SEARCH

SEARCH BY CITATION

Keywords:

  • Great Lakes;
  • MODIS;
  • WRF

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Model, Data Sources, and Methodology
  5. 3. Results
  6. 4. Conclusions and Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information

[1] In this study, remotely sensed lake surface temperature (LST) and lake ice cover (LIC) were integrated into the Advanced Research Weather Research and Forecasting (WRF) model version 3.2 to evaluate the simulation of lake-effect precipitation over the Great Lakes region. The LST was derived from the Moderate Resolution Imaging Spectroradiometer (MODIS), while the LIC was obtained from the National Ice Center (NIC). WRF simulations for the Great Lakes region were performed at 10 km grid spacing for the cold season from November 2003 through February 2008. Initial and lateral boundary conditions were provided by the North American Regional Reanalysis (NARR). Experiments were carried out to compare winter precipitation simulations with and without the integration of the satellite data. Results show that integration with MODIS LST and NIC LIC significantly improves simulation of lake-effect precipitation over the Great Lakes region by reduced latent heat and sensible heat fluxes. A composite analysis of lake-effect precipitation events further reveals that more accurately depicted low-level stability and vertical moisture transport forced by the observation-based LST and LIC contribute to the improved simulation of lake-effect precipitation.

1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Model, Data Sources, and Methodology
  5. 3. Results
  6. 4. Conclusions and Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information

[2] The Great Lakes exert significant influence on weather and climate in the region, especially on the downwind shores during the cold season. From late fall to winter, when arctic air masses sweep down, considerable temperature differences between the water surface and the overlying air often trigger lake-effect precipitation. This well-known lake effect of snowstorms enhances annual precipitation by as much as 200% over that in nearby areas without the lake effect influence [Scott and Huff, 1996].

[3] Past studies have investigated a broad range of aspects of lake-effect precipitation over the Great Lakes, including climatology [Norton and Bolsenga, 1993; Ellis and Leathers, 1996] and detailed physical processes of individual events [Braham and Kelly, 1982; Hjelmfelt, 1990; Sousounis and Fritsch, 1994; Ballentine et al., 1998; Kristovich and Laird, 1998; Sousounis and Mann, 2000; Liu and Moore, 2004]. Modeling investigation by Lavoie [1972] suggested that the air temperature difference between the lake surface and the 850-hPa level is the most important factor in triggering lake-effect precipitation. Wilson [1977] pointed out that when the observed 850-hPa level temperature is 7°C colder than that of the lake surface, downwind precipitation increased significantly. Ice cover also has an impact on precipitation through modifying surface evaporation and stability in the lower atmosphere, weakening the lake-effect precipitation. Based on model results, Niziol et al. [1995] found that lake ice cover (LIC) greatly reduces surface heat and moisture fluxes. By comparing observations with the Colorado State University mesoscale model [Pielke, 1974], Laird and Kristovich [2004] found that simulations of lake-effect precipitation are slightly improved when realistic LIC is included in the model. Based on aircraft measurements, Gerbush et al. [2008] found that LIC results in a reduction in both surface sensible and latent heat fluxes, and the reduction significantly influences the development of lake-effect snowstorms [Cordeira and Laird, 2008].

[4] Despite these earlier efforts, the extent to which the combined contribution of LIC and lake surface temperature (LST) to simulations of lake-effect precipitation has not been sufficiently studied. In particular, the impact of realistic LST and LIC conditions on simulations of lake-effect precipitation requires further analysis. Compared to the Moderate Resolution Imaging Spectroradiometer (MODIS) observed LST, positive temperature differences (warm biases) have been found in the LSTs from reanalysis data such as the North American Regional Reanalysis (NARR) [Mesinger et al., 2006] (discussed later and shown in Figure 2). Furthermore, LIC is unavailable from the NARR data, and the LIC data from other reanalyses [Kalnay et al., 1996; Kanamitsu et al., 2002] cover only the oceans and not lakes (e.g., the Great Lakes). Such discrepancies in LST and LIC have a strong potential to negatively impact simulated precipitation related to lake processes.

[5] In this study, we used the Weather Research and Forecasting (WRF) [Skamarock et al., 2008] model version 3.2, developed by the National Center for Atmospheric Research, to simulate lake-effect precipitation over the Great Lakes. Our intention was to explore the impact of remote sensing LST and LIC on precipitation simulations over the Great Lakes region through the WRF model. The physical processes during lake-effect precipitation events were examined as well. Effects of cumulus convection and microphysics schemes on modeling lake-effect precipitation have been explored by Theeuwes et al. [2010] and are not our focus. The paper is arranged as follows: section 2 describes the model, data sets, experiment design, and methodology, section 3 presents the results, and section 4 provides conclusions.

2. Model, Data Sources, and Methodology

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Model, Data Sources, and Methodology
  5. 3. Results
  6. 4. Conclusions and Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information

2.1. Model and Data

[6] In this study we used the coupled WRF version 3.2 and the Community Land Model version 3.5 (CLM3.5) [Jin et al., 2010] for the proposed simulations. This version of WRF presets the water body freezing point at 271.4 K when the fractional ice option is employed; this temperature setting can be used for treating only saline water such as in oceans. This setting therefore is not suitable for the fresh water of the Great Lakes. Instead, the freezing point in the Great Lakes was adjusted to 273.16 K to reflect reality.

[7] We selected the fractional ice cover option (added into WRF since version 3.1) for the simulations, since this option treats the model grid cells with an ice fraction between 0% and 100%. For a model grid cell in the lake, variables are averaged over the ice-covered and open water (i.e., without ice cover) fractions [Avissar and Pielke, 1989; Vihma, 1995]:

  • display math

where x is a quantity, and the subscripts i and w refer to the ice and open water components within a model grid cell, respectively. Here, 0 ≤ LIC ≤ 1.0.

[8] The observed LSTs used in this study were obtained from MODIS. The MODIS LST product is an 8-daily composite, including daytime and nighttime, configured onto a 0.05° (∼5.6 km) latitude/longitude grid [Wan et al., 2002, 2004; Coll et al., 2005; Hook et al., 2007]. Missing values for the Great Lakes due to clouds were replaced with values derived from solving the Poisson's equation via relaxation [Evans, 1998]. A simple linear method was employed to interpolate the 8-daily MODIS daytime and nighttime data to 6-hourly intervals for the lake surface boundary conditions of WRF. The gridded ice analysis (i.e., LIC) obtained from the National Ice Center (NIC) [Fetterer and Fowler, 2006] has a resolution of 2.5 km from December 2003 to February 2007 and of 1.8 km from December 2007 to February 2008. This LIC data set is a composite of the measurements from in situ, ship, and satellite observations as well as from model output. The uncertainty in ice concentration of this LIC data set is estimated to be between 5% and 10%, according to Partington et al. [2003]. The LIC data have multiday intervals (3 or 4 days), which were interpolated into 6-hourly data for use in WRF.

[9] Because the MODIS LST and NIC LIC data sets come from different sources, to maintain data consistency the following correction scheme between these two data sets was built:

  • display math
  • display math

[10] In addition, when ice and water coexisted within a model grid cell (0 < LIC < 1.0), the heat fluxes were calculated with the water and ice surface temperatures separately. In this case, the LST for the grid cell remained the same, but the water and ice surface temperatures (Tw and Ti, respectively) were calculated based on the following equations:

  • display math
  • display math
  • display math

[11] For these three equations LIC is always greater than 0 and less than 1.0, while in all the other cases (i.e., either pure water or pure ice) Tw or Ti is equal to LST.

[12] For observed precipitation, we used the University of Delaware monthly global gridded data set [Legates and Willmott, 1990] available from 1950 to 2008 at a 0.5° latitude × 0.5° longitude resolution. The 32-km NARR data were used for the lateral boundary and initial conditions for WRF (except for LST and LIC over the Great Lakes). The wind fields of the NARR data were treated as observations, based on the NARR's previous evaluation [Mesinger et al., 2006]. For model evaluation, surface and sounding observations from meteorological stations at Buffalo, New York were utilized (location indicated in Figure 1 with a black dot).

image

Figure 1. Domain of the WRF simulations. Boxes 1, 2, 3, 4, and 5 over the Great Lakes region represent the lake-effect regions. The black dot is the location of Buffalo, New York, at and near which there are one sounding station and two additional surface stations.

Download figure to PowerPoint

2.2. Model Experiments

[13] Three WRF experiments were conducted for the Great Lakes region, forced respectively with three sets of surface boundary conditions: (1) MODIS LST and NIC LIC; (2) NARR LST and NIC LIC; (3) MODIS LST assuming lakes with open water (i.e., without ice cover). Hereafter, we name these experiments as (1) MLST_Ice, (2) NLST_Ice, and (3) MLST_NoIce (in which M denotes MODIS and N denotes NARR; see Table 1). The MLST_Ice and NLST_Ice experiments were designed to aid in understanding the impact of LST on lake-effect precipitation, as well as to evaluate the NARR LST that has been broadly used in weather and climate modeling for the Great Lakes region. The comparison between the MLST_Ice and MLST_NoIce experiments serves to further our understanding of the role of LIC on lake-effect precipitation.

Table 1. Configurations of the WRF Experiments MLST_Ice, NLST_Ice, and MLST_NoIce
 Experiments
MLST_IceNLST_IceMLST_NoIce
Lake Surface Temperature (LST)MODIS LSTNARR LSTMODIS LST
Ice ConcentrationNICNICOpen water

[14] The choices of microphysics schemes (MPSs) and cumulus parameterization schemes (CPSs) in WRF were determined according to the precipitation root mean square error (RMSE) over the entire simulation domain. WRF version 3.2 includes 5 CPSs and 12 MPSs. The first set of WRF tests were performed for December 2003 with all 5 CPSs with an arbitrarily selected MPS, the Morrison scheme [Morrison et al., 2009]. The results show that the RMSE of precipitation ranges from 20.1 to 20.7 mm/month between the 5 CPSs, and the Grell-Devenyi ensemble CPS [Grell and Dévényi, 2002] generates the lowest RMSE (20.1 mm/month) and is selected for the rest of simulations. These tests also show that the RMSE range is very minor, implying that convection is not a dominant process that controls precipitation over the Great Lakes during the winter. The second set of WRF tests were performed also for December 2003 with all 12 MPSs with the Grell-Devenyi CPS, where the resulting RMSE of precipitation ranges from 19.6 to 26.7 mm/month. The Thompson MPS [Thompson et al., 2008] produces the lowest RMSE (19.6 mm/month). In the end, with all the above 17 WRF tests, the Grell-Devenyi CPS and Thompson MPS are chosen for exploring lake-effect precipitation over the Great Lakes. Other key physics schemes used in the simulations are: the CLM3.5 land surface model, the Bougeault and Lacarrère planetary boundary layer (PBL) scheme [Bougeault and Lacarrère, 1989], and the Dudhia shortwave radiation [Dudhia, 1989] plus the Rapid Radiative Transfer Model scheme [Mlawer et al., 1997].

[15] The simulation domain is centered at 45.2°N, 85°W with a 10 km horizontal grid spacing (Figure 1). The grid points of dimension are 190 × 150, with 35 vertical layers topped at the 100-hPa level. The simulations cover the cold season from 15 November through 28 February (29 for leap years) for the period of 2003–2008, reinitialized each year at 0000 UTC on 15 November. The last 17 days of November were treated as spin-up and discarded. We chose five geographical regions encompassing the downwind shores of lake-effect areas, outlined by the five boxes in Figure 1, where significant lake-effect precipitation often occurs [Norton and Bolsenga, 1993].

2.3. Composite Analysis

[16] Typical lake-effect events at Buffalo were selected for a composite analysis with the following method. First, hourly observations at Buffalo were used to identify precipitation events that lasted at least 6 h, regardless of magnitude. Then, radar reflectivity images from the National Climatic Data Center (NCDC) were used to select lake-effect events where the reflectivity developed over Lake Erie and Lake Ontario and onto the downwind shores but did not appear to be associated with large-scale synoptic conditions. This step was derived from the lake-effect radar morphology developed by Liu and Moore [2004]. Once a typical lake-effect event over Buffalo was identified, its life cycle was divided into two stages: (1) the initial stage as the first time precipitation was recorded at the Buffalo station, and (2) the demise stage as the last time precipitation was recorded during the event. These events and their total precipitation are listed in Table 2.

Table 2. The 11 Lake-Effect Events at Buffalo, New York, During the Five Winters (December, January, February) From 2003 to 2008a
Initial StageDemise StageTotal Daily
(UTC)(UTC)Precipitation (mm)
  • a

    The initial and demise stages indicate the first time and last time when precipitation was recoded at the Buffalo station (the black dot in Figure 1) during the event. Dates are given as yyyy/mm/dd.

2003/12/202003/12/201.5
1:5016:05 
2004/01/312004/01/313.8
1:4013:25 
2004/12/242004/12/2510.7
17:304:50 
2005/12/052005/12/061.0
14:4512:50 
2005/12/212005/12/212.5
10:5016:50 
2006/12/72006/12/083.0
23:5010:50 
2006/12/272006/12/270.5
1:5012:50 
2007/02/032007/02/045.6
10:508:50 
2007/02/062007/02/071.5
14:1502:40 
2007/02/102007/02/102
0:5011:50 
2007/02/222007/02/231.0
23:5016:50 

3. Results

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Model, Data Sources, and Methodology
  5. 3. Results
  6. 4. Conclusions and Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information

3.1. The Role of LST

[17] Past studies have found that LST has a considerable effect on the development of lake-effect snowstorms [Hjelmfelt, 1990; Kristovich and Laird, 1998]. For instance, numerical experiments conducted over the Great Salt Lake [Onton and Steenburgh, 2001] indicated that a 2°C increase (decrease) in LST results in 32% more (24% less) lake-effect precipitation. Here, we first examined LST averaged over the individual lakes of the Great Lakes to delineate the differences between the two sets of LST boundary conditions (prior to the simulations). The grids of the individual lakes were determined by land use types provided by the United States Geological Survey (USGS). Figure 2 depicts the monthly mean differences between NARR LST and MODIS LST from December 2003 to February 2008. In general, over these winter months, NARR LST tends to be warmer than MODIS LST by about 3°C, with an exception in February 2004 when NARR LST is cooler than MODIS LST over Lake Erie (Figure 2d). The maximum difference between these two LST data sets reached as high as 8.6°C in December 2005 over Lake Erie (Figure 2d).

image

Figure 2. The differences in the monthly LST between MODIS and NARR (NARR minus MODIS; °C) for the Great Lakes during the winters (December, January, February) from 2003 to 2008. The grids of each lake were determined by land use type data from the USGS. (a) Lake Superior, (b) Lake Michigan, (c) Lake Huron, (d) Lake Erie, and (e) Lake Ontario.

Download figure to PowerPoint

[18] During winter, LSTs are normally (but not necessarily) warmer than the air temperature near the surface. An intuitive expectation is that a higher LST creates greater instability when the overlying air mass is colder, which may lead to higher precipitation. To examine this, monthly precipitation simulated from MLST_Ice and NLST_Ice, as well as observations, is shown in Figure 3 for the 5-year average. In terms of the geographical distribution of precipitation, the two simulations agree with each other, but their precipitation magnitudes are noticeably different (Black hatchings are significant at 90% confidence level; Figures 3j, 3k, and 3l). For example, in December over central Lake Superior, the precipitation difference reached a maximum of 42 mm/month (Figure 3j). Meanwhile, when compared to observations, both simulations show a systematic bias of over-predicted precipitation, such as on the southern shores of Lake Superior (Figures 3a, 3b, 3d, 3e, 3g, and 3h). However, around Lakes Huron, Erie, and Ontario, the MLST_Ice experiment produces less precipitation than NLST_Ice, making it closer to the observations. In eastern Lake Ontario, the simulated precipitation is distinctly greater than the observations in both experiments. From the terrain map of WRF (not shown), the elevation of Lake Ontario is 75 m while the nearest mountain to the east of the lake is 500 m. Therefore, the overestimated precipitation in eastern Lake Ontario may also be attributed to orographic lifting that is too strong, as was documented in Niziol et al. [1995]. Such an orographic effect has also been found over Lake Michigan [Hjelmfelt, 1992]. Regarding the spatial pattern of the precipitation differences, Figure 4 shows the spatial correlation between the simulations and the observations. The correlation indicates that while both simulations capture the major pattern of observed precipitation, MLST_Ice appears to perform slightly yet consistently better in terms of spatial agreement.

image

Figure 3. Monthly precipitation (mm/month) averaged over the period of 2003–2008. Observations: (a) December, (b) January, and (c) February; MLST_Ice: (d) December, (e) January, and (f) February; NLST_Ice: (g) December, (h) January, and (i) February; NLST_Ice minus MLST_Ice: (j) December, (k) January, and (l) February. Black hatchings indicate the 90% confidence level (CL) based upon Student's t-test.

Download figure to PowerPoint

image

Figure 4. Spatial correlation coefficients of monthly precipitation between simulations and observations during the winters (December, January, February) from 2003 to 2008. (Red line represents the correlation coefficients between MLST_Ice and observations; blue line represents the correlation coefficients between NLST_Ice and observations.) The results are all significant at the 95% confidence level (CL) based upon Student's t-test.

Download figure to PowerPoint

[19] To examine quantitatively whether MLST_Ice improves the precipitation simulation as compared to NLST_Ice, we evaluated their mean error (ME) and RMSE:

  • display math
  • display math

where N is the number of grid points for evaluation, and P and O represent the model results and observations, respectively. To facilitate comparisons, we interpolated model grids to those of the observed precipitation. Both ME and RMSE (Figure 5) were computed for (a) the entire domain and (b) the lake-effect areas combined, calculated grid by grid between simulations and observations with monthly data. Here, ME indicates bias in the monthly magnitude of precipitation (perfect = 0), while RMSE quantifies bias in the precipitation variation (perfect = 0). Although ME fluctuates month to month, the mean values (red dots) suggest that MLST_Ice produces consistently less error in the precipitation magnitude than NLST_Ice, especially over the lake-effect domains (Figure 5b). There, the maximum difference between the two experiments occurs in December, with a mean value of 6.2 mm/month in MLST_Ice and 16.4 mm/month in NLST_Ice. Likewise, RMSE suggests that the precipitation of MLST_Ice is more realistic than that of NLST_Ice, as is reflected in their mean values (Figures 5c and 5d); this result corresponds to the lower LST as shown in Figure 2. These experiments are discussed further in section 3.3.

image

Figure 5. The box and whisker plots of the ME and RMSE of the simulated monthly precipitation (mm/month) for the winters (December, January, February) for the period of 2003–2008. (a and c) For the entire domain. (b and d) For the lake-effect domains. The thick horizontal line in each box, the top and bottom edges of the box, and the upper and lower whiskers represent the ME (Figures 5a and 5b) and RMSE (Figures 5c and 5d) for each of the five years. The upper (lower) whisker shows the maximum (minimum) value, and the red point represents the mean value.

Download figure to PowerPoint

3.2. The Role of Ice Cover

[20] It is known that lake ice cover weakens precipitation by prohibiting upward heat and moisture fluxes from the lake surface. Here, the experiments of MLST_Ice and MLST_NoIce were performed to examine the effect of LIC inputs on lake-effect precipitation simulation. For LST, both experiments used the surface boundary conditions of MODIS LST, owing to its better simulation results as shown above.

[21] We first examined the distribution of 5-year monthly mean NIC LIC over the Great Lakes (Figures 6a, 6c, and 6e). In general, ice cover gradually increases from December through February, but the development of ice cover over individual lake surfaces is different. For instance, the ice fraction over most of Lake Erie is more than 0.7 during February, while most of Lake Ontario is ice-free (Figure 6e). To detect the impact of ice cover on precipitation, the monthly precipitation difference in percentage between MLST_NoIce and MLST_Ice (i.e., MLST_Ice minus MLST_NoIce, then divided by MLST_Ice) is shown in Figures 6b, 6d, and 6f. Due to the small sample size of 5 winter months, however, statistical significance of the precipitation difference is generally low (i.e., less than the 90% level) and the results here should be considered suggestive rather than conclusive. It appears that MLST_Ice generally reduces precipitation, while the decrease seems proportional to the extent of ice cover over the lake surfaces. The maximum effect of LIC occurs along the downwind shores, such as southwestern Lake Superior, northern Lake Huron, and southeastern Lake Erie, causing a precipitation reduction as large as 46% (Figure 6f). However, positive precipitation differences between MLST_NoIce and MLST_Ice are observed around eastern Lake Superior and Lake Ontario. Sections of February mean vertical velocity across Lake Superior and Lake Erie (Figure 7a; orange and black lines) depict a branch of increased ascending motion (or reduced descent) over eastern Lake Superior (Figure 7b). Compared to Lake Erie (Figure 7c), the overall descent corresponds well with the reduced precipitation. These features suggest that ice cover may modify local and mesoscale circulations over the Great Lakes, complicating the precipitation generation process in MLST_Ice.

image

Figure 6. (a, c, and e) Monthly mean of the NIC ice fraction and (b, d, and f) monthly precipitation (%) difference between MLST_Ice and MLST_NoIce [(MLST_Ice -MLST_NoIce)/MLST_Ice] for the winters from 2003 to 2008.

Download figure to PowerPoint

image

Figure 7. Cross sections of monthly mean vertical velocity (February) over (b) Lake Superior and (c) Lake Erie. (a) Orange and black lines on Great Lakes map indicate the positions of cross sections for Lake Superior and Lake Erie, respectively. The lower boundary in Figures 7b and 7c is the terrain.

Download figure to PowerPoint

[22] To quantify the LIC impact on precipitation, the MEs of precipitation in the entire Great Lakes domain (Figure 8a) and the lake-effect areas (Figure 8b) are shown. From December to February, the discrepancy in the mean values (red point) between MLST_Ice and MLST_NoIce gradually increases in both domains as the season progresses and ice cover develops. The maximum difference occurs over lake-effect areas in February (Figure 8b), which shows an ME of 5.9 mm/month in MLST_Ice and 9.5 mm/month in MLST_NoIce. The presence of ice cover on the lakes could change surface albedo [Ingram et al., 1989; Curry et al., 1995] and influence energy exchange between the water surface and the overlying air – i.e., evaporation when there is no ice over the lake surface and sublimation when there is ice cover. In the WRF model, the default albedo for water is 0.08, and for ice it is 0.98. Because we used the fractional ice option in WRF, ice cover varies between 0% and 100%, and this results in a corresponding change in albedo. For example, if the ice fraction of a model grid cell is 0.7 over Lake Erie, then according to equation (1) this grid cell would have an albedo of 0.71 in MLST_Ice; this change in albedo would absorb much less solar radiation as compared to the open water setting in MLST_NoIce (where the albedo is 0.08). Ice cover also changes the method of energy transfer from the lake surface – i.e., evaporation and sublimation, the latter being much slower than the former due to the lower temperature and larger latent heat of vaporization. Thus, sublimation makes transport of water vapor into the atmosphere more difficult; this may be linked to the overall precipitation decrease revealed in Figure 6.

image

Figure 8. Same as Figure 5 but for the MLST_Ice and MLST_NoIce experiments.

Download figure to PowerPoint

[23] The differences in simulated latent heat flux and sensible heat flux between MLST_Ice and MLST_NoIce in February are illustrated in Figure 9; their average difference in latent heat flux is −3.3 W/m2 with a maximum difference of −106.8 W/m2 (Figure 9a). Clearly, ice cover decreases water vapor flux to the atmosphere, and this lends support to the reduced precipitation in MLST_Ice. With ice cover, the most substantial decreases in latent heat flux occur near the lake shores, consistent with the spatial distributions of ice cover (Figure 6e). On the other hand, decreased sensible heat flux with more lake ice cover would result in less energetic buoyant mixing in the lake-effect convective boundary layer. This subsequently decreases buoyant convection that contributes to the shallow convective boundary layer, suppressing the generation of precipitation (Figure 9b). The most significant difference in the latent heat flux and sensible heat flux occurs over Lake Erie, where the ice fraction is the largest. This further emphasizes the role of ice cover in affecting water vapor flux into the atmosphere and subsequently, the precipitation.

image

Figure 9. Monthly difference in (a) the latent heat flux (W m−2) and (b) the sensible heat flux (W m−2) between MLST_NoIce and MLST_Ice (MLST_Ice minus MLST_NoIce) in February averaged over 2003–2008.

Download figure to PowerPoint

3.3. Composite Analysis of Lake-Effect Events

[24] In view of the discrepancy in precipitation between MLST_Ice and NLST_Ice, we next examined the composite analysis of typical lake-effect precipitation events. Eleven such precipitation events were identified in section 2.4 and summarized in Table 2. The 850-hPa wind, 10 m wind, 2 m temperature, and precipitation differences (NLST_Ice minus MLST_Ice) are shown in Figure 10 to examine the dynamic cause of the precipitation difference. All variables were averaged within the duration of each event (from initial stage to demise stage). As shown in Figure 10a, the prevailing winds during the lake-effect precipitation events are mainly northwesterlies, consistent with the synoptic setting that creates lake effect. Over and downwind of the Great Lakes, the 2 m temperature of NLST_Ice is warmer than MLST_Ice (Figure 10b); this warming is distributed downstream and signifies a stronger low-level instability in NLST_Ice. Meanwhile, surface winds converge toward the warming regions (Figure 10c), such as southeastern Lake Superior, southeastern Lake Huron, southeastern Lake Michigan, eastern Lake Ontario, and central Lake Erie, based on the distribution of the convergence. The coupling of warming with convergence suggests a stronger lift being produced in NLST_Ice. Moreover, those convergence centers are consistently distributed on the downwind shores, hence supporting the increase in precipitation as shown in Figure 10d. In summary, the change in LST conditions appears to affect lake-effect precipitation through a combination of processes both thermodynamic (i.e., stability over and downwind of the lakes) and dynamic (i.e., downwind convergence), in addition to the mere change in water vapor fluxes.

image

Figure 10. Composite of (a) NARR 850 hPa wind field, (b) 2 m temperature difference (NLST_Ice minus MLST_Ice), (c) difference in 10 m wind and divergence (NLST_Ice minus MLST_Ice), and (d) difference in precipitation and 10 m wind (NLST_Ice minus MLST_Ice) for the 11 lake-effect events. The black dot shows the location of Buffalo, New York. Black hatchings indicate the 90% confidence level (CL) based upon Student's t-test.

Download figure to PowerPoint

[25] As a further examination of the atmospheric conditions, we plotted in Figure 11 the vertical profiles of temperature and water vapor mixing ratios in Buffalo. Although the precipitation difference is not remarkable in Buffalo, the vertical profile did reveal a noticeable difference between NLST_Ice and MLST_Ice (i.e., green and blue lines). We note that the twice-daily sounding observations (0000 UTC and 1200 UTC) do not align completely with the initial and demise stages of the lake-effect precipitation events; however, these observations are the only data available to represent the atmospheric sounding. In general, both MLST_Ice and NLST_Ice performed reasonably well in simulating the temperature and moisture profiles for the PBL, the observed height of which often ranges from 1 to 3 km in the Great Lakes [Chang and Braham, 1991; Kristovich, 1993]. However, the results from MLST_Ice are apparently closer to the observations. From the surface to the height of 1.6 km, where the two runs differ the most, the mean absolute error (MAE) of temperature is 1.02°C (0.66°C) in MLST_Ice for the initial stage (demise stage), whereas it is 1.87°C (1.48°C) in NLST_Ice (Table 3). Within the same layer of the atmosphere, the MAE of the mixing ratios is 0.10 g/kg (0.06 g/kg) in MLST_Ice for the initial stage (demise stage), while it is 0.18g/kg (0.10 g/kg) in NLST_Ice (Table 3). From the surface to the 1.6-km level, the average temperature and mixing ratio in MLST_Ice also show a corresponding improvement over NLST_Ice. In other words, MLST_Ice produced a colder and drier (i.e., more stable) PBL than NLST_Ice. Noteworthy are the significant biases in the upper atmosphere (above 1.6 km), where MLST_Ice and NLST_Ice do not differ much from each other. Apparently, these larger-scale biases do not contribute to the precipitation difference between the two runs.

image

Figure 11. Comparison of observations (purple dot) and simulations (blue and green lines) of the vertical temperature (°C) and water vapor mixing ratio (g/kg) composites of the 11 lake-effect events for the initial and demise stages at the Buffalo sounding station: (a and b) temperature; (c and d) water vapor mixing ratio.

Download figure to PowerPoint

Table 3. MAEs of the Vertical Temperature and Water Vapor Mixing Ratios From the Surface to a Height of 1.6 km for the 11 Selected Lake-Effect Event Simulations
 Temperature (°C)Water Vapor Mixing Ratio (g/kg)
MLST_Ice (Initial Stage)1.020.10
NLST_Ice (Initial Stage)1.870.18
MLST_Ice (Demise Stage)0.660.06
NLST_Ice (Demise Stage)1.480.10

4. Conclusions and Discussion

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Model, Data Sources, and Methodology
  5. 3. Results
  6. 4. Conclusions and Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information

[26] In this study the remotely sensed MODIS LST and NIC ice cover were integrated with the WRF model to examine their impact on the simulation of lake-effect precipitation. It is found that the simulated lake-effect precipitation is improved when forced by MODIS LST and NIC ice cover (i.e., realistic LST and ice conditions). Improvements are observed over the entire Great Lakes region and more significantly over the lake-effect areas. Ice cover also plays an important role in the precipitation magnitude. As compared to MLST_NoIce, MLST_Ice reveals a general decrease in latent heat flux and precipitation. Moreover, composite analysis of the lake-effect precipitation events suggests that more realistic low-level stability and vertical moisture transport lead to the improved precipitation simulation in MLST_Ice. These results therefore indicate that simulations of lake-effect precipitation forced by NARR would be biased due to the lack of ice cover data integrated in the model.

[27] Caution should be exercised when interpreting the simulation results presented here. First, the MODIS data is interpolated into 6-hourly intervals, a process that may not represent actual conditions completely, especially when temperature extremes occur. For example, a mass of cold arctic air can pass through the Great Lakes fairly quickly. It is impossible for data averaged over 8 days to reproduce such a variation in LST. As a result, the simulation results here are more representative for long-term climatology than they are for daily weather. Second, it is known that the diurnal variation in the lake-air temperature difference is correlated with lake-effect precipitation [e.g., Kristovich and Spinar, 2005]. Therefore, to reproduce LST more accurately (diurnal variation), a lake model may be a good option for simulating lake-effect precipitation and regional climate.

Acknowledgments

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Model, Data Sources, and Methodology
  5. 3. Results
  6. 4. Conclusions and Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information

[28] This work was supported by the Utah Agricultural Experiment Station, the NOAA MAPP NA090AR4310195 grant, and the EPA RD83418601 grant.

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Model, Data Sources, and Methodology
  5. 3. Results
  6. 4. Conclusions and Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Model, Data Sources, and Methodology
  5. 3. Results
  6. 4. Conclusions and Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information
FilenameFormatSizeDescription
jgrd17740-sup-0001-t01.txtplain text document0KTab-delimited Table 1.
jgrd17740-sup-0002-t02.txtplain text document1KTab-delimited Table 2.
jgrd17740-sup-0003-t03.txtplain text document0KTab-delimited Table 3.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.