1. Introduction
There has been a rapid advance in the understanding of orographic precipitation during the past several years with the application of high-resolution models and new observational datasets. However, accurate precipitation forecasts in areas of steep terrain remain a challenge given the complex dynamical, thermodynamical, and cloud microphysical processes in these regions. As a result, several field studies have been completed to better understand orographic precipitation as well as to verify and improve bulk microphysical parameterizations (BMPs), such as the Improvement of Microphysical Parameterization through Observational Verification Experiment (IMPROVE) project in 2001 over the Oregon Cascade Range (Stoelinga et al. 2003), the Intermountain Precipitation Experiment (IPEX) over the Wasatch Mountains in 2000 (Schultz et al. 2002), and the Mesoscale Alpine Program (MAP) over the Alps in 1999 (Bougeault et al. 2001).
These field data and other studies have identified numerous deficiencies in BMPs (Manning and Davis 1997; Colle and Mass 2000; Garvert et al. 2005b; Colle et al. 2005a). For example, using the Thompson et al. (2004) scheme within the fifth-generation Pennsylvania State University–National Center for Atmospheric Research (NCAR) Mesoscale Model (MM5) resulted in snow overprediction aloft in the 13–14 December 2001 IMPROVE-2 event over the central Oregon Cascades (Garvert et al. 2005b). This resulted in excessive spillover of snow over the crest into the lee of the Cascades during a period of strong low-level cross-barrier flow (Garvert et al. 2005a,b). The model snow overprediction aloft was not reduced using a triple-moment BMP for the same IMPROVE event (Milbrandt et al. 2008). In contrast, for the IPEX third intensive observing period (IOP3) over the Wasatch Mountains of Utah (Colle et al. 2005b), the Thompson et al. (2004) scheme within MM5 at 1.33-km grid spacing produced too little snow aloft over the barrier.
The various BMPs available in mesoscale models generate large differences in the simulated hydrometeor and surface precipitation distributions over steep terrain (Colle et al. 2005a,b; Grubisić et al. 2005; Richard et al. 2007). For example, Richard et al. (2007) found that the Thompson et al. (2004) scheme underpredicted the surface precipitation for MAP IOP2b using the MM5, while there was overprediction in the Goddard BMP scheme (Tao and Simpson 1993). They also found that the Weather Research and Forecasting (WRF) Single-Moment 6-Class Microphysics Scheme (WSM6; Hong et al. 2004) predicted ∼10 mm more precipitation than the Thompson scheme (Thompson et al. 2004) using the WRF model (Skamarock et al. 2005).
Although there have been a growing number of studies investigating the BMP sensitivities for orographic precipitation, only a few studies have used in situ aircraft data to evaluate the different schemes (Garvert et al. 2005b; Colle et al. 2005a,b). This paper focuses on the 4–5 December 2001 IMPROVE-2 event (Colle et al. 2008), which featured ∼50% weaker cross-barrier flow, a lower freezing level (∼1 versus 2 km), slightly more instability at low levels, and one-half less surface precipitation than the 13–14 December 2001 case. As a result, this case provides unique observations and model validation for an event with less snow growth and riming over the Cascades than the 13–14 December event. Furthermore, there have been relatively few verification studies of the WRF BMPs (Hong et al. 2004; Hong and Lim 2006; Thompson et al. 2004, 2008; Chen and Sun 2002), and no formal comparisons of WRF BMPs using in situ aircraft observations over steep terrain.
Colle et al. (2008) focused on the kinematic and precipitation evolution for the 4–5 December 2001 case over Oregon Coast Range and Cascades. The objective of this paper is to describe the observed microphysical evolution for the same event as well as to evaluate and understand the differences in performance of the WRF BMPs. These results are compared with the well-documented 13–14 December 2001 IMPROVE-2 case. The motivational questions for this study are the following:
How do the observed microphysical structures compare between the 13–14 and the 4–5 December 2001 IMPROVE-2 events given the different ambient conditions?
How well do the BMP schemes within WRF simulate the precipitation and microphysical quantities for the 4–5 December IOP? Are the results consistent with the previously investigated 13–14 December IMPROVE-2 event?
What are the BMP sensitivities for the 4–5 December event and some reasons for the differences between the BMPs?
The paper is organized as follows. The data and methods are given in section 2 as well as an overview of the four microphysical schemes. Observed microphysical characteristics of the event are presented in section 3. Model comparisons with observations are presented in section 4. The sensitivity tests and discussion of the potential deficiencies in the microphysical schemes are described in section 5. The summary and conclusions are given in section 6.
2. Data and methods
a. Microphysical datasets and retrieval
Two research aircraft, the National Oceanic and Atmospheric Administration (NOAA) P-3 and the UW Convair-580, collected radar and microphysical data over the Cascades during the 4–5 December event (Fig. 1a). The Convair-580 was equipped with a Particle Measuring System (PMS) 2-DC and Forward Scattering Spectrometer Probe (FSSP). The P-3 was equipped with a tail Doppler radar and PMS 2D-C and 2D-P grayscale imaging probes to obtain ice particle concentrations as well as King (King et al. 1978) and Johnson-Williams (J-W) probes (Baumgardner 1983) to measure cloud liquid water (CLW) mass. The CLW difference between the King and J-W probes was less than 10%, thus the average value was used. To quantify the aircraft ice mass concentration uncertainty, several different mass diameter (m–D) relationships were used. First, as in Garvert et al. (2005b) and Woods et al. (2005), the ice mass concentrations were calculated using in situ observations of the number concentrations, classified ice crystal types, and the corresponding m–D relationships for each type [refer to Table 4 of Woods et al. (2005) for the specific relationships used]. This approach also provides an estimate of the graupel (heavily rimed snow) mass concentration. Three other estimates of snow mass were obtained using Brown and Francis (1995), as well as Heymsfield et al. (2002) with an area ratio of 0.7 and 0.5 (later referred to as BF, HEYMS, and AGGR, respectively). The BF gives the appropriate estimate for aggregates of unrimed bullets, columns, and side planes, while AGGR and HEYMS give the appropriate estimate for aggregates and aggregate/graupel hybrid, respectively. The ice mass estimates from the three methods vary by 20%–40% depending on the flight altitude. Since most of the ice particles observed along the flight tracks are aggregates of different habits with little riming, BF was used for the direct comparison with the model.
A radiometer at Santiam Junction (SJ in Fig. 1a) was used to obtain the integrated cloud liquid water depth (LWD) and water vapor depth as described in Woods et al. (2005). Together with the calculated integrated water vapor from the sounding at UW and BB (UW and BB in Fig. 1a), this provided a good estimation of the observed moisture evolution.
The NCAR S-band Doppler radar (S-Pol), located on the western foothills of the Oregon Cascades (SPOL in Fig. 1b) provided the plan position indicator (PPI) and range–height indicator (RHI) scans over the Cascades and the Coast Range. The dual-Doppler radar on the NOAA P-3 sampled the winds and reflectivities over the windward slopes and immediate lee of the Cascades during the IOP (Colle et al. 2008).
Hourly precipitation totals from the National Weather Service (NWS) cooperative sites, National Resources Conservation Service (NRCS) Snowpack Telemetry (SNOTEL) sites, and the special precipitation gauges deployed over the Oregon Cascades during IMPROVE (Stoelinga et al. 2003) were synthesized to verify the simulated precipitation amounts.
b. Model setup and configuration
The WRF, version 2.2, was utilized using the setup described in Colle et al. (2008). The National Centers for Environmental Prediction Aviation Model (NCEPAVN) analyses at 1° resolution were used for the initial (starting at 1200 UTC 4 December 2001) and boundary conditions at 36-, 12-, 4-, and 1.33-km grid spacing (see Fig. 1a of Colle et al. 2008). The 36- and 12-km domain applied the Kain and Fritsch (1993) convective parameterization, while no cumulus parameterization was used in the 4- and 1.33-km domains. A positive-definite advection (PDA) scheme for moisture and hydrometeors was used in order to prevent artificial moisture gain during the advective time steps (Skamarock 2006), which will be shown to be important in section 5.
Four WRF BMPs (Table 1) were evaluated in this study at 1.33-km grid spacing. The WRF single-moment microphysics with graupel (WSM6) is based on modifications from Lin et al. (1983) and Rutledge and Hobbs (1983; Hong et al. 2004; Hong and Lim 2006). This scheme includes separate formula for the ice crystal and ice nuclei number concentrations; a self-consistent mass, diameter, and number concentration relationships for ice; temperature-dependent snow intercept; and a maximum ice crystal diameter of 500 μm. The Purdue Lin (LIN) scheme (Chen and Sun 2002) follows the parameterization of Lin et al. (1983) and Rutledge and Hobbs (1984), but with the following modifications. LIN uses a saturation adjustment method from Tao et al. (1989) and a simple Kessler type (Kessler 1969) autoconversion, with a threshold cloud liquid water mixing ratio of 0.7 g kg−1. Snow deposition and riming are implemented as in Koenig (1971). It also uses a Kessler-type autoconversion for snow to graupel with a threshold snow of 0.6 g kg−1.
The Thompson scheme (THOM1) originated from the Reisner2 scheme in MM5 (Reisner et al. 1998), with modifications by Thompson et al. (2004). Some modifications include temperature-dependent snow intercept as in Houze et al. (1979), a temperature-dependent ice number concentration as in Cooper (1986), a Gamma distribution for graupel with a graupel mass–dependent intercept, a varying rain intercept to indirectly simulate drizzle drops, a more physical autoconversion formula as in Walko et al. (1995), and a snow riming and depositional growth ratio to determine the riming snow transition to graupel.
A new Thompson scheme was developed for WRFV2.2 (THOM2; Thompson et al. 2008). Some modifications include sums of two gamma functions for snow size distribution based on aircraft observations by Field et al. (2005), a gamma size distribution for cloud water with an aerosol-dependent dispersion factor, an exponential size distribution for cloud ice, varying rain intercept considering the snow-melted rain, varying snow density, and collision and collection processes parameterized using a stochastic collision equation.
3. Observed microphysical characteristics
As highlighted in Colle et al. (2008), the 4–5 December 2001 event featured a landfalling baroclinic trough over the Pacific Northwest. A detailed description of the kinematic and precipitation evolution of the case is presented in Colle et al. (2008); thus, this section focuses on the in situ precipitation and microphysical measurements by the S-Pol radar, radiometer, NOAA P-3, and Convair aircrafts.
As illustrated in Colle et al. (2008), there were numerous convective cells over the Coast Range and Cascades during this event. The precipitation structures observed over the Cascades during the airborne microphysical measurements are illustrated using a composite of reflectivities and winds at 3 km above mean sea level (MSL) from the NOAA P-3 along north–south legs 2–4 between 2351 UTC 4 December and 0051 UTC 5 December (Fig. 1a). There are some precipitation enhancements over the southwest-facing ridges, while other areas of enhanced precipitation originated over the Coast Range and propagated east-northeastward with the 15–20 m s−1 west-southwest flow at this level (cf. Fig. 11 of Colle et al. 2008). Figure 1b suggests that the 1.33-km WRF was able to realistically simulate the enhanced precipitation areas over the windward slopes and crest of the Oregon Cascades as well as the precipitation shadow in the lee of the Cascades.1
Figure 2 shows the S-Pol RHI reflectivities at 0013, 0035, and 0118 UTC 5 December and the Convair ice water contents (IWCs) for the ∼20-min period surrounding each radar time (CV in Fig. 1a). IWC fluctuates significantly along the flight tracks, with 2–3 times larger IWC within convective cells than the surrounding regions. At 0013 UTC at 4.6 km MSL (−26°C; Fig. 2a), the IWC is generally less than 0.1 g m−3, with reflectivities less than 4 dBZ. The 2D-C imagery suggests small unidentifiable ice crystals at this level (Fig. 2a). At 0035 UTC at 4.5 km MSL (Fig. 2b), the IWC is within the same range as the first leg with peaks at ∼0.16 g m−3 at 122.6°W, where the Convair entered the upper portion of a convective cell of ∼10 dBZ. At 0118 UTC (Fig. 2c), the IWC reaches ∼0.20 g m−3 over the eastern portion of the second cross-barrier leg as the Convair entered another convective cell over the Cascade crest at 4.2 km MSL (−22°C). Larger unidentifiable crystals are prevalent within these cells (Figs. 2b,c). In contrast, there is less than 0.1 g m−3 IWC between the convective cells at 4.2 MSL over the Willamette Valley.
Figure 3 shows the P-3 reflectivities and in situ IWC (sum of snow and graupel) and CLW amounts along leg 2 (2351–0009 UTC at 2.45 km MSL or −9.5°C) and leg 3 (0016–0032 UTC at 3.35 km MSL or −15°C) over the Cascade windward slope (track shown in Fig. 1a). There are large variations in the ice and water mass along the legs as a result of the convective cells. Within the enhanced precipitation areas (∼20 dBZ) along leg 2 (Fig. 3a), there are relatively large IWCs (∼0.3 g m−3) and relatively large aggregates of dendrites (2–3 mm). In contrast, between the precipitation cells near 44.25°N (Fig. 3a), the IWC is ∼0.05 g m−3 and the dendrites crystals are <2 mm. Both the reflectivities and IWCs along leg 3 are approximately 50% less than leg 2 (Fig. 3b), which suggests a rapid depositional and aggregate growth between 3.5 and 2.5 km MSL. There is very little riming observed in the 2D-P imagery for legs 2 and 3 (not shown). However, at Tombstone Pass, near the Cascade crest (TS in Fig. 1b), ground observers reported moderately rimed dendritic assemblages around 0000 UTC 5 December (Stoelinga et al. 2007). This suggests that riming was mainly prevalent below 2.5 km MSL during this event. Upstream of the Cascades along leg 1 (1.85 km MSL or −5.5°C), there are IWC and CLW amounts of 0.09 and 0.06 g m−3 (Table 2), respectively, associated with lightly rimed bullets and stellar type ice crystals (not shown).
Figure 4 shows the leg-averaged ice particle size distributions from NOAA P-3 legs 1–3 and the 4.6 km MSL Convair leg. The snow size distribution over the windward Cascades is less steep (broader) along the P-3 leg 2 at −10°C than the Convair at −26°C, thus highlighting the snow growth and aggregation at low levels. The snow size distribution for the P-3 has a slightly steeper exponential distribution for particles smaller than 1 mm. Field et al. (2005) labeled this behavior as the shoulder-feature in the size distribution related to aggregation. The snow intercept values for the 4–5 December event are generally 2–3 times smaller than the heavier precipitation event on 13–14 December for the same temperatures aloft (e.g., Fig. 12 of Woods et al. 2005; Fig. 8 of Garvert et al. 2005b). Thus, the snow intercept increases with greater snow mixing ratio and decreasing temperatures, which is why both variables are used to determine the intercept in some BMPs (Thompson et al. 2008).
4. WRF microphysical verification
a. Surface precipitation
Figure 5 shows the 12-h (2000 UTC 4 December–0800 UTC 5 December) surface precipitation from four WRF simulations using different BMPs at 1.33-km grid spacing. In general, all four BMPs produce two precipitation maxima, one along the Coast Range (20–40 mm) and another over the windward slopes and crest of the Cascades (20–50 mm). THOM1 and THOM2 predict a similar precipitation distribution and maximum of ∼45 mm over the windward ridges of the Cascades, with THOM2 generating approximately 10% more precipitation than THOM1 over the Cascade windward slopes. In contrast, the WSM6 and LIN schemes generate more localized precipitation maxima over the Cascade windward slopes, with a few 60–65-mm areas. The THOM1 and THOM2 schemes have more precipitation spillover in the lee of Cascades than the LIN and WSM6, with the 5-mm contour extending ∼10 km farther to the east in THOM1 and THOM2.
Figure 6a shows an east–west cross section of meridionally averaged precipitation for the four BMP simulations over the Coast Range and Cascades (black box in Fig. 5a) from 2000 to 0800 UTC 4–5 December 2001. The maximum precipitation over the Coast Range is ∼10 km to the west in LIN and WSM6 than THOM1 and THOM2, which suggests more rapid precipitation fallout in LIN and WSM6. The LIN scheme produces the largest average precipitation over the Coast Range (121 cm), followed by WSM6 (119 cm), THOM2 (96 cm), and THOM1 (91 cm). In contrast, the precipitation amounts over the Cascades for the various BMPs are within 10%, and their precipitation profiles have a similar shape. This suggests that the convective cells over the narrow Coast Range have a greater microphysical and precipitation sensitivity than the precipitation over the broader and higher Cascade barrier.
The surface gauge precipitation and the model percent of observed precipitation at each gauge site within the box in Fig. 5a were projected onto a series of longitude points. A Cressman analysis was applied one-dimensionally using a grid spacing of 0.5° longitude and a radius of influence of 0.3° of longitude in order to quantify the meridional average of gauge precipitation and the model percent of observed precipitation (Fig. 6b). Rain gauge data show two maxima (3–4 cm), one over the Coast Range and the other over the upper Cascades windward slope, with much lighter (<0.5 cm) precipitation in the lee of Cascades. THOM1 and THOM2 underpredict precipitation by ∼30% over the Coast Range and Willamette Valley. There is some overprediction over the lower Cascade windward slopes, especially for the LIN. There is dramatic overprediction (>200%) in the immediate lee for THOM1 and THOM2. This is the result of too much snow spillover aloft (not shown), which is consistent with the 13–14 December IMPROVE-2 study (Garvert et al. 2005b). About 60 km east of the Cascades crest, all simulations underpredict the precipitation by 30%–50%.
b. Microphysical differences in the WRF BMPs
The east–west cross section of hydrometeors meridionally averaged over the black box in Fig. 5a was constructed for the four BMP simulations (Fig. 7). Cloud ice is negligible below 5 km MSL except for WSM6, which uses a relative larger maximum ice size (500 μm) before converting ice to snow (Hong et al. 2004). THOM1 produces the largest CLW (0.2 g kg−1) extending up to 4.5 km MSL over the Coast Range and Cascades windward slopes, while the WSM6 and THOM2 have the least CLW (<0.1 g kg−1) concentrated below 3 km MSL, especially over the Cascades. As will be shown below, the much lower CLW in THOM2 is likely from strong depositional growth and overprediction of snow aloft, while lower CLW in WSM6 is due to the competition between abundant cloud ice and cloud water (i.e., Bergeron process) and the efficient removal by riming of snow.
Snow dominates over the Coast Range and Cascades for THOM2, while THOM1 generates comparable graupel over the Coast Range as LIN and WSM6. In THOM1, when the snow accretion of cloud water process (Psacw) is 5 times larger than the snow depositional growth, Psacw contributes to graupel growth; otherwise, Psacw contributes to snow growth (Thompson et al. 2004). Graupel is more dominant than snow for LIN and WSM6, with graupel extending up to 3.5 km MSL. LIN has the least amount of hydrometeor mass aloft, which suggests more rapid precipitation fallout. THOM1 and THOM2 generate more snow spillover into the lee of Cascades than WSM6 and LIN as a result of more snow aloft with the smaller fall speed than LIN and WSM6.
The relative humidity with respect to ice (RHI), which affects the ice and snow depositional growth, was evaluated over the Cascades. The RHI peaks around −20°C at 115% and 112% in THOM1 and THOM2, respectively (not shown). In contrast, WSM6 and LIN have RHIs of 102%–104%. The small RHI in WSM6 is likely from the large amount of cloud ice aloft (Fig. 7c), which depletes the ice supersaturation quickly. LIN uses the saturation adjustment method of Tao et al. (1989), which adjusts the water vapor to the mass-weighted combination of the saturation values over liquid water and ice. Since the saturation adjustment between the cloud ice and cloud water depends on the environmental temperatures, no explicit formulation for the depositional growth of ice is needed in the LIN scheme. This is in contrast to other BMPs, which adjusts the water vapor to saturation values over liquid water (Dudhia 1989). A separate 1.33-km WRF simulation using the Dudhia (1989) saturation adjustment in the LIN scheme increases the RHI to 130% at ∼5 km MSL and it reduces the surface precipitation over the Coast Range and Cascade windward slopes and crest by 10%–15% (not shown).
c. Cloud water and water vapor verification
Figure 8 shows the observed LWD from the radiometer at Santiam Junction, which is located ∼8 km west of the Cascade crest at 1.1 km MSL (SJ in Fig. 1a). This LWD represents mainly integrated cloud liquid water, since there is little rainwater at this height given the relatively low freezing level (∼1 km MSL). Before 0000 UTC 5 December, the LWD measurements oscillate between 0.20 and 0.50 mm in association with intermittent convective cells. Prior to the passage of the midlevel trough between 0000 and 0700 UTC 5 December, the LWD gradually increases from 0.20 to 0.55 mm. With the passage of the surface trough at 0900 UTC 5 December, the LWD decreases to 0.18 mm. Afterward, the LWD is ∼0.40 mm with the postfrontal convection. Overall, the LWDs during this event are 80% less than the 13–14 December event (cf. Fig. 7 of Garvert et al. 2005b). The freezing level was near 2 km above MSL and the cross-barrier flow was twice as strong in the 13–14 December event, both of which favor more supercooled water generation at low levels.
The four BMP simulations produce LWD oscillations of similar amplitude as the observed (∼0.1 mm) as a result of the convective cells and there is a gradual increase of LWD after 1800 UTC 4 December as observed; however, most schemes underpredict the LWD by 30%–70%. In general, THOM1 predicts the largest LWD (∼0.25 mm) from 2300 to 0200 UTC 4–5 December, followed by LIN (∼0.20 mm). WSM6 and THOM2 predict the smallest LWD (∼0.10 mm). The CLW underprediction occurs even though the 1.33-km WRF vertical motions are relatively well predicted along the NOAA P-3 flight legs (Colle et al. 2008).
Figure 9a shows the simulated and observed column integrated water vapor (IWV) at the UW and BB sounding sites (Fig. 1b). During the early period (from 1500 to 0500 UTC 4–5 December) the IWV increases at both sounding locations, and there is nearly 2 times more IWV at UW than in the immediate lee at BB. This difference is from the higher altitude at BB (1027 m MSL) and precipitation fallout upstream of BB (not shown). Before 0300 UTC 5 December, the THOM1 and THOM2 runs have approximately 0.1 cm (∼8%) more IWV than observed, while the LIN and WSM6 overprediction is 0.02–0.05 cm. In contrast, at the BB site there is a ∼20% IWV underprediction in LIN, while THOM1 and THOM2 are only slightly drier than observed (less than 10%) during 0300–0500 UTC 5 December.
To quantify the vertical distribution of moisture errors in the WRF, Fig. 9b shows the simulated water vapor mixing ratio error at UW and BB at 0300 UTC 5 December 2001. At UW, all the simulations overpredict the moisture in the boundary layer (below 1 km MSL), which contributes to the overpredicted IWV in Fig. 9a. Above 2 km MSL, the error is generally less than 0.2 g kg−1. All schemes are too dry below 3 km MSL in the lee at BB. LIN and WSM6 develop a large dry error from the windward to the lee, which suggests that too much water vapor was removed over the windward slopes.
d. Verification using radar and aircraft
Figure 10a shows a time series of the observed radial velocities in the vertical over the S-band profiler site (MB in Fig. 1a) for a 6-h period (0000 to 0600 UTC 5 December). The radial velocity represents a combination of the vertical air motions and hydrometeor fall speeds; however, the 1.33-km WRF suggests that 90% of the time-averaged vertical motion is from the fall speed at this valley location at 0.5 km MSL (not shown). As with other IMPROVE-2 events (Houze and Medina 2005), there are turbulent updraft and downdraft cells of ∼1 m s−1 within a vertical shear layer near 1.5 km MSL (cf. Fig. 13 of Colle et al. 2008). Above 3.5 km MSL, the average velocity is downward at about 1 m s−1 as a result of the falling snow particles. As the snow crystals become more rimed approaching the surface, the net downward motion gradually increases to 1.8 m s−1. There is a secondary reflectivity maximum near 3 km MSL at around −15°C (Fig. 15 in Colle et al. 2008), where the mean downward velocity decreases by ∼0.2 m s−1. The slower fall speed is consistent with rapid dendritic snow growth at this level (Fig. 3).
Figure 10b,c show the 6-h mean reflectivity-weighted snow/graupel fall speeds for the various 1.33-km WRF BMPs and their mixing ratios during the same 6-h period at the S-band site using 15-min WRF output. The reflectivity-weighted fall speed is calculated by assuming that the reflectivity is proportional to D6 for THOM1, LIN, and WSM6 (Thompson et al. 2008). The vertical air motions in WRF were not included in the analysis, since most of the WRF vertical velocities are from the hydrometeor fall speeds. The WSM6 and LIN schemes predict increasing fall speed with decreasing altitude, with maximum fall speeds of 3.9 and 5.7 m s−1 at 1.5 and 2 km MSL, respectively. In contrast, the fall speed in THOM1 decreases from 2.4 to 1.6 m s−1 toward the surface, mainly due to the increasing air density and the dominance of snow at low levels. THOM2 realistically predicts the observed increase in fall speed below 3 km MSL, but it has a ∼0.5 m s−1 overprediction below 3 km MSL and above the freezing level (∼1 km MSL). For THOM1 and THOM2 above the S-band site, snow dominates over graupel by more than a factor of 5, with peak snow mixing ratios near 2 km MSL. In contrast, nearly all hydrometeors in the LIN and WSM6 runs are graupel below 2 km MSL, which is leading to the excessive fall speeds.
Using the NOAA P-3 and Convair data (P3 and CV leg in Fig. 1a), the CLW, IWC, and vertical motions in the 1.33-km WRF runs were verified. The model fields were interpolated in time and space to the flight tracks using 15-min output from the 1.33-km WRF. To determine whether relatively small timing errors in the WRF had an impact on the results, two other interpolations were done using WRF data an hour before or after the aircraft times. The time-shifted verification results are generally within 10% of the aircraft time (not shown).
Figure 11 shows in situ measured and simulated CLW, IWC, and vertical motions (Convair did not measure vertical motions) for the Convair legs at 4.2–4.6 km above MSL (−22° to −26°C) between 0003 and 0115 UTC (leg CV in Fig. 1a). The four BMPs predict similar vertical motions along the flight track; however, only the THOM1 generates CLW amounts comparable to the observed (0.01–0.07 g m−3). The Convair measured IWC of ∼0.25 g m−3 at 4.2 km MSL within the convective cells and ∼0.05 g m−3 outside the cells. All schemes also predict significant fluctuations of IWC along the flight track reflecting the convective cell characteristics (Fig. 11). The WSM6, THOM1, and THOM2 schemes have mean IWCs within ∼20% of observed at this height (Table 2). In contrast, the LIN scheme underpredicts the IWCs by ∼30% at this altitude on average. The LIN scheme also produces about 0.02 g m−3 of graupel at this height, while none was observed (Table 2).
At 3.3-km MSL along leg 3 (Fig. 13), the P-3 has <0.05 g m−3 of CLW and ∼0.08 g m−3 of IWC, with little evidence of riming in the 2D-C imagery at this level (Fig. 3b). All schemes predict negligible CLW at this level, except THOM1, which produces 0.05 g m−3 CLW for part of the leg. However, WSM6, THOM1, and THOM2 produce approximately 3 times more IWC than observed at this altitude (Table 2). In contrast, LIN produces comparable amounts of snow and graupel at this level (not shown). The sum of snow and graupel is within 20% of observed for LIN; however, it was shown earlier that the LIN overpredicts precipitation at the surface because of the larger fall speed and precipitation rate (Figs. 6b and 10). At P-3 leg 3, the temperature diagnosed intercept in THOM1 is close to the observed (Fig. 12c), while the fixed intercept of LIN is still a factor of 3–4 smaller than observed. Meanwhile, THOM2 also overpredicts the IWC and has more snow particles <1 mm than observed.
Figure 14 compares the IWCs from the P-3 and 1.33-km WRF along leg 2 (2.45 km MSL or about −9.5°C). The four WRF BMPs realistically simulated the vertical motions of ±0.5–1.5 m s−1 along this leg. However, all four schemes underpredict CLW along the leg except the THOM1, which has a mean CLW of 0.13 g m−3. Both THOM1 and THOM2 overpredict the IWC along most of the leg 2 by 55%–65% on average (Fig. 14a; Table 2). LIN produces slightly less IWC (0.14 g m−3) than observed (0.20 g m−3), while WSM6 overpredicts the IWC by ∼35% (Table 2). All the BMP simulations underpredict the number concentrations for particles >2.5 mm along leg 2 (Fig. 12d); however, model IWC is still slightly larger than observed except for LIN (Table 2). This suggests that the density of observed large particles is smaller than that assumed in these BMPs (100 kg m−3).
5. Microphysical sensitivity tests
The partition of snow and graupel aloft impacts the fallout and surface precipitation distribution. Both LIN and WSM6 predict more graupel than THOM1 and THOM2 (Fig. 7). One question is why the WSM6 and LIN generate more graupel than the Thompson schemes? Several sensitivity runs were completed with the LIN scheme, such as using the same temperature-dependent snow intercept and graupel intercept as in THOM1, turning off the snow to graupel autoconversion, and using the same fall speed of snow, rain, and graupel as in THOM1. These sensitivity runs only marginally increased the snow aloft (by <20%) and did not change the surface precipitation significantly (not shown).
As shown in Lin et al. (1983), the production of graupel by collecting snow (Pgacs) can be one order of magnitude larger than the snow depositional growth (Psdep). Pgacs is not used in the Thompson schemes, thus, we hypothesized that Pgacs may favor the enhanced graupel growth in the WSM6 and LIN. A simulation with the Pgacs turned with (LIN_NOGACS) reduced the precipitation bull’s-eye and the precipitation is shifted downwind as compared with LIN (Fig. 15c), which reduced the surface overprediction over the windward slope (Fig. 16). The LIN_NOGACS produces more snow aloft (Figs. 7d and 17c), with the greater snow depositional growth reducing the CLW generation. A simulation with Pgacs turned off in the WSM6 generated the same result (not shown).
Snow overprediction aloft has been noted in mesoscale numerical simulations (Garvert et al. 2005b; Colle et al. 2005a; Milbrandt et al. 2008). For the 13–14 December IMPROVE-2 event using the THOM1 scheme, Colle et al. (2005a) found that snow depositional growth contributed ∼70% to the snow growth, followed by riming growth (∼30%). Snow diffusional growth depends on the environmental supersaturation, particle electrical capacitance, ventilation effect, etc. (Rogers and Yau 1989; Fukuta and Takahashi 1999). Particle capacitance is a function of the size and shape of the ice particle. Capacitance for a sphere (C = 0.5D, where D is the diameter) has been widely used in the snow deposition and sublimation parameterization in BMPs. Recently, Westbrook et al. (2008) showed that capacitance of snow aggregates was only half that of a sphere, which will reduce the snow deposition rate. A third sensitivity run (THOM1_0.5SDEP) that reduced the snow deposition and sublimation rates by one-half was completed to test the snow depositional growth uncertainty on model simulated microphysics and surface precipitation. This run increases the surface precipitation by ∼10% over the Coast Range and decreases it by ∼10% over the windward slopes of Cascades (Fig. 15b). Snow aloft is reduced by ∼60% over the Cascades in the snow depositional growth zone from −20° to −10°C, but the graupel and CLW are increased by ∼80% and 20% below 3 km MSL, respectively (Fig. 17b). As a result, the IWC for this run verifies better along the flight tracks; however, the CLW increases by 50% and verifies worse (Table 2). This implies that other errors in the BMPs other than snow depositional growth are likely important. In addition, other errors in the model, such as the upstream moisture bias and mountain wave lifting, might also contribute to the microphysical errors.
Table 3 summarizes the water budget variables and these DR/RT parameters. EVAP is less than one unit (109 kg) for all simulations and NETQ differs less than three units among the simulations. For brevity, these two variables are not included in Table 2. For box A in Fig. 5b, the DRs for the WSM6 (14%) and LIN (16%) are larger than THOM1 (9%) and THOM2 (9%). The DR over the Coast Range is approximately 40% less than the Cascades for WSM6 and LIN, while it is ∼60% less than the Cascades for THOM1 and THOM2. The four BMPs have comparable DRs over box B in Fig. 5b over the Cascades given their comparable water vapor flux and precipitation (Table 2). Corresponding to the larger surface precipitation and smaller hydrometeor mass aloft (cf. Figs. 8 and 9), LIN has the smallest RT (1200 and 1000 s for boxes A and B). The RT for WSM6 is 1900 and 1700 s for box A and B, respectively, while the RT for THOM1 is around 3000 s in box A and 2500 s in box B. THOM2 has a RT of 3100 s in box A and 2200 s in box B. Relatively larger condensed water mass and smaller precipitation rates over the Coast Range (especially for THOM1 and THOM2; cf. Fig. 6) contribute to the relatively larger RT over the Coast Range than the Cascades for all schemes. The smaller RTs in LIN and WSM6 are from their greater graupel production than other schemes.
Braun (2006) examined artificial source terms for cloud and precipitation mass within a hurricane simulation associated with setting negative mixing ratios to zero for the numerical advection of hydrometeors. He found that the accumulative effect of these artificial source terms contributed to a mass equivalent to 15%–20% of the surface precipitation. To better conserve mass, a PDA scheme has been developed for the WRF (Skamarock 2006) and it has reduced the large positive precipitation bias for convection (Skamarock and Weisman 2009). To test the impact of the PDA, four additional simulations of the BMPs were completed without the PDA scheme turned on (NOPD simulations in Table 2). The non-PDA simulations have 25%–45% more surface precipitation than the PDA simulations over box A and ∼10% more precipitation than the PDA over box B (Fig. 15d; Table 2). The SUM is slightly positive using the PDA (3%–8% of the HFLX for the PDA simulations), while the SUM is negative (∼20% of HFLX) for the non-PDA runs. A negative SUM implies the WRF generates more precipitation than available water sources. Overall, water is better conserved for the PDA simulations. The SUM is more (3–4 times) negative over box A than box B for non-PDA runs, which implies that the large horizontal gradients of hydrometeors associated with convective cells over the Coast Range tend to generate more negative values without PDA. The non-PDA runs also increase the hydrometeor mass aloft by ∼10%–20% in general (Fig. 17d). The non-PDA simulation has 30%–40% more precipitation than the PDA simulations over the Coast Range (Fig. 15d).
6. Summary and conclusions
This paper describes the microphysical structures and model verification of a moderate orographic rainfall event during 4–5 December 2001 IMPROVE-2 IOP6. Because of the relatively low freezing level (1 km MSL) and moderate cross-barrier flow (15–20 m s−1), this event features less CLW (20%–30%), half the snow mass, and half the surface precipitation as the previously documented 13–14 December 2001 IMPROVE-2 event.
In situ particle images and retrieved IWC from the Convair and P-3 suggest large vertical and horizontal variation of microphysical characteristics. Small unidentifiable cold-type ice particles above 4 km MSL grow in size via deposition and aggregation as they fall toward the warmer temperatures. Large aggregate and dendritic particles were observed in transient convective cells resulting from the increased depositional and aggregation growth, while outside of the convective cells, the ice particles were smaller and the IWC was less.
The 4–5 December event was simulated down to 1.33-km grid spacing using four BMPs within WRF. Model surface precipitation and microphysics were verified with rain gauge data as well as in situ radar and aircraft measurements. All simulations slightly underpredict the precipitation over the Coast Range and Willamette Valley, while there is no widespread overprediction over the windward slopes of the Cascades except in the LIN scheme. There is an underprediction farther east of the Cascades for all the simulations, which implies that the simulated cloud dissipated too rapidly within the lee subsidence. The LIN and WSM6 schemes have more localized and overpredicted areas of precipitation over the windward slope and less spillover into the lee of the Cascades because of the dominance of graupel aloft.
Graupel dominates over snow in WSM6 and LIN, while snow dominates in THOM2. Microphysics verification suggests BMPs tend to overpredict snow in the maximum snow depositional growth region aloft unless they unrealistically convert the snow to graupel and fall out as quickly as in LIN. THOM1 realistically predicts CLW along the flight tracks, while the other three schemes underpredict CLW, since too much snow was produced at the expense of the CLW. For the reflectivity-weighted snow/graupel fall speed, LIN and WSM6 predict a larger fall speed than observed given the dominance of graupel and associated larger fall speeds. THOM1 generates a decreasing fall speed to 1.6 m s−1 with decreasing altitude. THOM2 is able to reproduce the general pattern of the observed fall speed with ∼0.5 m s−1 larger fall speed below 3.5 km MSL.
Sensitivity tests with graupel collection of snow turned off in LIN and WSM6 reduce the graupel and increase the snow aloft significantly. Consequently, surface precipitation is more uniform over the Cascades and there is more precipitation spillover into the lee. Another sensitivity test using an ∼D2 mass–diameter relationship rather than ∼D3 with varying snow density generates a relatively large (∼10%) sensitivity on the microphysics aloft and surface precipitation. These BMP sensitivity simulations suggest the importance of the snow and graupel partitioning. A third sensitivity simulation with half the snow deposition and sublimation of the THOM1 scheme was completed to test the impact of using a capacitance that is likely overestimated for the assumed spherical snow crystals in the model. A simulation with half the capacitance reduces the IWC by ∼30% in the snow depositional growth region, which verifies better with the observations than THOM1. However, the CLW and graupel increases to compensate for decreased snow and, thus, the total hydrometeor mass aloft only changes slightly. The large precipitation of LIN and WSM6 over the Coast Range is mainly due to a larger precipitation efficiency associated with rapid graupel fallout. The positive definite advection (PDA) scheme in WRF was compared with the non-PDA approach, which set negative mixing ratios to zero and produced an artificial source of water. Although this artificial water mass is negligible at any given grid point, its accumulative impact over the simulation is substantial and contributes to 10%–20% of the total surface precipitation, especially within the convective cells over the Coast Range. It is shown that a PDS scheme for moisture and hydrometeors improves the water conservation and reduces the surface precipitation over the whole 1.33-km domain by ∼10%.
Two IMPROVE case studies have illustrated snow overprediction aloft over the Cascades (Garvert et al. 2005b and this paper), and a leeside dry bias. Although increasing the conversion to graupel decreases this overprediction aloft (LIN scheme), there is surface overprediction over the windward slopes due to the unrealistic fallout of hydrometeors. As a result, other aspects of these schemes need to be investigated, such as the snowfall speed and density changes aloft. It is also still possible that the model hydrometeor and precipitation errors are also caused by the incorrect upstream moisture transportation and mountain wave lifting resulting from the model initialization and other model parameterization (e.g., boundary layer) errors.
Acknowledgments
This research was supported by the National Science Foundation (ATM-0450444). We thank Drs. Christopher Woods and David Kingsmill for processing the Convair and NOAA P-3 microphysical data and Dr. Socorro Medina for the S-Pol data. We appreciate the constructive comments by Greg Thompson and the two anonymous reviewers that helped us to improve the manuscript. Use of WRF was made possible by the Microscale and Mesoscale Meteorological Division of the National Center for Atmospheric Research (NCAR), which is supported by the National Science Foundation.
REFERENCES
Baumgardner, D., 1983: An analysis and comparison of five water droplet measuring instruments. J. Climate Appl. Meteor., 22 , 891–910.
Bougeault, P., and Coauthors, 2001: The MAP special observing period. Bull. Amer. Meteor. Soc., 82 , 433–462.
Braun, S. A., 2006: High-resolution simulation of Hurricane Bonnie (1968). Part II: Water budget. J. Atmos. Sci., 63 , 43–64.
Brown, P. R. A., and P. N. Francis, 1995: Improved measurements of the ice water content in cirrus using a total-water probe. J. Atmos. Oceanic Technol., 12 , 410–414.
Chen, S-H., and W-Y. Sun, 2002: A one-dimensional time dependent cloud model. J. Meteor. Soc. Japan, 80 , 99–118.
Colle, B. A., 2008: Two-dimensional idealized simulations of the impact of multiple windward ridges on orographic precipitation. J. Atmos. Sci., 65 , 509–523.
Colle, B. A., and C. F. Mass, 2000: The 5–9 February 1996 flooding event over the Pacific Northwest: Sensitivity studies and evaluation of the MM5 precipitation forecasts. Mon. Wea. Rev., 128 , 593–617.
Colle, B. A., M. F. Garvert, J. B. Wolfe, C. F. Mass, and C. P. Woods, 2005a: The 13–14 December 2001 IMPROVE-2 event. Part III: Simulated microphysical budgets and sensitivity studies. J. Atmos. Sci., 62 , 3535–3558.
Colle, B. A., J. B. Wolfe, W. J. Steenburgh, D. E. Kingsmill, J. A. Cox, and J. C. Shafer, 2005b: High-resolution simulations and microphysical validation of an orographic precipitation event over the Wasatch Mountains during IPEX IOP3. Mon. Wea. Rev., 133 , 2947–2971.
Colle, B. A., Y. Lin, S. Medina, and B. Smull, 2008: Orographic modification of convection and flow kinematics by the Oregon coastal range and Cascades during IMPROVE-2. Mon. Wea. Rev., 136 , 3894–3916.
Cooper, W. A., 1986: Ice initiation in natural clouds. Precipitation Enhancement—A Scientific Challenge, Meteor. Monogr., No. 43, Amer. Meteor. Soc., 29–32.
Dudhia, J., 1989: Numerical study of convection observed during the winter monsoon experiment using a mesoscale two-dimensional model. J. Atmos. Sci., 46 , 3077–3107.
Field, P. R., R. J. Hogan, P. R. A. Brown, A. J. Illingworth, T. W. Choularton, and R. J. Cotton, 2005: Parametrization of ice-particle size distributions for mid-latitude stratiform cloud. Quart. J. Roy. Meteor. Soc., 131 , 1997–2017.
Fukuta, N., and T. Takahashi, 1999: The growth of atmospheric ice crystals: A summary of findings in vertical supercooled cloud tunnel studies. J. Atmos. Sci., 56 , 1963–1979.
Garvert, M. F., B. A. Colle, and C. F. Mass, 2005a: The 13–14 December 2001 IMPROVE-2 event. Part I: Synoptic and mesoscale evolution and comparison with a mesoscale model simulation. J. Atmos. Sci., 62 , 3474–3492.
Garvert, M. F., C. P. Woods, B. A. Colle, C. F. Mass, P. V. Hobbs, M. T. Stoelinga, and J. B. Wolfe, 2005b: The 13–14 December 2001 IMPROVE-2 event. Part II: Comparisons of MM5 model simulations of clouds and precipitation with observations. J. Atmos. Sci., 62 , 3520–3534.
Grubisić, V., R. K. Vellore, and A. W. Huggins, 2005: Quantitative precipitation forecasting of wintertime storms in the Sierra Nevada: Sensitivity to the microphysical parameterization and horizontal resolution. Mon. Wea. Rev., 133 , 2834–2859.
Heymsfield, A. J., A. Bansemer, S. Lewis, J. Iaquinta, M. Kajikawa, C. Twohy, M. R. Poellot, and L. M. Miloshevich, 2002: A general approach for deriving the properties of cirrus and stratiform ice cloud particles. J. Atmos. Sci., 59 , 3–29.
Hong, S-Y., and J. J. Lim, 2006: The WRF single-moment 6-class microphysics scheme (WSM6), 2006. J. Korean Meteor. Soc., 42 , 129–151.
Hong, S-Y., J. Dudhia, and S-H. Chen, 2004: A revised approach to ice microphysical processes for the bulk parameterization of clouds and precipitation. Mon. Wea. Rev., 132 , 103–120.
Houze, R. A., and S. Medina, 2005: Turbulence as a mechanism for orographic precipitation enhancement. J. Atmos. Sci., 62 , 3599–3623.
Houze, R. A., P. V. Hobbs, P. H. Herzegh, and D. B. Parsons, 1979: Size distributions of precipitation particles in frontal clouds. J. Atmos. Sci., 36 , 156–162.
Jiang, Q., and R. B. Smith, 2003: Cloud timescales and orographic precipitation. J. Atmos. Sci., 60 , 1159–1172.
Kain, J., and M. Fritsch, 1993: Convective parameterization for mesoscale models: The Kain–Fritsch scheme. The Representation of Cumulus Convection in Numerical Models, Meteor. Monogr., No. 24, Amer. Meteor. Soc., 165–170.
Kessler, E., 1969: On the Distribution and Continuity of Water Substance in Atmospheric Circulations. Meteor. Monogr., No. 32, Amer. Meteor. Soc., 84 pp.
King, W. D., D. A. Parkin, and R. J. Handsoworth, 1978: A hot wire liquid water device having fully calculated response characteristics. J. Appl. Meteor., 17 , 1809–1813.
Koenig, L. R., 1971: Numerical modeling of ice deposition. J. Atmos. Sci., 28 , 226–237.
Lin, Y. L., R. Farley, and H. D. Orville, 1983: Bulk parameterization of the snow field in a cloud model. J. Climate Appl. Meteor., 22 , 1065–1092.
Manning, K. W., and C. A. Davis, 1997: Verification and sensitivity experiments for the WISP94 MM5 forecasts. Wea. Forecasting, 12 , 719–735.
Milbrandt, J. A., M. K. Yau, J. Mailhot, and S. Bélair, 2008: Simulation of an orographic precipitation event during IMPROVE-2. Part I: Evaluation of the control run using a triple-moment bulk microphysics scheme. Mon. Wea. Rev., 136 , 3873–3893.
Reisner, J., R. M. Rasmussen, and R. T. Bruintjes, 1998: Explicit forecasting of supercooled liquid water in winter storms using the MM5 mesoscale model. Quart. J. Roy. Meteor. Soc., 124 , 1071–1107.
Richard, E., A. Buzzi, and G. Zängl, 2007: Quantitative precipitation forecasting in the Alps: The advances achieved by the Mesoscale Alpine Programme. Quart. J. Roy. Meteor. Soc., 133 , 831–846.
Rogers, R. R., and M. K. Yau, 1989: A Short Course in Cloud Physics. Pergamon Press, 293 pp.
Rutledge, S. A., and P. V. Hobbs, 1983: The mesoscale and microscale structure and organization of clouds and precipitation in midlatitude cyclones. VIII: A model for the “seeder–feeder” process in warm-frontal rainbands. J. Atmos. Sci., 40 , 1185–1206.
Rutledge, S. A., and P. V. Hobbs, 1984: The mesoscale and microscale structure and organization of clouds and precipitation in midlatitude cyclones. XII: A diagnostic modeling study of precipitation development in narrow cold-frontal rainbands. J. Atmos. Sci., 41 , 2949–2972.
Schultz, D. M., 2002: Understanding Utah winter storms: The Intermountain Precipitation Experiment. Bull. Amer. Meteor. Soc., 83 , 189–210.
Skamarock, W. C., 2006: Positive-definite and monotonic limiters for unrestricted-time-step transport schemes. Mon. Wea. Rev., 134 , 2241–2250.
Skamarock, W. C., and M. L. Weisman, 2009: The impact of positive-definite moisture transport on NWP precipitation forecasts. Mon. Wea. Rev., 137 , 488–494.
Skamarock, W. C., J. B. Klemp, J. Dudhia, D. O. Gill, D. M. Barker, W. Wang, and J. G. Powers, 2005: A description of the Advanced Research WRF, version 2. NCAR Tech. Note NCAR/TN-468+STR, 88 pp. [Available from UCAR Communications, P.O. Box 3000, Boulder, CO 80307.].
Smith, R. B., and I. Barstad, 2004: A linear theory of orographic precipitation. J. Atmos. Sci., 61 , 1377–1391.
Smith, R. B., Q. Jiang, M. Fearon, P. Tabary, M. Dorninger, and J. Doyle, 2003: Orographic precipitation and air mass transformation: An alpine example. Quart. J. Roy. Meteor. Soc., 129 , 433–454.
Smith, R. B., I. Barstad, and L. Bonneau, 2005: Orographic precipitation and Oregon’s climate transition. J. Atmos. Sci., 62 , 177–191.
Stoelinga, M. T., and Coauthors, 2003: Improvement of microphysical parameterization through observational verification experiment. Bull. Amer. Meteor. Soc., 84 , 1807–1826.
Stoelinga, M. T., J. D. Locatelli, and C. P. Woods, 2007: The occurrence of “irregular” ice particles in stratiform clouds. J. Atmos. Sci., 64 , 2740–2750.
Tao, W-K., and J. S. Simpson, 1993: Goddard cumulus ensemble model. Part I: Model description. Terr. Atmos. Oceanic Sci., 4 , 35–72.
Tao, W-K., J. S. Simpson, and M. McCumber, 1989: An ice-water saturation adjustment. Mon. Wea. Rev., 117 , 231–235.
Thompson, G., R. M. Rasmussen, and K. Manning, 2004: Explicit forecasts of winter precipitation using an improved bulk microphysics scheme. Part I: Description and sensitivity analysis. Mon. Wea. Rev., 132 , 519–542.
Thompson, G., P. R. Field, R. M. Rasmussen, and W. D. Hall, 2008: Explicit forecasts of winter precipitation using an improved bulk microphysics scheme. Part II: Implementation of a new snow parameterization. Mon. Wea. Rev., 136 , 5095–5115.
Walko, R. L., W. R. Cotton, M. P. Meyers, and J. Y. Harrington, 1995: New RAMS cloud microphysics parameterization. Part I: The single-moment scheme. Atmos. Res., 38 , 29–62.
Westbrook, C. D., R. J. Hogan, and A. J. Illingworth, 2008: The capacitance of pristine ice crystals and aggregate snowflakes. J. Atmos. Sci., 65 , 206–219.
Woods, C. P., M. T. Stoelinga, J. D. Locatelli, and P. V. Hobbs, 2005: Microphysical processes and synergistic interaction between frontal and orographic forcing of precipitation during the 13 December 2001 IMPROVE-2 event over the Oregon Cascades. J. Atmos. Sci., 62 , 3493–3519.
(a) Observed NOAA P-3 reflectivity (shaded every 3 dBZ) and Doppler radar-derived winds at 3 km MSL for legs 2–4 between 2350 UTC 4 Dec and 0050 UTC 5 Dec 2001. The black lines denote P-3 and Convair (CV) flight legs. (b) As in (a), but for the 1.33-km WRF at 0000 UTC 5 Dec 2001. The terrain is also shown every 0.6 km.
Citation: Monthly Weather Review 137, 4; 10.1175/2008MWR2653.1
S-Pol radar RHI scan reflectivities for the west–east cross section using the gray-shaded scale (every 4 dBZ) at (a) 0013, (b) 0035, and (c) 0118 UTC 5 Dec 2001. In situ ice mass concentrations (g m−3) from Convair are color shaded along the corresponding flight leg sections (for a ∼20-min period surrounding each S-Pol RHI scan time). The crystal images collected from the PMS 2D-C probe (with a 0.8-mm vertical dimension) along the Convair flight legs are also indicated.
Citation: Monthly Weather Review 137, 4; 10.1175/2008MWR2653.1
P-3 radar reflectivities (gray shaded every 3 dBZ) and (top bar) in situ ice mass concentrations and (bottom bar) cloud liquid water concentrations color filled every 0.02 g m−3 along (a) leg 2 at 2.45 MSL and (b) leg 3 at 3.35 km MSL. The crystal images collected from the PMS 2D-C probe (with a 1.6-mm vertical dimension) along the P-3 flight legs are also indicated.
Citation: Monthly Weather Review 137, 4; 10.1175/2008MWR2653.1
Particle size spectra measured by the 2D cloud probes aboard the Convair at 4.6 km MSL and the P-3 legs 1–3. The 2D-C data smaller than 1 mm and 2D-P data larger than 1 mm were combined for the P-3.
Citation: Monthly Weather Review 137, 4; 10.1175/2008MWR2653.1
Surface precipitation (every 15 mm starting from 5 mm) for the 1.33-km domain between 2000 UTC 4 Dec and 0800 UTC 5 Dec 2001 for the (a) THOM1, (b) THOM2, (c) WSM6, and (d) LIN schemes. The 1.33-km terrain is shaded every 300 m for reference. The box in (a) shows the region used for the west–east mean cross section. The boxes A and B in (b) show the region used for the vertical profile and water budget calculations.
Citation: Monthly Weather Review 137, 4; 10.1175/2008MWR2653.1
(a) Simulated 1.33-km WRF precipitation (mm) from 2000 UTC 4 Dec to 0800 UTC 5 Dec 2001 meridionally averaged across the boxed region in Fig. 5a. The average terrain profile is also indicated. The table at the top indicates the total precipitation (cm) across the box and within boxes A and B in Fig. 5b. (b) Percent of observed precipitation (using scale on right-hand side) from the 1.33-km WRF at the precipitation gauge sites in boxed area in Fig. 5a between 2000 UTC 4 Dec and 0800 UTC 5 Dec 2001 for the four BMP schemes. The circles are the observed 12-h precipitation at rain gauges. The thick dashed line denotes 100% of the observed precipitation.
Citation: Monthly Weather Review 137, 4; 10.1175/2008MWR2653.1
A west–east cross section across the box in Fig. 5a showing north–south-averaged CLW mixing ratios (shaded every 0.05 g kg−1), snow (black solid, every 0.05 g kg−1), graupel (black dashed, every 0.05 g kg−1), rain (black dotted, every 0.05 g kg−1), ice (black dot–dashed, every 0.05 g kg−1) for the (a) THOM1, (b) THOM2, (c) WSM6, and (d) LIN schemes. Model fields were averaged from 2300 to 0200 UTC 4–5 Dec 2001 (forecast hours 11–14) using 15-min model output.
Citation: Monthly Weather Review 137, 4; 10.1175/2008MWR2653.1
Vertically integrated LWD measured by the microwave radiometer at Santiam Junction (SJ in Fig. 1a) and from the four 1.33-km microphysical simulations from 1200 UTC 4 Dec to 1200 UTC 5 Dec 2001.
Citation: Monthly Weather Review 137, 4; 10.1175/2008MWR2653.1
(a) Observed (circles) and simulated hourly (solid and dashed lines) column-integrated water vapor depth at UW and Black Butte Ranch (UW and BB in Fig. 1b) from 1200 UTC 4 Dec to 1200 UTC 5 Dec 2001. Model water vapor mixing ratio errors in the vertical for the various WRF BMP runs at (b) UW and (c) BB at 0300 UTC 5 Dec 2001.
Citation: Monthly Weather Review 137, 4; 10.1175/2008MWR2653.1
(a) Vertical pointing S-band radar (location MB in Fig. 1b) vertical radial velocity (color shaded every 0.2 m s−1) from 0000 to 0600 UTC 5 Dec 2001. (right) The scatterplot of radial velocities and the 6-h mean radial velocity during the period (red line). (b) The 1.33-km WRF 6-h reflectivity-weighted mean snow/graupel fall speed from 0000 to 0600 UTC 5 Dec 2001 at the S-band site for the four microphysical schemes as well as the observed mean velocity profile (black solid). (c) The 6-h 1.33-km WRF mean mixing ratio profiles of snow (black) and graupel (gray) at the same location for the four microphysical schemes.
Citation: Monthly Weather Review 137, 4; 10.1175/2008MWR2653.1
(a) Observed (solid lines) and simulated cloud liquid water and ice mass concentrations along the Convair legs for the THOM1 (dotted lines) and THOM2 scheme (dashed lines). (b) As in (a), but for the WSM6 (dotted lines) and LIN (dashed lines) scheme. The model vertical motion is also shown. The two vertical arrows indicate the three Convair flight legs.
Citation: Monthly Weather Review 137, 4; 10.1175/2008MWR2653.1
(a) Different mass–dimension relationships used for observation retrievals (BF, AGGR, and HEYMS) and the various bulk microphysical simulations. CONST is the constant density (100 kg m−3) spherical assumption used in the LIN, WSM6, and THOM1 schemes. Observed and simulated particle size spectra for the (b) Convair leg at 4.6 km MSL, (c) P-3 leg 3, and (d) P-3 leg 2 for the various microphysical runs (LIN, WSM6, THOM1, THOM1_MD, and THOM2).
Citation: Monthly Weather Review 137, 4; 10.1175/2008MWR2653.1
(a) Observed (solid lines) and simulated cloud liquid water, ice mass concentrations, and vertical motions along the P-3 leg 3 for the THOM1 (dotted lines) and THOM2 scheme (dashed lines). (b) Same as (a), but for the WSM6 (dotted lines) and LIN (dashed lines) scheme.
Citation: Monthly Weather Review 137, 4; 10.1175/2008MWR2653.1
(a) Observed (solid lines) and simulated cloud liquid water, ice mass concentrations (IWC), and vertical motions along the P-3 leg 2 for the THOM1 (dotted lines) and THOM2 scheme (dashed lines). (b) Same as (a), but for the WSM6 (dotted lines) and LIN (dashed lines) scheme.
Citation: Monthly Weather Review 137, 4; 10.1175/2008MWR2653.1
12-h precipitation difference (mm) between (a) THOM1_MD, (b) THOM1_0.5SDEP, and (d) THOM1_NOPD and the THOM1. Solid contours means the precipitation is larger than THOM1, while dashed contours show less precipitation than the THOM1. The contour interval is every 5 mm for (a) and 3 mm for (b) and (d). (c) As in (a), but for the 12-h precipitation difference between the LIN_NOGACS and LIN. The terrain is shaded for reference.
Citation: Monthly Weather Review 137, 4; 10.1175/2008MWR2653.1
Percent of observed precipitation from the 1.33-km WRF for the THOM, three THOM1 sensitivity runs (THOM1_MD, THOM1_0.5SDEP, THOM1_NOPD), LIN, and LIN_NOGACS at the precipitation gauge sites in the boxed area in Fig. 5a between 2000 UTC 4 Dec and 0800 UTC 5 Dec 2001. The inset box highlights the line types of the various experiments, while the LIN is denoted by the gray dashed–dotted line. The circles are the observed 12-h precipitation at rain gauges. The thick dashed line denotes 100% of the observed precipitation.
Citation: Monthly Weather Review 137, 4; 10.1175/2008MWR2653.1
A west–east cross section across the box in Fig. 5a showing north–south-averaged mixing ratios of CLW (shaded every 0.05 g kg−1), snow (black solid, 0.05 g kg−1), graupel (black dashed, 0.05 g kg−1), rain (black dotted, 0.05 g kg−1), and ice (black dotted, 0.05 g kg−1) for the (a) THOM1_MD, (b) THOM1_0.5SDEP, (c) LIN_NOGACS, and (d) THOM1_NOPD runs. Model fields were temporally averaged from 2300 to 0200 UTC 4–5 Dec 2001 (forecast hours 11–14) using the model outputs every 15 min.
Citation: Monthly Weather Review 137, 4; 10.1175/2008MWR2653.1
Summary of different WRF simulations performed. The indices c, i, r, s, and g denote cloud droplets, cloud ice, rain, snow, and graupel, respectively. Mixing ratio and number concentration is indicated by q and N, respectively. The Marshall–Palmer (MP) distribution (exponential distribution) is also noted. Pgacs represents graupel accretion of snow; Psdep and Pssub are snow depositional and sublimational growth, respectively.
Observed CLW, total (snow and graupel) IWC, and graupel (g m3) vs the WRF at 1.33-km grid spacing averaged along the P-3 and Convair flight legs shown in Fig. 1a.
Water budget results for boxes A and B (separated by vertical bars; see Fig. 5b for boxes). Water flux and precipitation are in units of 109 kilograms. See text for further details.
A comprehensive verification of the WRF kinematics and precipitation structures for this event was presented in Colle et al. (2008).