1. Introduction
Precipitation is a principal component in the land and atmospheric moisture budgets, regulating the water availability and quality in the Great Lakes watersheds and various processes including surface runoff, lake levels, soil moisture, and groundwater reserves. It is affected both by local feedback mechanisms—that is, regional evapotranspiration resulting in recycled precipitation (Dominguez et al. 2006; Lamb et al. 2012)—and large-scale processes in the form of atmospheric moisture advection into the region (Zangvil et al. 2004; Banacos and Schultz 2005; Lélé and Leslie, 2016). Diagnostic analysis of the moisture budget components can explain the large-scale atmospheric controls and regional mechanisms behind precipitation patterns (Li et al. 2013; Wang et al. 2017, Dagan et al. 2019) and can elucidate moisture transport processes (Zangvil et al. 2001; Trenberth et al. 2011), land–atmosphere interactions (Zangvil et al. 2004; Lamb et al. 2012), regional climate (Li et al. 2010), and hydrological extremes (Yang et al. 2014; Şahin et al. 2015). Variations in any moisture budget component can result in fluctuations in precipitation seasonality and distribution that can significantly alter the mean state of the regional hydroclimate (Seager and Henderson 2013).
Past studies have used moisture budget analyses to understand precipitation variability in different regions across the globe. For example, Şahin et al. (2015) used reanalysis data to investigate the moisture flux characteristics over the Mediterranean Basin and found that the large-scale dynamical conditions over the Atlantic Ocean (i.e., the Azores high and the Icelandic low) control the wet or dry conditions within the basin. Li et al. (2013) used reanalysis datasets to link summer (June–August) precipitation intensification with the mean seasonal moisture transport in the southeastern United States. They concluded that dynamic processes driven by atmospheric circulation regulate more than 90% of the variation in moisture transport and resulting precipitation, whereas thermodynamic processes (specific humidity) have a subdued role within that region. Similarly, Zangvil et al. (2001) studied the correlation between moisture budget components and summer precipitation (May–August) over the midwestern United States and found that the changes in moisture flux divergence (MFD) are the dominant contributor to the variability in precipitation at daily, monthly, and seasonal time scales. Within the Great Lakes region, Li et al. (2010) studied annual trends of moisture budget components using the NOAA National Centers for Environmental Prediction North American Regional Reanalysis (NCEP NARR; Mesinger et al. 2006) and also highlighted the role of the large-scale moisture transport in the variation of precipitation, noting that better depiction of moisture transport processes in weather and climate models is needed to adequately simulate the hydrological cycle in the region.
In the Great Lakes region, precipitation has a well-defined seasonality with a summer maximum and drier conditions in the remaining seasons (Basile et al. 2017). Various mechanisms generate this pattern including large-scale systems, convective storms, and lake-effect precipitation (Grover and Sousounis, 2002), with the dominant mechanism varying by season. Synoptic-scale systems such as weather fronts and low pressure systems have the largest contribution to the annual precipitation magnitude, primarily due to the high frequency and intensity of these events (Grover and Sousounis 2002). Because of their immense volume and surface area, the Great Lakes also influence the hydroclimate and consequently the precipitation dynamics of the region (Li et al. 2010; Notaro et al. 2013; Fujisaki-Manome et al. 2017). The lake-induced feedback mechanisms are more prominent downwind of the lakes, i.e., generally in east to southeast direction, and their magnitude scales with the size of the lake (Scott and Huff 1996). Moreover, lake ice cover can significantly alter evaporation rates, influence precipitation magnitude and phase, and induce lake-effect snowstorms (Notaro et al. 2015).
Studies have found notable changes in precipitation over the Great Lakes region, often attributed to anthropogenically driven climate change (Hayhoe et al. 2010; Kim et al. 2015). These changes are projected to intensify in the future and can profoundly impact the ecology, wildlife, and water quality of the Great Lakes. However, despite its significance, precipitation remains difficult to simulate in climate models. For example, many CMIP5 models fail to capture the observed precipitation seasonal patterns and magnitudes in the Great Lakes region (Basile et al. 2017). Therefore, understanding the physical mechanisms controlling observed and modeled precipitation is crucial to improve its representation in climate models. Within this region, the mechanistic processes behind the month-to-month variations in precipitation require a comprehensive analysis to understand the shifts in its patterns.
The primary objective of this study is to understand the mechanisms that define the seasonality of precipitation in the Great Lakes region through an assessment of the atmospheric moisture budget. Because the Great Lakes have a prominent role in the regional water cycle, we also assess their role in modulating the moisture budget components. A secondary objective is to determine appropriate reanalysis products for hydroclimatic assessments in this region by evaluating their performance in representing the moisture budget. Analyzing precipitation using a single dataset can introduce inherent biases, therefore, we use five reanalysis products including European Centre for Medium-Range Weather Forecasts (ECMWF) interim reanalysis (ERA-Interim; Dee et al. 2011) and ERA5 (Hersbach et al. 2020), version 2 of the NASA Modern-Era Retrospective Analysis for Research and Applications (MERRA-2; Gelaro et al. 2017), NCEP Climate Forecast System Reanalysis (CFSR; Saha et al. 2010), and NCEP NARR (Mesinger et al. 2006).
2. Data and methods
The study domain over the Great Lakes region (94°–74°W and 40°–51°N) encompasses the hydrological basins of the five lakes: Superior, Michigan, Huron, Erie, and Ontario. Lake Superior is the deepest (mean depth of 148 m) and largest by volume of the five. Lakes Ontario and Michigan have a mean depth of 85–86 m followed by Huron (59 m), while Lake Erie is the smallest both in terms of volume and depth (19 m) (Assel et al. 2003). These physical properties, together with their geographical location, can create variations in the lake–atmosphere feedbacks among the five lakes. The region has moderate topographic relief with the Appalachian Mountains in the southeast, the St. Lawrence lowlands on the eastern edge of the domain, the Laurentian Plateau to the northeast and northwest, and the Hudson Bay lowlands in the north (Fig. 1).
Study domain and topographic features of the Laurentian Great Lakes region. The analysis is conducted over the red-outlined box (94°–74°W and 40°–51°N) encompassing the hydrological basins of the five Great Lakes.
Citation: Journal of Climate 34, 2; 10.1175/JCLI-D-19-0952.1
We use five reanalysis products at 6-hourly temporal resolution, including ERA5 (~31-km native resolution), ERA-Interim (~79 km), and NCEP–CFSR (~38 km), NASA MERRA-2 (~50 km), and NCEP NARR (~36 km). For consistency, we conduct the analysis on regridded data for the reanalyses at 0.5° × 0.5° rectilinear grid, except MERRA-2 for which the native resolution of 0.5° × 0.625° is used. The details of these datasets are provided in the online supplemental material (Table S1). We specifically focus on three reanalyses for concision: ERA-Interim, MERRA-2, and CFSR. The monthly climatology of the moisture budget components in ERA5 is very similar to ERA-Interim (Fig. 1; discussed in section 3a); however, Hersbach et al. (2020) notes the inability of ERA5 to simulate the correct surface temperatures over the Great Lakes, resulting in an anomalously strong annual lake surface temperature cycle, that can affect the latent and sensible heat fluxes and other meteorological fields. NARR directly assimilates precipitation observations (Shafran et al. 2004; Mesinger et al. 2006), which can introduce spatiotemporal inconsistencies in precipitation time series depending on observational data availability and gauge distribution.
We also use monthly precipitation data from three observation-based gridded datasets at 0.5° × 0.5° spatial resolution: University of East Anglia CRU TS4.03 (CRU) (Harris et al. 2014), University of Delaware Global Land Data v4.01 (UoD) (Willmott and Matsuura 2001), and DWD-WMO GPCC precipitation dataset v7.0 (GPCC) (Schneider et al. 2014). It is important to note that these datasets only employ land-based gauge measurements and therefore may not directly capture precipitation over the lake surfaces. Estimate of statistically significant differences in precipitation spatial patterns of the reanalyses is calculated from the CRU TS4.03 time series using Student’s t test (Decremer et al. 2014). The climatological time period of 1981–2010 is used for all analysis in this study. We classify December–March (DJFM) as the winter season when the mean temperature is approximately 267 K, April–May as spring (AM; ~282 K), June–September as summer (JJAS; ~291 K), and October–November as autumn (ON; ~278 K). The winter season in this region has a longer duration, with the mean monthly temperature for each of the four DJFM months below 273 K, and the spring and autumn transitory seasons have a shorter duration.
Seasonal climatology of the moisture budget components [Eq. (1)] averaged over the study domain in Fig. 1, including (a) total precipitation, (b) evaporation, (c) change in total precipitable water (dPW), (d) vertically integrated moisture flux divergence (VIMFD), and (e) the residual term. Observation-based precipitation datasets are shown as dashed lines in (a). The 95% confidence intervals for P, dPW, E, and VIMFD are also shown.
Citation: Journal of Climate 34, 2; 10.1175/JCLI-D-19-0952.1
An important consideration in the computation of the VIMFD term is the temporal resolution of the variables used, especially the wind vector (zonal u and meridional υ winds). Seager and Henderson (2013) discussed the errors introduced by the time resolution of the archived data of models, which is coarser than the model time step. To assess the effect of temporal resolution on VIMFD computations, we tested monthly, daily, and 6-hourly time series for ERA-Interim (Fig. S1 in the online supplemental material), and we observed substantial winter and spring bias in VIMFD when using the coarser time steps. At the monthly time scale, this bias is strong enough to change the sign of the term from convergence to divergence. For example, in the winter season the VIMFD is strongly convergent when using the 6-hourly data (DJFM mean of −1.29 mm day−1), which weakens for the daily data (−0.96 mm day−1), and is divergent for the monthly data (+0.32 mm day−1). We also note a negative summer bias in the coarser temporal datasets, with the intensity of summertime divergence decreasing as the temporal resolution becomes coarser from 6-hourly to monthly (e.g., for the month of August the VIMFD magnitudes are 0.61, 0.51, and 0.44 mm day−1 from fine to coarse resolutions). These differences are likely due to the changing direction of winds at finer spatiotemporal resolutions associated with subsynoptic and mesoscale weather systems. Averaging over longer periods may not capture the high-frequency meteorology and can thus considerably alter the spatial patterns and the sign of the divergence computations at individual grid cells. However, as we move to finer temporal resolutions (e.g., from 6-hourly to 3-hourly), the strength of these systems diminish and their overall contribution to moisture fluxes seasonal cycle is inconsequential. Furthermore, 6-hourly resolution is the highest available for all datasets and is therefore used in this study.
The residual term R of Eq. (1) arises as a result of various sources of errors that result in an artificial term that lacks physical meaning yet cannot be disregarded (Dominguez et al. 2006, Li et al. 2010). Coarse spatial and temporal discretization, segmenting the atmosphere into fewer vertical levels, physical parameterization and data assimilation limitations of the models, and methodological errors introduce biases in the moisture budget, resulting in nonclosure of the mass-balance of the system that is incorporated in the residual term.
The remaining two terms of Eq. (1) include the rates of evaporation E (also incorporating transpiration from vegetated surfaces) and total precipitation P. Precipitation in the models is represented by the resolved large-scale (microphysics) and subgrid convective parameterization schemes (Lopez 2007) and constitutes both solid and liquid phases. Past studies have shown that various processes and mechanisms can affect representation of precipitation in the reanalysis, including land surface processes such as evapotranspiration that feeds the recycled precipitation (Gottschalck et al. 2005; Yang et al. 2009; Lamb et al. 2012), and assimilation of observations (Saha et al. 2010; Zhang et al. 2012) that can violate the governing equations and result in hydrological imbalance (Bosilovich et al. 2015). For example, NARR assimilates hourly precipitation observations (Mesinger et al. 2006), which affects the simulation of precipitation as seen in the subsequent analysis. The reanalyses also tend to overestimate precipitation magnitudes as compared to observations (Bosilovich et al. 2008, 2011). Like precipitation, the parameterization of evapotranspiration in the models also contains biases and uncertainties that are due to limited observations and estimations using bulk flux formulas (Trenberth et al. 2011).
3. Results and discussion
a. Diagnosis of moisture budget components
1) Precipitation
The observed precipitation climatology of the Great Lakes region shows peak magnitudes in June–July, followed by a brief decline in August and a secondary peak in September (Fig. 2a). From the observational datasets, the winter (DJFM) conditions are the driest, constituting only around 23% of the annual precipitation, whereas the summer months of June to September (JJAS) represent more than 42% of the annual total. The shorter spring (AM) and autumn (ON) seasons have near-equal contribution of ~17% each to the annual magnitudes. This seasonality is somewhat represented in the reanalysis products, albeit with varying magnitudes. MERRA-2 has the highest magnitude among the reanalyses, especially during spring and early summer months, while NARR has the smallest overall magnitude, with the annual total approximately 10% less than observation-based estimates (788 mm for NARR as compared with 884 mm for CRU and GPCC), but otherwise captures the observed seasonal cycle. The spatial patterns of precipitation for NARR (Figs. 3q–t) show that the magnitudes directly over the lakes are relatively lower than other datasets. We also note drier conditions over the U.S.–Canada border, which is likely an artifact of precipitation observation assimilation. The distribution of ground observations, buoy data over the lakes, and precipitation datasets assimilated by NARR shows a sharp contrast in the density of observations between the United States and Canada, where the latter has a relatively sparse distribution of observations (Shafran et al. 2004). In MERRA-2, the spring months have higher precipitation over land toward the south, while in summer the northeastern region of the domain has significantly higher precipitation compared to other datasets (Figs. 3j,k). In general, precipitation has a high standard deviation because of strong interannual variability (~0.4–0.9 mm day−1 depending on the season; Fig. S2a in the online supplemental material), yet for MERRA-2 the standard deviation is particularly high, especially in the spring–early summer months (as high as 0.9 mm day−1 for June). The ERA5 and ERA-Interim seasonality is very similar but slightly diverges in magnitude from August onward, when ERA5 has higher precipitation until December.
Seasonal spatial patterns of total precipitation (mm day−1) averaged over the 1981–2010 period for (left) winter (DJFM), (left center) spring (AM), (right center) summer (JJAS), and (right) autumn (ON). (a)–(d) CRU TS v4.03 precipitation patterns for comparison with (e)–(h) ERA-Interim, (i)–(l) NASA MERRA-2, (m)–(p) NCEP CFSR, and (q)–(t) NCEP NARR. Stippling shows regions of statistically significant difference in reanalysis from CRU at 5% significance.
Citation: Journal of Climate 34, 2; 10.1175/JCLI-D-19-0952.1
One reanalysis that does not reproduce the observed precipitation pattern is NCEP CFSR, which has a bimodal behavior with a strong wet bias in the autumn, winter, and early spring months (nearly 40% higher than observed) and a dry bias in the late spring and summer precipitation. Moreover, the decline from June to August is greater in CFSR, with a decrease of nearly 28% as compared with 8% in the gridded observation datasets. Overall, CFSR suppresses precipitation during the warm season and enhances it in the remaining months. The spatial patterns for CFSR (Figs. 3m–p) show statistically significant differences from CRU in precipitation distribution throughout the domain in the winter and spring seasons, and in the northern part of the domain in autumn. The summer mean patterns, on the other hand, are not significantly different from CRU, despite strong differences in the mean monthly magnitudes (Fig. 2a).
2) Evaporation
Evaporation is at its maximum in the summer, ranging from 2.7 ± 0.08 for CFSR to 3.4 ± 0.07 mm day−1 for NARR (mean ± 95% confidence interval; Fig. 2b), driven by evapotranspiration over the land surface (Figs. 4c,g,k,o). The higher heat capacity of the lakes results in a delayed heating and cooling response in the summer and winter seasons, respectively, as compared with the surrounding land and the overlying air (Scott and Huff 1996). In winter the three larger lakes (Michigan, Superior, and Huron) have higher evaporation rates (Figs. 4a,e,i,m), whereas in spring the evaporation rate considerably drops over the lakes due to a delayed cooling response (Figs. 4b,f,j,n). This continues into the summer when the lake-surface evaporation is smaller in magnitude than the land component of the domain. Overall, the lake evaporation contrasts the surrounding land by providing extra moisture during the dry winter months and reducing the evaporation magnitude in the summer months.
Seasonal (as labeled) spatial patterns of the evaporation rate (mm day−1) averaged over the 1981–2010 period for (a)–(d) ERA-Interim, (e)–(h) MERRA-2, (i)–(l) CFSR, and (m)–(p) NARR.
Citation: Journal of Climate 34, 2; 10.1175/JCLI-D-19-0952.1
Similar to precipitation, we also observe a slight skewness in the seasonal cycle of the CFSR evaporation (Fig. 2b), where the average wintertime rate is the highest for CFSR at 1.03 ± 0.03 mm day−1 as compared with 0.68 ± 0.03 and 0.80 ± 0.03 mm day−1 for ERA-Interim and MERRA-2, respectively. Over land, CFSR has slightly greater evaporation in winter (Fig. 4i), whereas during the summer months it has lower magnitudes as compared with the other datasets (Fig. 4k). NARR, on the other hand, has the highest evaporation rates in the summer months, specifically June–August, with the average over the three months of 3.73 ± 0.08 mm day−1 as compared with 3.58 ± 0.07 and 3.29 ± 0.05 mm day−1 for MERRA-2 and ERA-Interim, respectively (Fig. 2b). Much of this increase is due to higher rates over land toward the south of the domain (Fig. 4o). Similar to the precipitation patterns, we note decreased evaporation over the U.S.–Canada border, which is likely associated with fewer surface observations and/or different datasets assimilated for the Canada region (Shafran et al. 2004).
3) Total precipitable water
The change in total precipitable water (dPW) increases the atmospheric moisture content from February to July (Fig. 2c). Between July and August, there is a sharp drop in the atmospheric moisture content by approximately 0.34 ± 0.10 mm day−1, which is ubiquitous in all reanalyses and corresponds with the precipitation decline. The change in total precipitable water is highest in May, and, while there is no direct correlation with other moisture components, the seasonal maximum may be due to combined effects of moderate evaporation and moderate to high moisture flux convergence [section 3a(4)]. We note two deviations in the ERA5 climatology where the month of January is relatively drier as compared with other datasets whereas the month of October is less dry. The standard deviation during these two months is also the highest for ERA5 (~0.53 mm day−1 in each month), whereas for ERA-Interim the standard deviation is 0.15 and 0.24 mm day−1 for January and October, respectively (Fig. S2c in the online supplemental material). All products have a remarkably similar seasonal cycle despite no commonality in other moisture budget quantities. We also note that our computed dPW compares well to the change in the “total column water vapor” variable from ERA-Interim that is computed on model levels (dashed line in Fig. 2c).
4) Vertically integrated moisture flux divergence
(i) Monthly climatology
The domain-averaged VIMFD monthly climatology (Fig. 2d) shows convergence from September to May and divergence in the summer months. However, the strength and extent of summer divergence vary among the reanalyses. For CFSR and MERRA-2, the months of June–August are divergent (maxima of 1.02 ± 0.22 and 0.97 ± 0.25 mm day−1 in August, respectively), while for the remaining products only July and August are divergent with a relatively weaker magnitude (e.g., 0.54 ± 0.22 mm day−1 for ERA-Interim, 0.57 mm day−1 ± 0.22 for ERA5, and 0.37 ± 0.21 mm day−1 for NARR in August). The convergence rate (negative VIMFD) is highest in the months of April and November with comparable magnitudes for the four reanalysis products, except NARR, which has a weaker absolute VIMFD magnitude throughout the year.
To understand the mechanistic processes behind moisture fluxes across the boundaries, we use ERA-Interim dataset for concision and because it has a relatively better representation of all the moisture budget components. We specifically focus on the summer (JJAS) and winter (DJFM) months to understand their contrasting behavior in the seasonal climatology. In winter, the synoptic-scale (submonthly) weather systems dominate the VIMFD magnitudes (Fig. S1 in the online supplemental material), bringing substantial moisture into the domain and resulting in overall convergence. The meridional moisture transport yield convergence (Fig. S3e in the online supplemental material) as the moisture inflows through the south, while fluxes across the north boundary are relatively weaker (Figs. 5c,d). In the zonal direction, the fluxes intensify from west to east, with higher moisture being removed through the east boundary (Figs. 5a,b). The strength of this zonal divergence is weaker than the meridional convergence, ensuing net convergence in winter. In summer, there is a strong moisture influx across the south boundary that is prominent on the southwestern edge from 950 to 800 hPa (Fig. 5g), while the north boundary has relatively weaker outflow (Fig. 5h) resulting in meridional convergence. As in the winter, the westerly fluxes bring in moisture and carry it out through the eastern boundary (Figs. 5e,f). However, unlike winters, the summer moisture gradient between the two boundaries is greater (supplemental Fig. S3d), resulting in much stronger divergence (supplemental Fig. S3h). Overall, the zonal divergence is the dominant contributor toward the net summer divergence (Fig. 2d).
Vertical profile of moisture flux across boundaries in (top) winter (DJFM) and (bottom) summer (JJAS) for ERA-Interim: (a),(b),(e),(f) zonal flux qu and (c),(d),(g),(h) meridional flux qυ for which positive sign shows northward flow. The units for specific humidity q and winds (u, υ) are kilograms of water per kilogram of air and meters per second, respectively.
Citation: Journal of Climate 34, 2; 10.1175/JCLI-D-19-0952.1
(ii) Spatial patterns
The spatial patterns of the VIMFD conform with the seasonality, showing net convergence in winter, autumn, and spring and overall divergence in the summer (Fig. 6). In spring, we observe enhanced net convergence in the atmospheric column over the lakes and divergence zones downwind, which is more prominent in CFSR (Figs. 6d–f). The summer (Figs. 6g–i) and autumn seasons (Figs. 6j–l) exhibit more spatial variability in the convergence and divergence patterns. In summer, Lake Superior mostly has convergence overhead, with divergence toward the downwind (eastern) shore, while over Lakes Michigan and Huron, there are mixed patterns depending on the location and reanalysis product. Both MERRA2 and CFSR have strong divergence downwind of lakes that is not as pronounced in ERA-Interim. The domain also shows patterns of alternating convergence and divergence over land to the southeast of Lakes Erie and Ontario between 40°–45°N and 80°–74°W, where the Appalachian mountainous terrain can also influence the fluxes. The effect of topography on VIMFD is more prominent in CFSR, especially during the summer months, producing alternating convergence and divergence patterns on the southeastern edge of the domain that merges with the effects of the lakes (Fig. 6i). It is important to note that the native model resolution of the reanalyses will also affect the convergence/divergence patterns, where finer resolution models may produce more resolved details as we observe in CFSR (Table S1 in the online supplemental material). These localized spatial patterns provide some indication that the topography and water bodies may influence the regional distribution of convergence and divergence zones.
Seasonal (as labeled) spatial patterns of VIMFD (mm day−1) averaged over the 1981–2010 period for (left) ERA-Interim, (center) MERRA-2, and (right) CFSR. Blue colors (positive sign) show divergence, and red colors (negative sign) show convergence.
Citation: Journal of Climate 34, 2; 10.1175/JCLI-D-19-0952.1
(iii) Lake effects on local MFD patterns
The Great Lakes influence the regional climate by altering various processes associated with temperature, water vapor mixing ratio, total precipitation, wind patterns, and vertical motion. For example, Notaro et al. (2013) found that the instability effect of lakes in the cold season (due to warmer lake temperatures) can result in low-level convergence and an anomalous ascent, with the reverse effect in summers. The presence of lakes increases turbulent fluxes (sensible and latent) in the winter and enhances stability in the summer due to colder lake surface temperature than the overlying air (Notaro et al. 2013). A cross section at 44°N through Lakes Michigan and Huron (Fig. 7) indicates a similar influence of the lakes on the vertical profile of atmospheric MFD. In summer, air diverges directly over the lake surface in the planetary boundary layer (PBL) and converges downwind of the two lakes. In the troposphere, this behavior is reversed with convergence above lakes and divergence above land (Figs. 7g–i). In other seasons, the patterns are shifted; for example, in spring we observe convergence over the lake throughout the atmospheric column and divergence downwind. The vertical profile in the autumn and winter are similar at the 44°N cross section, albeit with higher magnitudes in autumn. Despite the strong magnitudes of convergence/divergence in the PBL, the overall behavior for the atmospheric column is dominated by the magnitudes in the troposphere. While the three reanalyses (ERA-Interim, MERRA-2, CFSR) have comparable patterns of MFD in the atmospheric column (Fig. 7), the nuanced differences in the extent and orientation of the divergent fluxes produce differences in the VIFMD patterns (Fig. 6), which can be influenced by the native model resolution in addition to the magnitude of overall moisture transport across the boundaries and how the models represent the lakes (specifically lake cover and surface temperature).
Atmospheric cross section of MFD ∇ ⋅ qu at 44°N averaged over the 1981–2010 period. The locations of Lake Michigan (87°–88°W) and Lake Huron (82°–84°W) are marked in black along the x axis. The units for divergence are kilograms of water per kilogram of air per second.
Citation: Journal of Climate 34, 2; 10.1175/JCLI-D-19-0952.1
To further understand the potential role of the lakes in inducing convergence/divergence patterns, we apply a conceptual framework for an idealized two-dimensional summer case (Fig. 8a). During the summer, land heats more rapidly than the lakes, increasing both the sensible and latent heat fluxes. The latent heat fluxes add moisture due to higher land surface evapotranspiration and the sensible heat fluxes drive rising vertical motion, inducing moisture convergence in the PBL over land. As this moist warm air rises in the troposphere, westerly winds transport moisture eastward, resulting in convergence downwind. Over the lake in the troposphere, the westerlies continue to carry some moisture eastward, and the subsidence of air toward the PBL will replace the moisture mass loss by surface divergence. This divergence over the lake surface arises because more moisture is being carried out through lateral transport than is being added into the layer from lake surface evaporation. In this simplified theoretical framework, the horizontal transport downwind (eastward) and alternating ascension/subsidence patterns are driven by the temperature contrast between land and lake surfaces together with the synoptic wind patterns.
(a) Conceptual schematic of the MFD vertical profile in summer resulting from the land–lake temperature contrast during the early summer months when the lake surface is cooler than land. ERA-Interim is used to show (b),(c) 2-m air and surface temperature difference (K); (d),(e) evaporation rate (mm day−1); and (f),(g) the atmospheric cross section of MFD at 44°N (kilograms of water per kilogram of air per second) for (left) January and (right) June. The vertical winds (m s−1) are scaled by a factor of 100 to make them comparable to the zonal winds u.
Citation: Journal of Climate 34, 2; 10.1175/JCLI-D-19-0952.1
In reality, the lake surface temperature is not uniform, nor are the land surface evaporation rates homogeneous. These subtle differences, and the divergence in three-dimensional space, produce more complex patterns (Figs. 8f,g for ERA-Interim) but generally conform to the theoretical framework. We detail this framework for two months, January and June, corresponding to winter and summer months, respectively, when the temperature gradient between the lake surface and the overlying air is the highest. In late spring to early summer, the lake surface is much colder than the overlying air (Fig. 8c) with negligible evaporation from the lake surface (Fig. 8e), while the higher surface temperature over land will increase latent heat fluxes and add moisture into the air. In June, ERA-Interim shows divergence over the two lake surfaces and convergence in the troposphere at approximately 900–700 hPa (Fig. 8g). In the PBL, the convergence zones have positive vertical velocities (i.e., warm moist air rising), while in troposphere, they have corresponding subsidence toward the divergence zone below as shown in the conceptual schematic. We also note some eastward flow in the tropospheric convergence zones.
In January, the lake surface temperature is warmer than the surrounding land and the overlying air temperature (Fig. 8b) and has higher evaporation rates (Fig. 8d). The MFD patterns reverse as compared with June, specifically with the emergence of convergence over the eastern shores of the lakes (Fig. 8f). Again, the PBL convergence zones correspond with upward flow and the divergence zones correspond with subsidence. Near the eastern edge of the domain (around 76°W), there is strong convergence in the PBL, likely driven by the Appalachian topography.
5) Residual
Because the reanalyses assimilate observations, they do not comply with mass, energy, or momentum conservation, and therefore an additional term arises in the moisture budget (Bosilovich et al. 2011; Trenberth et al. 2011; Bosilovich et al. 2015). The residual is an artificial term arising from the nonclosure of the mass balance equation for which the magnitudes can be nontrivial and comparable to VIMFD magnitudes (Fig. 2e). Seager and Henderson (2013) discuss the errors that can arise due several factors, including the diagnostic computation of the divergence term, the use of longer time steps, and coarse vertical resolution of the atmospheric profile. They suggest that representation of some processes in the models can be improved at the finer resolutions. Dynamical constraints in the model (requirement of internal consistency in u and υ winds, temperature, etc.), data assimilation (e.g., incorporation of satellite data), and perpetually evolving observation and numerical systems can also introduce spurious shifts in the time series (Dee et al. 2011; Trenberth et al. 2011; Gelaro et al. 2017), thus affecting the mass balance. Furthermore, the seasonal cycle we observe in each model’s residual (Fig. 2e) may also be influenced by limitations in its representation of the parameterized quantities, specifically precipitation and evaporation.
We note that MERRA-2 and CFSR have a high residual (up to 0.89 and 0.73 mm day−1, respectively) during the summer and autumn months, indicating that the products may not be adequately capturing processes in these seasons. NARR has an error introduced in MFD computations that is due to regridding of the vector quantities, which are on a Lambert conformal conic grid and require interpolation to a rectilinear grid. Regridding vector quantities, particularly vector pairs such as zonal (u) and meridional (υ) wind components individually, introduces errors in the calculation of divergence. NARR also assimilates precipitation data, which can render the mass balance assessment physically inconsistent and result in considerable residual magnitudes. The two ERA products have smaller residual magnitudes demonstrating a more representative mass balance (R generally less than 0.30 mm day−1). Small alterations in ERA5 precipitation and evaporation climatology have superficially resulted in an overall improved budget closure, especially during the winter and spring months. Nevertheless, it is important to note that small residual is not necessarily indicative of a better representation of climatic variables in the model and can just be a mathematical artifact.
b. Using the moisture budget to understand precipitation seasonality
The moisture budget components (mm day−1) for (a) February, (b) April, (c) June, and (d) November. The dotted bar shows the net effect of the local evapotranspiration and large-scale moisture transport (E − VIMFD). The 95% confidence intervals for precipitation, dPW, E, and VIMFD are also shown.
Citation: Journal of Climate 34, 2; 10.1175/JCLI-D-19-0952.1
These monthly contributions signify three features: 1) precipitation seasonality in the Great Lakes region is influenced by a combined role of the two moisture sources (i.e., regional evapotranspiration and net moisture transport through the boundaries), and their relative contribution varies markedly from month-to-month; 2) error in precipitation magnitude can arise due to misrepresentation of either of the two origins (E and VIMFD) in the models; and 3) in some cases (particularly MERRA-2 and CFSR) the mass balance is violated by producing surplus moisture than is available from the remote and local moisture sources. For CFSR, Trenberth et al. (2011) also found limitations with its global precipitation representation, stating that the analysis increment (periodic changes in temperature, wind, humidity, and geopotential fields due to model adjustment to observations) provides extra moisture for precipitation over land when the evapotranspiration estimate is low.
A visible feature of the precipitation seasonality in the Great Lakes region is the decline from July to August, followed by an increase in September. The July–August decline is particularly prominent for MERRA-2 (−0.42 mm day−1) and CFSR (−0.39 mm day−1) and is slightly weaker for ERA-Interim (−0.23 mm day−1) and NARR (−0.22 mm day−1), while for ERA5 it is insignificant (−0.1 mm day−1) (Fig. 10a). Similarly, the August–September precipitation increment is also the highest for CFSR (0.49 mm day−1) and MERRA-2 (0.24 mm day−1) (Fig. 10b). We assess this behavior in month-to-month differences in Fig. 10 (corresponding monthly magnitudes for July – September are provided in Fig. S4 in the online supplemental material). The increase in strength of VIMFD from July to the August maxima ranges from 0.18 mm day−1 for CFSR to 0.31 mm day−1 for NARR (Figs. 10a and 2c). Simultaneously, there is a decrease in evaporation rates, ranging from −0.73 mm day−1 for NARR to −0.22 mm day−1 for ERA5. Thus, the two potential sources for atmospheric moisture concurrently drop in all the reanalyses, resulting in reduced water vapor in the atmospheric column, which plays a cooperative role in generating the decline in precipitation and atmospheric moisture content from July to August (Fig. 10a).
Magnitude difference in the moisture budget components for (a) August − July and (b) September − August. The same legend is used as in Fig. 9. Monthly magnitudes for July, August, and September are provided in Fig. S4 in the supplemental material.
Citation: Journal of Climate 34, 2; 10.1175/JCLI-D-19-0952.1
From August to September (Fig. 10b), the decrease in the evaporation rate ranges from −1.07 to −0.61 mm day−1. In contrast, there is an increase in convergence of external moisture flux, which exceeds the drop in evaporation (except for NARR), and the difference between the two sources results in net positive moisture content and an increase in precipitation rate from August to September. We note that MERRA-2 again has higher moisture content than is available from the two sources resulting in a positive residual. For CFSR, a smaller decline in evaporation rate results in higher moisture and a greater increase in precipitation rate. We also note that NARR has a high residual term, and despite the drop in the net moisture content it produces a small increase in the precipitation rate, primarily due to assimilation of observation-based precipitation data.
4. Conclusions
This study aims to identify the drivers of precipitation seasonality in the Great Lakes region through interpretation of the moisture budget components. The monthly climatology indicates the presence of a distinct precipitation cycle, with its magnitude peaking in the early summer, declining in August, followed by an increase in September, which has not been captured in prior studies discussing the regional precipitation (Li et al. 2010; Basile et al. 2017). Here, we explained this behavior through the complementary influence of synoptic-scale moisture influx and local latent heat fluxes that alters the moisture content in the atmosphere and can thus produce drying or moistening behavior. The change in total precipitable water is an order of magnitude smaller than the other moisture budget quantities; however, the atmospheric moisture content also drops at the same rate as precipitation from July to August, signaling the start of drier atmospheric conditions. The vertically integrated moisture flux divergence seasonal cycle shows contrasting behavior in the winter and summer months, with net divergence in summer and convergence in the remaining months, which is influenced by synoptic-scale processes that drive moisture fluxes across the boundaries. In summer, the strength of zonal divergence is the dominant contributor to the net summer divergence, whereas in winter and the two shoulder seasons, the meridional convergence dominates over the zonal divergence.
Previous regional assessments highlighted the importance of large-scale moisture transport in the interannual variability of precipitation (Grover and Sousounis, 2002; Li et al. 2010). Our seasonal analysis suggests that both the potential moisture sources (regional evapotranspiration and remote moisture influx) together drive the precipitation seasonality, and misrepresentation of any component can result in an anomalous precipitation simulation. In some instances, addition of surplus moisture, not accounted for by the local or remote moisture, produces irregularities in the seasonal cycle leading to large residual magnitudes, which may arise due to issues in the parameterization schemes of the models, systematic errors from coarse resolutions or methodological constraints, and data assimilation in the reanalyses. The Great Lakes also play a role in the precipitation cycle by moderating the two moisture sources. The contrasting surface-air temperature gradients affect seasonal differences in turbulent fluxes over the water bodies, which can influence the localized moisture flux convergence/divergence spatial patterns.
We also explored the biases in the reanalyses, and concluded that MERRA-2, CFSR, and NARR fail to capture important aspects of the atmospheric moisture budget in the Great Lakes region. For example, CFSR exhibits a highly uncharacteristic bimodal precipitation seasonality, MERRA-2 overestimates spring and early summer precipitation magnitudes, and NARR has subdued precipitation and VIMFD magnitudes. NARR also shows significant differences in precipitation magnitudes over the U.S.–Canada border, which is likely due to limitation in precipitation observation assimilation. In ERA5, the failure to simulate correct lake surface temperatures (Hersbach et al. 2020) can result in erroneous spatial patterns of the moisture budget components.
It is important to understand the seasonal variability of precipitation within the region and how it is resolved in the various reanalyses, as these products are often used for hydrological and atmospheric modeling. The mechanisms driving the precipitation patterns and their quantification are equally important because any irregularities in historical or present-day representation of precipitation patterns can be propagated and amplified in future projections, significantly altering the outcomes. Future projections indicate that, while the annual precipitation magnitude will unequivocally increase (d’Orgeville et al. 2014), precipitation patterns will have a disproportionate shift in seasons, with total magnitudes and frequency of high-intensity events increasing in winter and spring (Hayhoe et al. 2010; Basile et al. 2017) and decreasing in summer (Hayhoe et al. 2010; Kim et al. 2015; Peltier et al. 2018). This work also highlights the importance of appropriately representing the Great Lakes in models: owing to their capacity to retain heat for longer durations, their different heating/cooling behavior under the warming climate can have a direct effect on the regional moisture budget and consequently on precipitation patterns.
Acknowledgments
This study was supported by National Science Foundation Grant OCE-1600012. We thank Prof. Ashley E. Payne and the three anonymous reviewers for their critical analysis and constructive feedback.
REFERENCES
Assel, R., K. Cronk, and D. Norton, 2003: Recent trends in Laurentian Great Lakes ice cover. Climatic Change, 57, 185–204, https://doi.org/10.1023/A:1022140604052.
Banacos, P. C., and D. M. Schultz, 2005: The use of moisture flux convergence in forecasting convective initiation: Historical and operational perspectives. Wea. Forecasting, 20, 351–366, https://doi.org/10.1175/WAF858.1.
Basile, S. J., S. A. Rauscher, and A. L. Steiner, 2017: Projected precipitation changes within the Great Lakes and western Lake Erie basin: A multi-model analysis of intensity and seasonality. Int. J. Climatol., 37, 4864–4879, https://doi.org/10.1002/joc.5128.
Berrisford, P., and Coauthors, 2011. The ERA-Interim archive: Version 2.0. ERA Rep. Series 1, 27 pp., https://www.ecmwf.int/sites/default/files/elibrary/2011/8174-era-interim-archive-version-20.pdf.
Bosilovich, M. G., and Coauthors, 2008: Evaluation of global precipitation in reanalyses. J. Appl. Meteor. Climatol., 47, 2279–2299, https://doi.org/10.1175/2008JAMC1921.1.
Bosilovich, M. G., F. R. Robertson, and J. Chen, 2011: Global energy and water budgets in MERRA. J. Climate, 24, 5721–5739, https://doi.org/10.1175/2011JCLI4175.1.
Bosilovich, M. G., and Coauthors, 2015: MERRA-2: Initial evaluation of the climate. NASA Tech. Memo. NASA/TM-2015-104606/Vol. 43, 145 pp., https://gmao.gsfc.nasa.gov/pubs/docs/Bosilovich803.pdf.
Dagan, G., P. Stier, and D. Watson-Parris, 2019: Analysis of the atmospheric water budget for elucidating the spatial scale of precipitation changes under climate change. Geophys. Res. Lett., 46, 10 504–10 511, https://doi.org/10.1029/2019GL084173.
Decremer, D., and Coauthors, 2014: Which significance test performs the best in climate simulations? Tellus, 66A, 23139, https://doi.org/10.3402/tellusa.v66.23139.
Dee, D. P., and Coauthors, 2011: The ERA-Interim reanalysis: Configuration and performance of the data assimilation system. Quart. J. Roy. Meteor. Soc., 137, 553–597, https://doi.org/10.1002/qj.828.
Dominguez, F., and Coauthors, 2006: Impact of atmospheric moisture storage on precipitation recycling. J. Climate, 19, 1513–1530, https://doi.org/10.1175/JCLI3691.1.
d’Orgeville, M., and Coauthors, 2014: Climate change impacts on Great Lakes Basin precipitation extremes. J. Geophys. Res., 119, 10 799–10 812, https://doi.org/10.1002/2014JD021855.
Fujisaki-Manome, A., and Coauthors, 2017: Turbulent heat fluxes during an extreme lake-effect snow event. J. Hydrometeor., 18, 3145–3163, https://doi.org/10.1175/JHM-D-17-0062.1.
Gelaro, R., and Coauthors, 2017: The Modern-Era Retrospective Analysis for Research and Applications, version 2 (MERRA-2). J. Climate, 30, 5419–5454, https://doi.org/10.1175/JCLI-D-16-0758.1.
Gottschalck, J., and Coauthors, 2005: Analysis of multiple precipitation products and preliminary assessment of their impact on global land data assimilation system land surface states. J. Hydrometeor., 6, 573–598, https://doi.org/10.1175/JHM437.1.
Grover, E. K., and P. J. Sousounis, 2002: The influence of large-scale flow on fall precipitation systems in the Great Lakes Basin. J. Climate, 15, 1943–1956, https://doi.org/10.1175/1520-0442(2002)015<1943:TIOLSF>2.0.CO;2.
Harris, I., and Coauthors, 2014: Updated high-resolution grids of monthly climatic observations—The CRU TS3.10 dataset. Int. J. Climatol., 34, 623–642, https://doi.org/10.1002/joc.3711.
Hayhoe, K., J. Van Dorn, T. Carley, N. Schlegal, and D. Wuebbles, 2010: Regional climate change projections for Chicago and the U.S. Great Lakes. J. Great Lakes Res., 36, 7–21, https://doi.org/10.1016/j.jglr.2010.03.012.
Hersbach, H., and Coauthors, 2020: The ERA5 global reanalysis. Quart. J. Roy. Meteor. Soc., 1999–2049, https://doi.org/10.1002/qj.3803.
Kim, J., V. Y. Ivanov, and S. Fatichi, 2015: Climate change and uncertainty assessment over a hydroclimatic transect of Michigan. Stochastic Environ. Res. Risk Assess., 30, 923–944, https://doi.org/10.1007/s00477-015-1097-2.
Lamb, P. J., D. H. Portis, and A. Zangvil, 2012: Investigation of large-scale atmospheric moisture budget and land surface interactions over U.S. Southern Great Plains including for CLASIC (June 2007). J. Hydrometeor., 13, 1719–1738, https://doi.org/10.1175/JHM-D-12-01.1.
Lélé, M. I., and L. M. Leslie, 2016: Intraseasonal variability of low-level moisture transport over West Africa. Climate Dyn., 47, 3575–3591, https://doi.org/10.1007/s00382-016-3334-x.
Li, L. F., W. H. Li, and A. P. Barros, 2013: Atmospheric moisture budget and its regulation of the summer precipitation variability over the Southeastern United States. Climate Dyn., 41, 613–631, https://doi.org/10.1007/s00382-013-1697-9.
Li, X. P., and Coauthors, 2010: Hydroclimate and variability in the Great Lakes region as derived from the North American Regional Reanalysis. J. Geophys. Res., 115, D12104, https://doi.org/10.1029/2009JD012756.
Lopez, P., 2007: Cloud and precipitation parameterizations in modeling and variational data assimilation: A review. J. Atmos. Sci., 64, 3766–3784, https://doi.org/10.1175/2006JAS2030.1.
Mesinger, F., and Coauthors, 2006: North American Regional Reanalysis. Bull. Amer. Meteor. Soc., 87, 343–360, https://doi.org/10.1175/BAMS-87-3-343.
Notaro, M., and Coauthors, 2013: Influence of the Laurentian Great Lakes on regional climate. J. Climate, 26, 789–804, https://doi.org/10.1175/JCLI-D-12-00140.1.
Notaro, M., V. Bennington, and S. Vavrus, 2015: Dynamically downscaled projections of lake-effect snow in the Great Lakes basin. J. Climate, 28, 1661–1684, https://doi.org/10.1175/JCLI-D-14-00467.1.
Peltier, W. R., and Coauthors, 2018: Uncertainty in future summer precipitation in the Laurentian Great Lakes basin: Dynamical downscaling and the influence of continental-scale processes on regional climate change. J. Climate, 31, 2651–2673, https://doi.org/10.1175/JCLI-D-17-0416.1.
Saha, S., and Coauthors, 2010: The NCEP Climate Forecast System Reanalysis. Bull. Amer. Meteor. Soc., 91, 1015–1057, https://doi.org/10.1175/2010BAMS3001.1.
Şahin, S., and Coauthors, 2015: Large scale moisture flux characteristics of the Mediterranean basin and their relationships with drier and wetter climate conditions. Climate Dyn., 45, 3381–3401, https://doi.org/10.1007/s00382-015-2545-x.
Schneider, U., and Coauthors, 2014: GPCC’s new land surface precipitation climatology based on quality-controlled in situ data and its role in quantifying the global water cycle. Theor. Appl. Climatol., 115, 15–40, https://doi.org/10.1007/s00704-013-0860-x.
Scott, R. W., and F. A. Huff, 1996: Impacts of the Great Lakes on regional climate conditions. J. Great Lakes Res., 22, 845–863, https://doi.org/10.1016/S0380-1330(96)71006-7.
Seager, R., and N. Henderson, 2013: Diagnostic computation of moisture budgets in the ERA-Interim reanalysis with reference to analysis of CMIP-archived atmospheric model data. J. Climate, 26, 7876–7901, https://doi.org/10.1175/JCLI-D-13-00018.1.
Shafran, P., J. Woollen, W. Ebisuzaki, W. Shi, Y. Fan, R. W. Grumbine, and M. Fennessy, 2004: Observational data used for assimilation in the NCEP North American Regional Reanalysis. 14th Conf. on Applied Climatology, Seattle, WA, Amer. Meteor. Soc., 1.4, https://ams.confex.com/ams/84Annual/techprogram/paper_71689.htm.
Trenberth, K. E., and C. J. Guillemot, 1995: Evaluation of the global atmospheric moisture budget as seen from analyses. J. Climate, 8, 2255–2272, https://doi.org/10.1175/1520-0442(1995)008<2255:EOTGAM>2.0.CO;2.
Trenberth, K. E., J. T. Fasullo, and J. Mackaro, 2011: Atmospheric moisture transports from ocean to land and global energy flows in reanalyses. J. Climate, 24, 4907–4924, https://doi.org/10.1175/2011JCLI4171.1.
Wang, Z. Q., and Coauthors, 2017: Atmospheric moisture budget and its regulation on the variability of summer precipitation over the Tibetan Plateau. J. Geophys. Res., 122, 614–630, https://doi.org/10.1002/2016JD025515.
Willmott, C. J., and K. Matsuura, 2001: Terrestrial air temperature and precipitation: Monthly and annual time series (1950–1999). Center for Climatic Research, Department of Geography, University of Delaware, accessed 1 May 2019, http://climate.geog.udel.edu/~climate/html_pages/download.html.
Yang, Q., and Coauthors, 2014: Atmospheric moisture budget and spatial resolution dependence of precipitation extremes in aquaplanet simulations. J. Climate, 27, 3565–3581, https://doi.org/10.1175/JCLI-D-13-00468.1.
Yang, S., and Coauthors, 2009: Variations of U.S. regional precipitation and simulations by the NCEP CFS: Focus on the Southwest. J. Climate, 22, 3211–3231, https://doi.org/10.1175/2009JCLI2532.1.
Zangvil, A., D. H. Portis, and P. J. Lamb, 2001: Investigation of the large-scale atmospheric moisture field over the midwestern United States in relation to summer precipitation. Part I: Relationships between moisture budget components on different timescales. J. Climate, 14, 582–597, https://doi.org/10.1175/1520-0442(2001)014<0582:IOTLSA>2.0.CO;2.
Zangvil, A., D. H. Portis, and P. J. Lamb, 2004: Investigation of the large-scale atmospheric moisture field over the midwestern United States in relation to summer precipitation. Part II: Recycling of local evapotranspiration and association with soil moisture and crop yields. J. Climate, 17, 3283–3301, https://doi.org/10.1175/1520-0442(2004)017<3283:IOTLAM>2.0.CO;2.
Zhang, L., A. Kumar, and W. Q. Wang, 2012: Influence of changes in observations on precipitation: A case study for the Climate Forecast System Reanalysis (CFSR). J. Geophys. Res., 117, D08105, https://doi.org/10.1029/2011JD017347.
