marked in Fig. 2) of the valley. Thus the spatial distribution and intensity of the accumulated precipitation largely arose from the favorable positioning of the surface cyclone with respect to the topography and the interaction of the low-level flow with orographic features in the Saguenay region.

3. Modeling strategy

a. Model description

The cyclone is simulated using the Canadian Mesoscale Compressible Community (MC2) model (Benoit et al. 1997). Two high-resolution (20 km) simulations, initialized at 0000 UTC 19 July 1996, were performed over domain B (Fig. 3a). The lateral boundary conditions were updated every 3 h using a coarse simulation (80 km) over domain A (Fig. 3a) initialized at the same time. One of the high-resolution simulations was initialized with the NCEP reanalysis (experiment NCEP). The other control run (experiment CTL) used an improved initial condition obtained using the 1DVAR scheme.

The number of grid points along x, y and z directions for the coarse and the high resolution runs is 160 × 100 × 25 and 195 × 195 × 25, respectively. The model physical processes include surface fluxes, atmospheric boundary layer (ABL) transport (Benoit et al. 1989), short- and longwave radiation (Garand 1983) invoked every 30 min, deep cumulus convection, and explicit microphysics (Kong and Yau 1997). Deep cumulus convection is represented by the Kuo (Kain–Fritsch) cumulus parameterization scheme (CPS) in the coarse (high resolution) run (Kuo 1974; Kain and Fritsch 1990).

b. One-dimensional variational assimilation scheme

The initial temperature and moisture field used in the CTL run was obtained by assimilating surface precipitation data.

The 1DVAR problem is expressed in terms of a functional J(x)
i1520-0493-134-5-1371-e1
and the gradient of J(x) at x = xn as
i1520-0493-134-5-1371-e2
where x = (T, qυ)T is the state vector of dimension N = 2Nk, Nk is the number of vertical levels, and T, qυ are vectors of vertical profiles of temperature and specific humidity; xb = background state vector at the observation location; yo = instantaneous observed precipitation rate; B = a priori (background) error covariance matrix; R = observational error variance; HT = KTFT is the transpose (or adjoint) of operator H, KT is the transpose (or adjoint) of the tangent-linear operator K where K represents the Jacobian of the Kain–Fritsch CPS. A perturbative approach (Fillion and Errico 1997) is used to to perform the adjoint computations. Note that the Jacobians were kept fixed during the minimization and represents a good approximation for the majority of vertical convective profiles but fails in certain cases where convection is highly nonlinear when the minimization becomes pathological (e.g., convection deactivated; Fillion and Errico 1997); and H = nonlinear observation operator. The operator H links the state vector x with the instantaneous precipitation rate observations and is a composite of two operators (Fillion and Errico 1997):
i1520-0493-134-5-1371-eq1
where C is a one-dimensional nonlinear operator (i.e., Kain–Fritsch CPS) that acts in the vertical on the state vector x to yield a vector of changes Δx (i.e., ΔT, Δqv). The linear operator F is given by y = FΔx, where y is the estimated observation value of the instantaneous precipitation rate. The minimization of the functional J is attained by passing J and its gradient [Eqs. (1) and (2)] to a variable-storage quasi-Newton module.

Note that no large-scale precipitation process was included in the observation operator since we assumed that deep convection brings most of the contribution to the instantaneous rain rate in our case study.

1) Specification of background field

The coarse- and high-resolution model simulations are initialized at 0000 UTC 19 July using the NCEP reanalysis. Initial tests with the 1DVAR scheme using NCEP reanalysis data for the background field indicated that in contrast to observations, deep convection was initiated at a smaller number of grid points. The convective trigger function in the Kain–Fritsch CPS depends on the vertical motion (Fritsch and Chappell 1980), with stronger upward motion favoring deep convection. But the horizontal coarse-resolution (about 250 km) wind field of the reanalysis data yielded weaker vertical motion. This led to the sparseness of initial points of deep convection in the 1DVAR scheme. To overcome this problem we employed the horizontal higher-resolution (24 km) CMC operational analyses for the background field.

2) Specification of background error statistics

A Gaussian autocorrelation model is used to specify the background error for both temperature and specific humidity (Fillion and Errico 1997). The background error variances are allowed to vary in the vertical and are deduced from the ones used by Fillion and Errico (1997).

3) Specification of observed precipitation rate data and error statistics

The operational hourly accumulation of radar-derived precipitation data over the continental United States constitutes a potential source of data in the 1DVAR analysis (Lin and Mitchell 2005). However a lack of radar precipitation estimates over Canada precludes its use in the present study. Thus, only rain gauge observations were used in this study.

Figure 3a shows the spatial distribution of the rain gauges. The hourly precipitation accumulation data over the United States (Canada) is produced and quality controlled by the National Climate Data Center (Meteorological Service of Canada; Hammer and Steurer 1997; Mekis and Hogg 1999).

The hourly rainfall accumulation data from the rain gauges are used to construct the 0000 UTC 19 July 1996 rainfall map (Fig. 3b). Specifically, an objective analysis of the 1-h accumulated precipitation amount (2300 UTC 18 July–0000 UTC 19 July) is performed using the Barnes (1964) scheme with a radius of influence of 35 km. Rain gauge–derived objective precipitation analyses have also been used by Milbrandt and Yau (2001) and Grassotti et al. (2003). The former study used the precipitation analysis to verify the QPF produced by the MC2 model while the latter study intercompared a gauge-adjusted radar analysis, a radar-only analysis, and individual rain gauge measurements.

Figure 3b shows that at the model initial time, surface precipitation occurs in association with precipitating systems depicted by the radar measurement over the Great Lakes, Lake Ontario, Pennsylvania, and Louisiana (Fig. 1e). However weak precipitation rates associated with cold cloud shields are observed over Quebec (Fig. 1e).

Since the only physical process considered by the 1DVAR scheme is deep convection, the scheme should be applied to areas exhibiting deep convection. A rain-rate threshold is needed to identify regions of deep convection. For simplicity, we set the threshold to a value greater than zero (i.e., all precipitating areas) as we have only the hourly accumulated rainfall at our disposal and the instantaneous rain rate inferred from the hourly accumulation and a spatial smoothing of the rain rate by the Barnes objective analysis scheme will underestimate the actual rain rate.

The standard deviation error for observed precipitation rate from various observing systems (e.g., surface gauges, satellite passive microwave retrievals, or ground-based radars) can be as high as the background precipitation rate (Garand and Grassotti 1995). Thus, the standard deviation error in precipitation is set to 25% of the background field. Although this choice is arbitrary, we shall show in the next section that the results are not sensitive to the specification of this error.

c. Generation of initial temperature and moisture fields

The 1DVAR scheme yields temperature and moisture increments by assimilating surface precipitation (Fig. 3b). These increments are used to enhance the temperature and specific humidity fields of the NCEP reanalysis. The enhancement is applied only at the model initial time.

The impact of the assimilation of precipitation on the initial temperature and moisture fields can be gauged by the difference in convective available potential energy (CAPE) between the CTL and the NCEP runs (Fig. 4a). The CAPE is computed by mixing the air in the lowest 125 hPa. Using a shallower 60-hPa mixed layer yields similar results. Recall from section 2 that the initial condition is characterized by four precipitating systems over regions P1, P2, P3, and P4 (Fig. 3b). Figure 4a shows that all the regions, with the exception of P4, exhibit a significant change in CAPE. The decrease in CAPE over P1 and P2 arises from a large decrease (between 0.5 and 2.0 g kg−1) in specific humidity in the lowest 125-hPa layer between the CTL and NCEP runs (Fig. 4b). Some of the decrease in specific humidity is also accompanied by an increase in temperature (P1 and P2 in Fig. 4b). Thus the 1DVAR scheme warms and dries the ABL over P1 and P2. For P3, CAPE increases along the Atlantic Seaboard but decreases to the northwest. The increase (decrease) in CAPE is due to the increase (decrease) in specific humidity between the two experiments (Fig. 4b). As for regions P1 and P2, some of the decrease in specific humidity is accompanied by an increase in temperature. Thus, the assimilation of precipitation results in a drier ABL over much of P1, P2, and P3, but a moister ABL along the Atlantic Seaboard.

Although there is no significant change in CAPE over P4, a skew T–logp plot of the domain average (over P4) temperature and dewpoint temperature shows that CTL exhibits warming and moistening (relative to the NCEP run) between 1000 and 500 hPa (Fig. 5). The lower part of the troposphere over P4 is rendered stable (postconvective “onion-shaped” sounding) while there is warming and moistening at the midlevels. Thus, the midtroposphere is rendered more favorable for the formation of grid-scale precipitation as the dry layer between 900 and 500 hPa in experiment NCEP is eliminated.

4. Verification of the control simulation CTL

a. Precipitation

Since all of the precipitation occurred between 1200 UTC 19 July and 1200 UTC 21 July, we compared the simulated rainfall accumulation during the last 48 h against a gridded dataset of observed precipitation (Fig. 2). Note that the rain gauge density is high in southeastern Quebec where much of the rainfall occurred (Fig. 3a).

Overall, the 48-h accumulated precipitation (PR48) from CTL qualitatively reproduces the observed spatial distribution of precipitation (Figs. 6a and 2). Three maxima with magnitudes of 225, 190, and 158 mm are shown in the simulated PR48 as compared to the observed maxima of 246, 170, and 150 mm, respectively. Thus, the CTL run reproduces realistically the location and intensity of PR48 associated with the Saguenay flood cyclone. On the other hand, the PR48 simulated by the NCEP run exhibits only two precipitation maxima and the southernmost maximum is located east of the observed (Figs. 6b and 2). A plot of the hourly precipitation accumulation from the rain gauge in the Saguenay region (Fig. 2) suggests that much of the rain fell between 1800 UTC 19 July and 1800 UTC 20 July (not shown). During the first 12–15 h of this period, the surface cyclone track (Fig. 6c) and the 500-hPa geopotential height minimum (not shown) in the NCEP simulation is located east of the track in CTL and results in the occurrence of the precipitation maximum east of the observed. The magnitudes of the two simulated PR48 maxima in the NCEP run are 250 and 240 mm. The northern maximum is overpredicted with respect to the observed. Thus, the simulated PR48 in the NCEP run misses the central peak, places the southern maximum east of the observed, and overpredicts the northern maximum.

A quantitative evaluation of the simulated PR48 is performed through the computation of the threat score (TS) and bias score (BS) defined by Anthes (1983) as
i1520-0493-134-5-1371-eq2
and
i1520-0493-134-5-1371-eq3
where C is the number of grid points correctly forecast to receive a threshold amount of precipitation, F is the number of grid points forecast to receive this amount, and R is the number of grid points where the threshold amount is observed. These parameters F, R, and C can alternatively be defined as the area enclosed by the forecast isohyet, observed isohyet, and the intersection of the observed and forecast isohyets, respectively. The values for TS range from 0 to 1, with a higher value suggesting a better forecast. The BS can be a positive number, with values above (below) 1 denoting an overprediction (underprediction) of the area of precipitation.

The TS and BS are computed by interpolating the simulated PR48 to the location of the rain gauges. Since our objective is to evaluate the skill of the model in simulating the large precipitation values, the rain gauges reporting a PR48 less than 50 mm are excluded from the TS and BS computation. Following Hamill (1999), the BSs are equalized to those of the CTL before computing the TSs. Figure 6d shows the TS and BS for PR48 thresholds between 50 and 225 mm for the CTL and NCEP simulations. The TS of the CTL is greater than that in NCEP when threshold exceeds 105 mm. However, the TS is a little better in NCEP for threshold values between 75 and 105 mm. This suggests that relative to NCEP, experiment CTL shows significant skill in capturing the high rainfall values. However, for PR48 thresholds between 50 and 200 mm, both CTL and NCEP significantly overpredict the area of precipitation (Fig. 6d).

The TS and BS evaluated the spatial distribution of the simulated PR48. The temporal behavior of the simulated precipitation is examined by comparing the rainfall accumulation from CTL and NCEP against rain gauge observations (Fig. 7). As for the computation of TS and BS, the 3-hourly accumulated precipitation from CTL and NCEP are interpolated to the location of the rain gauges.

Overall, the time evolution of the accumulated precipitation in CTL is in close agreement with the observed (Fig. 7). The CTL exhibits little or no delay in the onset of precipitation (Fig. 7). In other words, the assimilation of precipitation in CTL almost removes the model precipitation spinup. In contrast, a spinup period of up to 6 h is evident in NCEP (Fig. 7). Thus, the CTL not only reasonably captures the spatial distribution and the temporal evolution of the precipitation but also exhibits a significantly reduced spinup period.

b. Wind and mass fields

Figure 8 compares the CTL and observed soundings at Caribou, Sept-Iles, and Maniwaki, Canada (locations shown in Fig. 2). Compared to the observed, the model sounding shows saturation in the lower troposphere (900–650 hPa) over Caribou and Sept-Iles with saturation confined to a shallow layer near 900 hPa over Maniwaki (Fig. 8c). Thus, below the tropopause, the simulated (at 60 h) moisture, temperature, and wind profiles compare reasonably well with the observed (Figs. 8a–c). However, the simulated temperature deviates from the observed temperature in the stratosphere (approximately between 400 and 100 hPa) at Caribou and Sept-Iles (Figs. 8a,b). The time evolution of the simulated soundings at these locations compare favorably with the observed (not shown).

Figure 2 shows a comparison of the location of the surface cyclone from CTL and from the NCEP reanalysis. The simulated cyclone tracks northwest (south) of the observed between 0000 UTC 19 July and 0000 UTC 20 July (0600 UTC 20 July and 0600 UTC 21 July). The northwestward bias in the simulated track was also reported by Milbrandt and Yau (2001). Figure 8d shows the time evolution of the central minimum SLP every 6 h from the CTL simulation and the corresponding NCEP reanalysis. During the first 30 h, the simulated SLP agrees to within about 2 hPa with the observed. However, the simulated cyclone deepens considerably compared to the observed and attains its lowest central pressure of 973 hPa at 42 h (observed 980 hPa at 36 h). Milbrandt and Yau (2001) showed that the Saguenay flood cyclone is affected strongly by latent heat release as no deepening of the storm occurs when latent heating from condensational processes (subgrid and grid scale) was turned off. Thus, the overprediction of the area of precipitation in CTL (Fig. 6d) may result in a deeper cyclone being simulated (Fig. 8d). Even though the CTL cyclone is deeper, the rate of filling between 42 and 60 h is similar to the observed storm. Thus CTL reasonably reproduces the observed mass and wind fields. On the other hand, the SLP in NCEP is higher than in CTL most of the time although the temporal variation remains similar. The location of the surface cyclone in NCEP is always east of that in CTL after 1800 UTC 19 July (Fig. 6c).

5. Discussion on QPF improvement in CTL

Milbrandt and Yau (2001) showed that the correct placement of the surface cyclone relative to the topography in the Saguenay region is important for the proper prediction of the location and intensity of the precipitation maxima. They also demonstrated that in turn, the track of the surface cyclone is affected by the interaction of two upper-level troughs (Fig. 1b), referred to as the northern and southern trough, respectively. To determine whether the assimilation of precipitation can affect the interaction of the troughs to place the CTL cyclone in the proper location to improve QPF, we apply the potential vorticity (PV) inversion technique to the difference PV field between CTL and NCEP (Hoskins et al. 1985; Huo et al. 1998, 1999; Milbrandt and Yau 2001).

The PV concept is useful because of its invertibility and conservation properties. Davies and Emanuel (1991) developed a piecewise PV inversion technique to study the contribution of upper- and lower-level processes on cyclogenesis. For a fully baroclinic and compressible flow, the Ertel potential vorticity (EPV) is given by
i1520-0493-134-5-1371-eq4
where ρ, η, and θ, respectively, denote the air density, the three-dimensional absolute vorticity, and the potential temperature. EPV is conserved following motion in an inviscid adiabatic flow (Rossby 1940; Ertel 1942). In the piecewise PV inversion technique, the total EPV field is decomposed into a mean and an anomaly field. The latter is composed of individual PV anomalies. The PV inversion technique is applied to deduce the wind, height, and temperature fields associated with the anomalies (Davies 1992; Huo et al. 1998).

The total PV anomaly is defined as a deviation from a mean PV that is a 6-day average computed from the NCEP reanalysis data starting from 1200 UTC 17 July 1996. The total PV anomaly is partitioned into perturbations, including those associated with tropopause depression (Qd), and latent heat release in the lower troposphere (Qm). Specifically, Qd is defined as all positive PV anomalies in dry air (relative humidity less than 30%) between 800 and 200 hPa. The quantity Qm denotes all positive PV anomalies in moist air (relative humidity greater than 70%) between 900 and 500 hPa generated mainly by latent heat release from condensation but also include PV advected out of saturated precipitation regions (Huo et al. 1999; Milbrandt and Yau 2001).

Figure 9 depicts the 12-h accumulated precipitation starting from 0000 UTC 19 July. In the region under the southern trough, the accumulated precipitation is larger in NCEP (Fig. 9b) than in CTL (Fig. 9a). This phenomenon is consistent with the result that the assimilation of precipitation data yields a drier and warmer environment under the southern trough in CTL (Fig. 4b), accompanied by a reduction in CAPE (Fig. 4a).

Figure 10 shows the difference field of layer-averaged Qm and Qd between CTL and NCEP at 1200 UTC 19 July. Over the region of the southern trough, the layer-averaged difference Qm is mainly negative (Fig. 10a), consistent with the reduced latent heating in CTL relative to NCEP. The larger latent heat release in NCEP caused stronger negative PV anomaly aloft. As a result, the layer-averaged difference Qd becomes positive in the region of the southern trough (Fig. 10b).

Figure 11 shows that the zonal wind component associated with the difference Qd anomaly at 300 hPa obtained using the PV inversion technique. The difference zonal wind is positive in the region of the southern trough. It becomes negative ahead of the northern trough. The zonal wind is therefore stronger in CTL than in NCEP ahead of the northern trough. The results for other levels from 500 to 200 hPa are qualitatively similar (not shown) except that the magnitudes of the positive and negative zonal wind differences are smaller than that at 300 hPa. The weaker zonal flow in NCEP accelerates the eastward propagation of the northern trough and placed the NCEP surface cyclone between 40 and 140 km east of CTL between 1800 UTC 19 July and 0300 UTC 20 July (Fig. 6d). Since most of the rainfall occurred during this period (not shown), the eastward displacement of the cyclone in the NCEP simulation relative to the observed location led to an improper interaction of the surface flow with topography and the surface precipitation is placed east of the observed.

6. Sensitivity experiments

a. 1DVAR parameters

In section 3b the standard deviation error for the observed precipitation rate was set to 25% of the background field. The sensitivity of the QPF is studied by performing two experiments with the standard deviation error set to 5% and 50%, with other aspects the same as in CTL.

Figure 12a shows that the TS for PR48 does not change significantly between 50 and 200 mm. For threshold values exceeding 200 mm, since the TS and BS are computed by interpolating the simulated PR48 to the location of the rain gauges, and only two rain gauge locations in the sensitivity experiments have PR48 larger than 200 mm, the results are not considered significant. On the other hand, the BS shows little sensitivity when the standard deviation error varies between 5% and 25% (Fig. 12b), but decreases for PR48 thresholds between 120 and 200 mm when the standard deviation error is set to 50%. Thus the TS (BS) shows no significant (some) sensitivity to the variation of the observed precipitation error statistics.

To understand the lower BS scores evident in the sensitivity experiment with the standard deviation error set to 50% (50P), we note that grid-resolved precipitation is the dominant contributor to PR48 (not shown). Since formation of grid-resolved precipitation depends on ascending motion occurring over a deep saturated column of air (Molinari and Dudek 1992), it suggests that any reduction in the area of accumulated precipitation (or BS) is associated with the moisture field over P4 and/or the strength of vertical motion.

The 1DVAR assimilation procedure seeks to adjust the temperature and moisture profiles that fit, in a least square sense, the observation within both background and observation errors (Marecal and Mahfouf 2002). A large observation error (i.e., 50%) implies that the retrieved temperature and moisture profile will be closer to the background (i.e., NCEP reanalysis). However, a skew T–logp plot of the area-averaged (over P4) sounding from 50P is almost identical to CTL suggesting that the strength of the vertical motion may have contributed to the reduced area of precipitation (not shown).

The time series of the area-averaged (region enclosed by 150-mm PR48 contour in CTL) total accumulated precipitation shows that the temporal evolution of precipitation in 50P is virtually identical to CTL. However, the precipitation amount in 50P is less than CTL beyond 24 h (Fig. 13a). The reduction in the area of precipitation beyond 24 h is closely linked to the evolution of the upward motion. A plot of the 500-hPa vertical motion in p coordinates reveals that beyond 24 h, the strength of upward motion in 50P is less than CTL (Fig. 13b). A similar conclusion is reached by employing area averages performed for other PR48 thresholds. Thus, a weaker upward motion in 50P contributed to the lower BS.

On the other hand, a significant increase in CAPE occurs over P2 and some regions of P3 in 50P relative to CTL at model initial time (not shown). Elsewhere, over P1 and P4, there is no change in CAPE between 50P and CTL. Between 0000 and 0600 UTC 20 July, the surface cyclone center is displaced 40–60 km south of the CTL (not shown). At other times, the cyclone center is located within one grid point of the CTL. Milbrandt and Yau (2001) showed that the vertical motion resulting from the interaction between topography and the low-level flow associated with the cyclone can contribute up to about 25% of PR48. Thus, a displacement of the cyclone center and the associated horizontal flow field leads to a reduction in the strength of the upward motion. The weaker upward motion results in a smaller area of precipitation and a lower BS.

b. Initial conditions

Recall that the CMC operational analyses was specified as the background field (section 3a) and the temperature and moisture increments given by the 1DVAR scheme are used to enhance the NCEP reanalysis (section 3c). Thus, the total increment of temperature and moisture (i.e., 1DVAR minus background) is composed of contributions from the 1DVAR scheme and the CMC profiles. In other words, the CTL initial condition contains corrections from the 1DVAR scheme and those arising from replacing the NCEP reanalysis with CMC profiles. A sensitivity experiment is performed to study the impact of the 1DVAR scheme on the QPF by setting the 1DVAR increments to zero. The results show that the TS is less than CTL for PR48 threshold greater than 135 mm and is zero for PR48 threshold greater than 180 mm (not shown). This suggests that temperature and moisture increments from the 1DVAR scheme have significant impact in the prediction of high rainfall values.

As a result of the improved initial conditions by the assimilation of precipitation over regions P1, P2, P3, and P4 (see Fig. 3b), the TS in CTL for large precipitation amount is significantly higher than in NCEP (section 4a). We examine the impact of assimilating rainfall associated with individual precipitation systems by performing four sensitivity experiments, referred to as No-P1, No-P2, No-P3, and No-P4, respectively. The No-P1 experiment excludes temperature and moisture profiles obtained from 1DVAR scheme over P1 in the model initial conditions. The other experiments are defined similarly.

Figure 14a shows the TS from CTL and the four sensitivity experiments. The TS in the No-P1 experiment is not significantly different from CTL. On the other hand, the TS over the entire PR48 threshold range (except between 170 and 180 mm) in the No-P4 experiment is smaller than CTL. The total precipitation in No-P4 is largely made up of contributions from grid-resolved precipitation (Figs. 15a,b). Grid-resolvable precipitation is favored in the presence of saturated ascent over a deep tropospheric column (Molinari and Dudek 1992). Since the assimilation of precipitation over P4 at model initial time moistens the midtroposphere and renders it favorable for the formation of grid-scale precipitation (Fig. 5), there is a significant loss of skill in predicting precipitation in the No-P4 experiment. The TS in the No-P2 experiment is similar to CTL with lower values occurring between 150 and 215 mm. Since the BS between 150 and 200 mm is similar in No-P2 and CTL, it suggests that the lower TS in No-P2 arises from the placement of heavy precipitation regions east of the observed (Fig. 15c). This follows from the fact that the BS reflects the under- or overprediction of the area of precipitation compared to the observed whereas the TS represents the degree of overlap between the predicted and observed area of precipitation. The TS in the No-P3 experiment is similar to CTL between 110 and 190 mm. However the TS is greater than CTL between 75 and 105 mm and is accompanied by a BS that is lower than that in CTL. This behavior is also exhibited in PR48 from NCEP (Figs. 6d and 14a), suggesting that assimilation of precipitation over P3 in CTL adversely impacts the TS and BS between 75 and 105 mm.

Hence, the assimilation of precipitation in CTL over P4 and P2 impact the midtropospheric moisture and the placement of the precipitation, respectively. Since the placement of the precipitation and the surface cyclone track is interrelated, our results suggest that the displacement of the NCEP-simulated surface cyclone tracks (Fig. 6c) east of CTL is due largely to the initial temperature and moisture profiles over P2.

When compared to observations, all the sensitivity experiments overpredict precipitation between 50 and 185 mm. The No-P1 and No-P3 experiments significantly overpredict the area of precipitation relative to CTL for PR48 between 160 and 225 mm. In No-P1, the ABL is significantly more humid than CTL (Fig. 4b). Although in No-P3, the Atlantic Seaboard is drier relative to CTL (Fig. 4b), much of the other area over P3 is significantly moister than CTL. This suggests that low-level moisture content over the ABL largely leads to an overprediction of the precipitation in No-P1 and No-P3, and may partly explain the overprediction of precipitation in the NCEP run for PR48 between 195 and 225 mm (Fig. 6d). Therefore, assimilation of precipitation over P2 and P4 favorably impacts the TS while the assimilation over P1 and P3 positively influences the BS.

7. Summary and conclusions

In this study a one-dimensional variational assimilation technique is used to improve model initial conditions and extend the predictability of the Saguenay flood cyclone that occurred between 19 and 21 July 1996.

The operational Canadian Meteorological Center (CMC) forecasts initialized at 0000 UTC 19 July 1996 poorly predicted the location and intensity of the precipitation. This was attributed to errors in the initial conditions arising from the lack of mesoscale resolution data to characterize the mesoscale precipitation systems. Four distinct precipitation systems were identified over regions P1–P4 (Fig. 3b) at the model initial time (0000 UTC 19 July).

The initial temperature and moisture fields were improved using the 1-DVAR scheme. The 24-km horizontal resolution CMC operational analyses provided the background field. A Gaussian autocorrelation model was used to specify the background error for both temperature and specific humidity and the standard deviation error of the observed precipitation was set to 25% of the background field. The NCEP reanalysis was enhanced with temperature and specific humidity profiles from the 1DVAR scheme at the model initial time (0000 UTC 19 July) by employing the Barnes objective analysis scheme. The assimilation of precipitation resulted in a drier ABL over much of the area occupied by the precipitation systems with the exception of a moist ABL along the Atlantic Seaboard. Elsewhere (over P4) there was an increase in moisture between 900 and 500 hPa.

Two 60-h-long numerical experiments initialized at 0000 UTC 19 July were performed at high horizontal resolution (20 km). The control (CTL) and the NCEP simulations were initialized with the enhanced and the NCEP reanalysis data, respectively.

The CTL simulation qualitatively reproduced the location and intensity of the 48-h accumulated precipitation. On the other hand, the PR48 simulated by the NCEP run exhibited only two precipitation maxima and they were situated east of those observed. The threat score of CTL was greater than NCEP except for thresholds between 75 and 105 mm. Thus, the CTL simulation showed significant improvement in the predictive skill for large rainfall values. However, both CTL and NCEP significantly overpredicted the area of precipitation relative to the observed. This led to a deeper-than-observed cyclone being simulated in CTL. The CTL run exhibited a reduced model spinup period and captured the temporal evolution of precipitation. Below the tropopause, the simulated moisture, temperature, and wind profiles at 60 h from CTL compared reasonably with the observations.

One of the less understood parameters in the 1DVAR scheme is the standard deviation error of the observed precipitation data. In the CTL simulation, it was assumed to be 25% of the background field. Two sensitivity experiments were performed with the standard deviation error set to 5% and 50%. No significant change in TS occurred but the BS decreased for PR48 between 120 and 200 mm for a standard deviation error of 50%. The decrease in BS was attributed to a weaker upward motion resulting from the interaction between the low-level flow associated with the cyclone and the topography that yielded a reduced grid-resolved precipitation. Thus, the TS (BS) showed no significant (some) sensitivity to the variation of the observed precipitation error statistics.

We performed sensitivity experiments to study the impact of the individual precipitation systems on the TS and BS of the CTL run by removing the associated temperature and moisture profiles obtained from the 1DVAR scheme. While no significant impact on the TS occurred due to the assimilation of precipitation over P1 and P3 (except between 75 and 105 mm), the assimilation of precipitation over P2 and P4 significantly affected the TS. The lack of assimilation of precipitation data over P2 may explain the poor skill seen in the NCEP run. The improved skill for heavy precipitation arose due to the correct placement of the surface precipitation. The assimilation of precipitation over P4 increased the moisture between 900 and 500 hPa and favored formation of grid-resolved precipitation. On the other hand, the decrease in TS between 75 and 105 mm observed in CTL is caused by the assimilation of precipitation over P3 leading to a humid ABL. All the sensitivity experiments overpredicted the area of precipitation as a result of a very humid ABL.

A PV inversion technique was applied to study the time evolution of the cyclone track. Two distinct positive upper-level PV anomalies were associated with the northern and southern trough. The upper-level PV anomaly associated with the northern trough influenced the cyclone track. However, its propagation was influenced by the wind field associated with the upper-level PV anomaly associated with the southern trough.

The southern trough was strongly influenced by condensational heating over P2. The weaker zonal wind component associated with the upper-level PV anomaly of the southern trough in NCEP accelerated the eastward propagation of the northern trough and resulted in an incorrect placement of the surface cyclone and precipitation.

In conclusion, we showed that the specification of the mesoscale moisture and temperature field in the vicinity of the southern trough (i.e., over P2 and P4) using the 1DVAR scheme yielded an improved QPF. Due to the short-lived nature of the southern trough the 1-DVAR scheme was only applied at the model initial time. Bosart and Sanders (1991) studied the climatology of trough merger events and found that two-thirds of such merger events were associated with explosive cyclogenesis. Milbrandt and Yau (2001) and Huo et al. (1999) studied trough merger events and concluded that the southern trough only impacted the initial evolution of the surface cyclone. This suggests that an improvement in the QPF of precipitation may be expected in other trough-merger events if the initial temperature and moisture fields in the vicinity of the southern trough are improved. Although only a single case is discussed in this paper, the 1DVAR scheme is presently being applied to other trough-merger events, the results of which will be reported in the future. Since precipitation constitutes an important input to a hydrological model used in flood forecasting, an improved QPF is likely to result in an accurate flood forecast (Lin et al. 2002; Benoit et al. 2003).

Acknowledgments

We wish to thank the Forecast Systems Laboratory (FSL) for providing us with the rain gauge data over the United States and Dr. N. Donaldson and Ms. R. Diaconesco of Environment Canada for providing the rain gauge data over Ontario and Quebec. The U.S. composite radar reflectivity data were provided by the Global Hydrology Resource Center (GHRC) at the Global Hydrology and Climate Center, Hunstville, Alabama. The AVHRR satellite data were obtained from the NOAA Satellite Active Archive (SAA) System. This research was supported by the Canadian Foundation for Climate and Atmospheric Science.

REFERENCES

  • Anthes, R. A., 1983: Regional model of the atmosphere in middle latitudes. Mon. Wea. Rev., 111 , 13061335.

  • Barnes, S. L., 1964: A technique for maximizing details in numerical weather map analysis. J. Appl. Meteor., 3 , 396409.

  • Benoit, R., , J. Cote, , and J. Mailhot, 1989: Inclusion of a TKE boundary layer parameterization in the Canadian regional finite-element model. Mon. Wea. Rev., 117 , 17261750.

    • Search Google Scholar
    • Export Citation
  • Benoit, R., , M. Desgagne, , P. Pellerin, , S. Pellerin, , Y. Chartier, , and S. Desjardins, 1997: The Canadian MC2: A semi-implicit semi-Lagrangian wide-band atmospheric model suited for fine-scale process studies and simulation. Mon. Wea. Rev., 125 , 23822415.

    • Search Google Scholar
    • Export Citation
  • Benoit, R., , N. Kouwen, , W. Yu, , S. Chamberland, , and P. Pellerin, 2003: Hydrometeorological aspects of the real-time ultrafinescale forecast support during the special observing period of the MAP. Hydrol. Earth Syst. Sci., 7 , 877889.

    • Search Google Scholar
    • Export Citation
  • Bosart, L. F., , and F. Sanders, 1991: An early-season coastal storm: Conceptual success and model failure. Mon. Wea. Rev., 119 , 28312851.

    • Search Google Scholar
    • Export Citation
  • Davies, C. A., 1992: A potential vorticity diagnosis of the importance of the initial structure and condensational heating in observed extratropical cyclogenesis. Mon. Wea. Rev., 120 , 24092428.

    • Search Google Scholar
    • Export Citation
  • Davies, C. A., , and K. A. Emanuel, 1991: Potential vorticity diagnostics of cyclogenesis. Mon. Wea. Rev., 119 , 19291953.

  • Davolio, S., , and A. Buzzi, 2004: A nudging scheme for the assimilation of precipitation data into a mesoscale model. Wea. Forecasting, 19 , 855871.

    • Search Google Scholar
    • Export Citation
  • Ertel, H., 1942: Ein nueur hydrodynanischer wirbelsatz. Meteor. Z., 59 , 277281.

  • Fillion, L., , and R. Errico, 1997: Variational assimilation of precipitation data using moist convective parameterization schemes. Mon. Wea. Rev., 125 , 29172942.

    • Search Google Scholar
    • Export Citation
  • Fillion, L., , and S. Belair, 2004: Tangent linear aspects of the Kain–Fritsch moist convective parameterization scheme. Mon. Wea. Rev., 132 , 24772494.

    • Search Google Scholar
    • Export Citation
  • Fritsch, J. M., , and C. F. Chappell, 1980: Numerical prediction of convectively driven mesoscale pressure systems. Part I: Convective parameterization. J. Atmos. Sci., 37 , 17221733.

    • Search Google Scholar
    • Export Citation
  • Fritsch, J. M., and Coauthors, 1998: Quantitative precipitation forecasting: Report of the eighth prospectus development team, U.S. Weather Research Program. Bull. Amer. Meteor. Soc., 79 , 285299.

    • Search Google Scholar
    • Export Citation
  • Garand, L., 1983: Some improvements and complements to the infrared emissivity algorithm including a parameterization of the absorption in the continuum region. J. Atmos. Sci., 40 , 230244.

    • Search Google Scholar
    • Export Citation
  • Garand, L., , and C. Grassotti, 1995: Toward an objective analysis of rainfall rate combining observations and short-term forecast model estimates. J. Appl. Meteor., 34 , 19621977.

    • Search Google Scholar
    • Export Citation
  • Grassotti, C., , R. N. Hoffman, , E. R. Vivoni, , and D. Entekhabi, 2003: Multiple-timescale intercomparison of two radar products and rain gauge observations over the Arkansas–Red River basin. Wea. Forecasting, 18 , 12071229.

    • Search Google Scholar
    • Export Citation
  • Guo, Y-R., , Y-H. Kuo, , J. Dudhia, , D. Parsons, , and C. Rocken, 2000: Four-dimensional variational data assimilation of heterogeneous mesoscale observations for a strong convective case. Mon. Wea. Rev., 128 , 619643.

    • Search Google Scholar
    • Export Citation
  • Hamill, T. S., 1999: Hypothesis tests for evaluating numerical precipitation forecasts. Wea. Forecasting, 14 , 155167.

  • Hammer, G. R., , and P. M. Steurer, 1997: Data set documentation for hourly precipitation data. NOAA/NCDC Tech. Doc. 3240, Documentation Series, Asheville, NC, 1–18.

  • Hoskins, B. J., , M. E. McIntyre, , and R. W. Robertson, 1985: On the use and significance of isentropic potential vorticity maps. Quart. J. Roy. Meteor. Soc., 111 , 877946.

    • Search Google Scholar
    • Export Citation
  • Huo, Z., , D-L. Zhang, , and J. Gyakum, 1998: An application of potential vorticity inversion to improving the numerical prediction of the March 1993 superstorm. Mon. Wea. Rev., 126 , 424436.

    • Search Google Scholar
    • Export Citation
  • Huo, Z., , D-L. Zhang, , and J. Gyakum, 1999: Interaction of potential vorticity anomalies in extratropical cyclogenesis. Part I: Static piecewise inversion. Mon. Wea. Rev., 127 , 25462561.

    • Search Google Scholar
    • Export Citation
  • Kain, J. S., , and J. M. Fritsch, 1990: A one-dimensional entraining/detraining plume model and its application in convective parameterization. J. Atmos. Sci., 47 , 27842802.

    • Search Google Scholar
    • Export Citation
  • Kasahara, A., , A. P. Mizzi, , and L. J. Donner, 1992: Impact of cumulus initialization on the spinup of precipitation forecasts in the Tropics. Mon. Wea. Rev., 120 , 13601380.

    • Search Google Scholar
    • Export Citation
  • Kong, F. Y., , and M. K. Yau, 1997: An explicit approach to microphysics in MC2. Atmos.–Ocean, 35 , 257291.

  • Krishnamurti, T. N., , H. S. Bedi, , and K. Ingles, 1993: Physical initialization using the SSM/I rain rate. Tellus, 45A , 247269.

  • Kuo, H. L., 1974: Further studies of the parameterization of the influence of cumulus convection on large scale flow. J. Atmos. Sci., 31 , 12321240.

    • Search Google Scholar
    • Export Citation
  • Lin, C. A., , L. Wen, , M. Beland, , and D. Chaumont, 2002: A coupled atmospheric-hydrological modeling study of the 1996 Ha! Ha! River basin flash flood in Quebec, Canada. Geophys. Res. Lett., 29 .1026, doi:10.1029/2001GL013827.

    • Search Google Scholar
    • Export Citation
  • Lin, Y., , and K. E. Mitchell, 2005: The NCEP stage II/IV hourly precipitation analyses. Preprints, 19th Conf. on Hydrology, San Diego, CA, Amer. Meteor. Soc., 72–73.

  • Marecal, V., , and J-F. Mahfouf, 2002: Four-dimensional variational assimilation of total column water vapor in rainy areas. Mon. Wea. Rev., 130 , 4358.

    • Search Google Scholar
    • Export Citation
  • Mekis, E., , and W. D. Hogg, 1999: Rehabilitation and analysis of Canadian daily precipitation time series. Atmos.–Ocean, 37 , 5385.

  • Milbrandt, J. A., , and M. K. Yau, 2001: A mesoscale modeling study of the 1996 Saguenay flood. Mon. Wea. Rev., 129 , 14191440.

  • Molinari, J., , and M. Dudek, 1992: Parameterization of convective precipitation in mesoscale numerical models: A critical review. Mon. Wea. Rev., 120 , 326344.

    • Search Google Scholar
    • Export Citation
  • Nagarajan, B., , M. K. Yau, , and D-L. Zhang, 2001: A numerical study of a mesoscale convective system during TOGA COARE. Part I: Model description and verification. Mon. Wea. Rev., 129 , 25012520.

    • Search Google Scholar
    • Export Citation
  • Olson, D. A., , N. W. Junker, , and B. Korty, 1995: Evaluation of 33 years of quantitative precipitation forecasting at the NMC. Wea. Forecasting, 10 , 498511.

    • Search Google Scholar
    • Export Citation
  • Rossby, C. G., 1940: Planetary flow patterns in the atmosphere. Quart. J. Roy. Meteor. Soc., 66 , 6887.

  • Stensrud, D. J., , and J. M. Fritsch, 1994: Mesoscale convective systems in weakly forced large-scale environments. Part II: Generation of a mesoscale initial condition. Mon. Wea. Rev., 122 , 20682083.

    • Search Google Scholar
    • Export Citation
  • Treadon, R. E., , H-L. Pan, , W-S. Wu, , Y. Lin, , W. S. Olson, , and R. J. Kuligowski, 2003: Global and regional moisture analyses at NCEP. Proc. ECMWF/GEWEX Workshop on Humidity Analysis, Reading, United Kingdom, ECMWF, 33–48.

  • Tsuyuki, T., , K. Koizumi, , and Y. Ishikawa, 2003: The JMA mesoscale 4D-Var system and assimilation of precipitation and moisture data. Proc. ECMWF/GEWEX Workshop on Humidity Analysis, Reading, United Kingdom, ECMWF, 59–68.

  • Verret, R., , L. Lefaivre, , J-G. Desmarais, , and T. Robinson, 1996: Intense precipitation in Quebec, July 19–20, 1996. Can. Meteor. Center Rev., 3 , 129.

    • Search Google Scholar
    • Export Citation
  • Zhang, D-L., , and J. M. Fritsch, 1986: Numerical simulation of the meso-β-scale structure and evolution of the 1977 Johnstown flood. Part I: Model description and verification. J. Atmos. Sci., 43 , 19131943.

    • Search Google Scholar
    • Export Citation
  • Zou, X., , and Y-H. Kuo, 1996: Rainfall assimilation through an optimal control of initial and boundary conditions in a limited-area mesoscale model. Mon. Wea. Rev., 124 , 28592882.

    • Search Google Scholar
    • Export Citation
Fig. 1.
Fig. 1.

The NCEP reanalyses of 850-hPa temperature (dashed contours every 2°C) superposed with SLP (solid contours every 4 hPa) for (a) 0000 UTC 19 Jul 1996 and (c) 1200 UTC 20 Jul 1996. (b), (d) The corresponding 500-hPa geopotential height (solid contours every 6 decameters) superposed with absolute vorticity (dashed contours every 2 × 10−5 s−1). Subjectively drawn front and trough lines are shown in (a), (c), and (b), respectively. The AVHRR satellite brightness temperature (shaded) with U.S. radar reflectivity composite (solid contours every 10 dBZ) are shown for (e) 0000 UTC and (f) 1200 UTC 19 Jul 1996. Blank regions in (e) and (f) depict regions with missing satellite data.

Citation: Monthly Weather Review 134, 5; 10.1175/MWR3128.1

Fig. 2.
Fig. 2.

The 48-h objectively analyzed rain gauge precipitation data for the period between 1300 UTC 19 Jul and 1300 UTC 21 Jul (shaded and contoured) is shown along with central SLP tracks from NCEP reanalysis (thin solid), the CTL (dashed–dotted), and NCEP (thick solid) simulation data. The three precipitation maxima are shown in mm. The central SLP locations are depicted every 6 h between 0000 UTC 19 Jul and and 1200 UTC 21 Jul. Maniwaki, Caribou, and Sept-Iles are the upper-air radiosonde stations. R1, R2, and R3 depict the location of three rain gauges whose measurements are used in evaluating the temporal variation of precipitation.

Citation: Monthly Weather Review 134, 5; 10.1175/MWR3128.1

Fig. 3.
Fig. 3.

(a) The coarse (A) and fine mesh domain (B) is shown with spatial distribution of the rain gauges. (b) The objectively analyzed rain gauge data show the precipitation rate (mm h−1) computed from the hourly rainfall accumulation ending 0000 UTC 19 Jul 1996. The 0.1, 1.0, 5.0, 10.0, 25.0, and 50.0 mm h−1 contours are plotted with the shaded area depicting regions with a rain rate greater than 0.1 mm h−1. The boxes P1, P2, P3, and P4 in (b) represent the areas occupied by the four distinct precipitation systems.

Citation: Monthly Weather Review 134, 5; 10.1175/MWR3128.1

Fig. 4.
Fig. 4.

(a) The difference in CAPE in J kg−1 between the CTL and NCEP runs (negative dashed and positive shaded) is shown at 0000 UTC 19 Jul 1996 over the regions P1–P4. (b) The shaded area (solid line) shows the corresponding difference moisture (temperature) in g kg−1 (°C). The 0.5° and 1.0°C temperature contours are shown. The CAPE is computed by mixing the lowest 125-hPa air layer and the moisture and temperature difference field represents a layer average over the lowest 125 hPa.

Citation: Monthly Weather Review 134, 5; 10.1175/MWR3128.1

Fig. 5.
Fig. 5.

A skew T–logp plot of the area-averaged temperature and dewpoint profiles over P4 (Fig. 3b) at 0000 UTC 19 Jul 1996. Thick solid (dotted) lines indicate temperature and dewpoint from CTL (NCEP).

Citation: Monthly Weather Review 134, 5; 10.1175/MWR3128.1

Fig. 6.
Fig. 6.

The spatial distribution of the 48-h precipitation accumulation (mm) ending 1200 UTC 21 Jul 1996 is shown for (a) CTL and (b) NCEP. (c) The surface cyclone location between 1800 UTC 19 Jul and 1200 UTC 21 Jul from CTL (C) and NCEP (N) is plotted every 3 h. (d) The TS (with bias equalized to CTL) and BS (unequalized) is solid (dashed) for the CTL (NCEP).

Citation: Monthly Weather Review 134, 5; 10.1175/MWR3128.1

Fig. 7.
Fig. 7.

A time series of the total accumulated precipitation (mm), plotted every 3 h, from rain gauge (OBS), CTL, and NCEP is shown for (a) R1, (b) R2, and (c) R3. The rain gauge locations R1, R2, and R3 are shown in Fig. 2. Rain gauge observations are absent before 9 h.

Citation: Monthly Weather Review 134, 5; 10.1175/MWR3128.1

Fig. 8.
Fig. 8.

A skew T–logp plot of the observed (solid) and model sounding (dashed) from CTL at 1200 UTC 21 Jul 1996 for (a) Caribou, (b) Sept-Iles, and (c) Maniwaki (see Fig. 2). The observed wind is plotted at the rightmost end of (a)–(c). A full barb is 5 m s−1. (d) The central SLP (hPa) from CTL, NCEP, and NCEP reanalysis (NCEP-REANALYSIS) is shown as a function of time.

Citation: Monthly Weather Review 134, 5; 10.1175/MWR3128.1

Fig. 9.
Fig. 9.

The 12-h accumulated precipitation (mm) ending at 1200 UTC 19 Jul superposed with the 500-hPa geopotential height (decameters) from (a) CTL and (b) NCEP. The 1200 UTC 19 Jul northern and southern troughs at 500 hPa are shown by the thick dotted line.

Citation: Monthly Weather Review 134, 5; 10.1175/MWR3128.1

Fig. 10.
Fig. 10.

The CTL-minus-NCEP layer-averaged potential vorticity anomaly (PVU) (a) Qm and (b) Qd at 1200 UTC 19 Jul. The Qd (Qm) is averaged between 500 and 200 hPa (900 and 500 hPa). The contours −0.2, −0.1, 0.1, and 0.2 (−0.5, −0.1, 0.1, and 0.5) PVU are shown for Qd (Qm). Shaded regions indicate values greater (less) than 0.1 (−0.1) PVU for Qd (Qm), where 1 PVU = 10−6 m2 K s−1 kg−1.

Citation: Monthly Weather Review 134, 5; 10.1175/MWR3128.1

Fig. 11.
Fig. 11.

The geopotential height (decameters) superposed with the CTL-minus-NCEP zonal wind component (m s−1) at 300 hPa associated with the southern trough (Figs. 9a,b) PV anomaly Qd at 1200 UTC 19 Jul. The northern and southern troughs are shown by the thick solid lines. Although the southern trough is not evident, it is subjectively depicted based on its location at 500 hPa (Fig. 9a).

Citation: Monthly Weather Review 134, 5; 10.1175/MWR3128.1

Fig. 12.
Fig. 12.

The PR48 TS (with bias equalized to CTL) and BS (unequalized) from CTL and two sensitivity experiments performed by setting the standard deviation error for observed precipitation rate to 5% (OBS ERR 5%) and 50% (OBS ERR 50%) of the background field. The standard deviation error in CTL is assumed to be 25%.

Citation: Monthly Weather Review 134, 5; 10.1175/MWR3128.1

Fig. 13.
Fig. 13.

A time series of area-averaged accumulated precipitation (mm) and the 500-hPa vertical motion in p coordinates (Pa s−1) from (a) CTL and (b) 50P are shown every 3 h. The area average is performed over the region enclosed by the 150-mm precipitation contour in Fig. 6a.

Citation: Monthly Weather Review 134, 5; 10.1175/MWR3128.1

Fig. 14.
Fig. 14.

The PR48 TS (with bias equalized to CTL) and BS (unequalized) from CTL and four sensitivity experiments: No-P1, No-P2, No-P3, and No-P4. The sensitivity experiment No-P1 refers to the absence of the temperature and moisture profiles used in CTL obtained through the 1DVAR scheme over P1. The other sensitivity experiments are defined similarly.

Citation: Monthly Weather Review 134, 5; 10.1175/MWR3128.1

Fig. 15.
Fig. 15.

(a) The 48-h accumulated convective and (b) grid-resolved precipitation (mm) ending 1200 UTC 21 Jul 1996 from experiment No-P4. (c) The 48-h accumulated precipitation (convective and grid resolved) from No-P2.

Citation: Monthly Weather Review 134, 5; 10.1175/MWR3128.1

Save