1. Background and motivation
Arid-to-semiarid regions located in subtropical zones are projected to experience some of the most adverse impacts of climate change. There is likely to be an increase in heat and aridity resulting from the retreat of the midlatitude jet and expansion of subtropical highs (e.g., Archer and Caldeira 2008; Seidel et al. 2008; Lu et al. 2009), for example, as summarized in the recent Climate Change Assessment for the Southwest (Garfin et al. 2013) and Intergovernmental Panel on Climate Change (IPCC) Fifth Assessment Report (IPCC 2013). Another conclusion within these two climate-assessment reports is that there will be an increase in precipitation intensity and more extreme weather because of the exponential increase in atmospheric water vapor capacity of the atmosphere with a higher mean temperature (e.g., Meehl et al. 2000). Significant positive trends in observed atmospheric water vapor exist globally and within the United States (e.g., Karl and Knight 1998; Karl and Trenberth 2003; Groisman et al. 2005; Willett et al. 2007; Santer et al. 2007). Recent increases in precipitation extremes have been attributed to increases in greenhouse gases (e.g., Karl and Trenberth 2003; Min et al. 2009, 2011).
Our geographic area of interest in this study is the southwestern United States, henceforth referred to as the Southwest. Observed 20-yr return-period thresholds of daily maximum precipitation in the Southwest have also exhibited an upward trend (Kunkel et al. 2013). Any long-term increases in precipitation intensity should be most apparent during the North American monsoon (NAM; Adams and Comrie 1997) in late summer (July–early September) because this is the period of warm-season severe weather caused by convective thunderstorms. There is observational evidence to suggest that monsoon precipitation is becoming more extreme in the Southwest and in northwestern Mexico (e.g., Anderson et al. 2010; Petrie et al. 2014; Chang et al. 2015).
We must necessarily depend on global and regional atmospheric climate models (GCMs or RCMs) to generate future projections and retrospective simulations of the NAM for impacts assessment, but these tools have caveats. GCMs, used for purposes of climate projection or seasonal forecasting, and global atmospheric reanalyses are generally challenged to represent the NAM as a salient climatological feature, in terms of its seasonal maximum in precipitation that occurs during July and August and/or its retreat in early autumn (e.g., Castro et al. 2007, 2012; Geil et al. 2013; Bukovsky et al. 2015). Regional models at the meso-β scale, even when “perfect” boundary forcing of an atmospheric reanalysis is applied, tend to overestimate monsoon precipitation in mountainous regions and underestimate precipitation that is associated with organized, propagating convection at lower elevations (Castro et al. 2012; Bukovsky et al. 2013).
With regard to the question of changes in extreme precipitation during the NAM, GCMs and RCMs with a grid spacing on the meso-β scale or coarser are inadequate to explicitly represent monsoon thunderstorms, and their statistical representation of precipitation extremes depends on the spatial resolution, improving with finer grid spacing (e.g., Tripathi and Dominguez 2013). RCM simulations at the meso-γ scale (on the order of 1-km grid spacing) are referred to as convective-permitting models (CPMs), because thunderstorms may be explicitly represented without the use of convective parameterization. The overall value added of CPMs, in particular with respect to the representation of precipitation extremes and organized convective structures, has been well established (e.g., Prein et al. 2015; Kendon et al. 2017). In the context of numerical weather prediction–type (NWP type) simulations, CPMs are required to reasonably represent mesoscale convective systems (MCSs) during the NAM, one of the most common meteorological triggers for severe weather events (Cassell et al. 2017, manuscript submitted to Mon. Wea. Rev.).
The current study evaluates historical changes in the intensity of NAM precipitation within the Southwest in the context of CPM simulations of severe weather events during a historical period 1951–2010. We select the specific severe weather event days from a long-term dynamically downscaled reanalysis on the basis of thermodynamic conditions of atmospheric instability and precipitable water that the authors have already established are robust precursors of severe weather during the NAM in the Southwest (Mazon et al. 2016). If long-term increases in atmospheric moisture are present in the Southwest during the NAM and are represented in the dynamically downscaled reanalysis, then we hypothesize that there should also be long-term increases in the model-simulated precipitation intensity in the CPM severe weather event simulations. Our objective is to create a historical database of CPM severe weather event simulations during the NAM that can be used for 1) purposes of climate impacts assessment within the Southwest and 2) a future comparison with similarly simulated severe weather events using dynamically downscaled global climate models. In more broad terms, we hope to demonstrate the value added of CPMs for evaluating changes in monsoonal precipitation extremes in an arid-to-semiarid region.
2. Severe weather event selection and NWP-type modeling
a. Severe weather event selection
The severe weather event days for NWP-type simulations at CPM grid spacing are selected from an existing long-term RCM that is described in Chang et al. (2015). This RCM simulation dynamically downscales NCEP-1 global reanalysis data (Kalnay et al. 1996) during the complete historical period 1948–2010 with version 3 of the Weather Research and Forecasting (WRF) Model (Skamarock et al. 2008) at 35-km grid spacing for a domain that spans the contiguous United States and Mexico. Specific severe weather event days are chosen on the basis of favorable thermodynamic characteristics for monsoon convection in the Southwest (Mazon et al. 2016), as based on the model variables of daily maximum column-integrated precipitable water (PW) and most unstable convective available potential energy (MUCAPE). When both modeled PW and MUCAPE are in the top 20% of the distribution during the period June–September (i.e., 122 days per year during the monsoon period) within the Southwest, the event is flagged as a severe weather event day. Identified severe weather event days using this method applied to Tucson, Arizona, radiosonde data correspond well to National Weather Service observed storm reports in southern Arizona, with a hit rate of approximately 60% (Mazon et al. 2016). A convective day is defined as starting at 1200 UTC to ensure uniformity with radiosonde observational data. The locations of seven operational radiosonde stations throughout the Southwest have been previously used to objectively define the PW and MUCAPE characteristics for the entire region with this downscaled reanalysis, as described in Mazon et al. (2016) and Castro [2017; note that the material cited in several places in the present paper as being from Castro (2017) was also presented in an unpublished 2014 poster presentation by M. Jares et al. (https://ams.confex.com/ams/94Annual/webprogram/Paper237080.html)].
(a) Positions of domain 2 (d02) and domain 3 (d03) with 10- and 2.5-km resolution, respectively, within the larger 35-km coarse-resolution domain. Domain 2 uses forcing data from a downscaled reanalysis as described in Chang et al. (2015), and domain 3 is a convective-permitting nested domain within domain 2. (b) Elevation (m) within CPM domain 3.
Citation: Journal of Applied Meteorology and Climatology 56, 9; 10.1175/JAMC-D-16-0358.1
b. Regional atmospheric model simulations of severe weather events
Thermodynamically favorable severe weather event days identified in the WRF dynamically downscaled reanalysis during a retrospective “present day” period (1991–2010) are first simulated with a CPM grid spacing of 2.5 km. Results from these simulations are compared with hourly observed precipitation (derived from gauge and radar data) to verify that the atmospheric model can reasonably represent the diurnal cycle of precipitation and its relationship to convective organization and propagation. Severe weather event days within a retrospective “historical past” period (1951–70) are then simulated. Differences in the behavior of precipitation between the present-day and historical-past periods can help to reveal the impact of long-term observed changes in atmospheric moisture and instability on severe monsoon weather. The total number of simulated days in the present-day and historical-past periods is 255 and 268, respectively.
The WRF Model experimental design for simulation of severe weather events uses a two-domain nesting strategy, with an intermediate domain of 10-km grid spacing and a CPM meso-γ domain of 2.5-km grid spacing. The intermediate domain utilizes data from the aforementioned long-term dynamically downscaled NCEP-1 reanalysis as boundary forcing and covers the Southwest and northwestern Mexico. The 2.5-km CPM domain is centered over central and southern Arizona and extends over the entirety of Arizona and New Mexico, including portions of California, Colorado, Nevada, Texas, and Utah (Fig. 1), and it is over this geographic area that we display our results. WRF Model parameterization options on the CPM domain are nominally similar to what is used for generating real-time WRF quasi-operational NWP monsoon forecasts at the University of Arizona Department of Hydrology and Atmospheric Sciences (UA-HAS), but with two notable differences. Common parameterization options include a bulk microphysics scheme (Thompson et al. 2004), Mellor–Yamada–Janjić planetary boundary layer scheme (Janjić 1990, 1996, 2001) with eta surface layer (Janjić 1996, 2001), the Noah-MP land surface model (Niu et al. 2011), Dudhia shortwave radiation (Dudhia 1989), and the Rapid Radiative Transfer Model longwave radiation (Mlawer et al. 1997). On the intermediate domain, the Kain–Fritsch cumulus parameterization scheme is applied (Kain and Fritsch 1993; Kain 2004), with the modified convective trigger and CAPE closure assumption of Truong et al. (2009) that better accounts for dynamic pressure effects in complex terrain. As shown in Truong et al. (2009), this modified Kain–Fritsch scheme better represents propagating MCS-type convection in complex terrain in an RCM with meso-β grid spacing. An urban canopy model, with optimized anthropogenic parameters for the Southwest as adapted from Grossman-Clarke et al. (2010), is applied at grid points with a defined urban land-use classification, including the cities of Tucson; Phoenix, Arizona; Albuquerque, New Mexico; and Las Vegas, Nevada.
The main objective of the CPM simulation design is to achieve a more reasonable representation of organized, propagating MCS-type convection that we know a priori is not represented well in RCM simulations at the meso-β scale. Although we arguably could have spent additional time in testing various parameterization options (e.g., convection, microphysics, and planetary boundary layer) to optimize a model precipitation solution, that is not the objective of this study. UA-HAS WRF operational products have been used by variety of public and private sector entities in Arizona during the past decade, especially and including for extreme-weather events during the monsoon. A good example is the 5 July 2011 Phoenix dust storm (haboob) event (Raman et al. 2014). So our goal is to assess long-term climatological changes in precipitation within an existing and robust operational NWP framework.
For a given identified severe weather event that satisfies the thermodynamic threshold criteria as previously described, the CPM WRF simulation is performed as follows in an NWP-type mode. The event simulation is initialized at 0600 UTC [2300 mountain standard time (MST)] of the day prior to the event, and the simulation is executed for 30 h, ending at 1200 UTC (0500 MST) on the day following the event. The initialization in the evening prior to the event allows for 6 h of model spinup time, consistent with UA-HAS operational forecast practices. Model output from the CPM grid is saved hourly, because this temporal resolution resolves well the diurnal cycle of convection.
c. Observational verification data for precipitation
The quality-controlled stage-IV combined NEXRAD–gauge precipitation product (http://data.eol.ucar.edu/codiac/dss/id=21.093) is used to compare observed precipitation with WRF-modeled precipitation on the CPM simulation domain. The specific dataset that we use is the Hourly Precipitation Data digital dataset DSI-3240 archived at the National Centers for Environmental Information, formerly known as the National Climatic Data Center (NCDC). Although we consider the stage-IV data as observational truth in the context of this study, it is important to note that these data have problems with respect to estimating precipitation in complex terrain because of issues around lack of rain gauge observations and radar beam blockage (e.g., Adams et al. 2014; Minjarez-Sosa et al. 2017). Hourly station data from the Cooperative Observer Program (COOP) are used to verify modeled precipitation diurnal cycle in some specific locations (http://www.ncdc.noaa.gov/cdo-web/search?datasetid=PRECIP_HLY). To evaluate long-term changes in precipitation, we use three sources of gridded daily precipitation data that are available for the contiguous United States since 1950: 1) Climate Prediction Center (CPC) 0.25° daily U.S. unified gauge-based analysis (Higgins et al. 1996), 2) gridded meteorological data produced originally by the Surface Water Modeling group at the University of Washington, at ⅛° resolution (Maurer et al. 2002), and 3) an updated version of these same data, at 1/16° resolution (Livneh et al. 2013). In both the Maurer and Livneh datasets, the COOP station precipitation data are the basis for the precipitation observations. We define the set of precipitation events in all sources of observational data as the top 20% of daily precipitation events that occur during the monsoon within a 20-yr period at a given location so as to be directly comparable in sample size to the subset of severe weather event days that are defined by thermodynamic criteria and simulated on the convective-permitting model grid.
3. Statistical-analysis methods of precipitation distributions and changes
The representation of the likelihood of receiving a specific rainfall amount is best accomplished by fitting a theoretical probability density function (PDF). A theoretical PDF is a continuous function with no discontinuity that can be determined for every local grid point on the map. These distributions make it possible to estimate the likelihood of rainfall being within a specified range. The gamma distribution typically yields a good PDF fit to a total precipitation distribution except in the upper tail. Changes in mean precipitation from one time period to another can be assessed using a t statistic.
4. Long-term changes in atmospheric thermodynamic conditions in the Southwest
The requisite conditions for NAM thunderstorm development have been previously discussed by the authors in Mazon et al. (2016) and Lahmers et al. (2016). To summarize, favorable thermodynamic conditions are a primary requirement for development of any monsoon thunderstorms in the NAM region. Moist, rising air occurs during the day over mountain ranges because of the differential heating of the mountains relative to the surrounding air. In a conditionally unstable atmosphere, water vapor in rising air may condense to form cumuliform clouds, extending through the entire depth of the troposphere when monsoon thunderstorms are fully mature. Thunderstorms begin to develop over mountain ranges in late morning to early afternoon and produce precipitation by late afternoon. If dynamic conditions are favorable, namely, there is the presence of a transient inverted trough and/or surge of moisture from the Gulf of California, monsoon thunderstorms in Arizona may organize into MCSs that propagate westward off the mountains during the late-afternoon and evening hours.
Atmospheric instability and moisture during the monsoon have substantially changed over the past 30 years within the Southwest in the context of the downscaled reanalysis, as reported earlier (Castro 2017; Lahmers et al. 2016) and shown in Fig. 2. The mean difference in MUCAPE from the downscaled reanalysis (Fig. 2a) shows an overall increase in atmospheric instability, excluding the Gulf of California. This increase of MUCAPE is maximized over northern Arizona and the southern parts of Nevada and Utah. The corresponding results for PW (Fig. 2b) show that PW has increased throughout the entire Southwest as well. The increase is maximized at the northern end of the Gulf of California and extends northward to the area where increases in MUCAPE are observed. The largest increase in PW occurs over central Arizona where Phoenix is located. The increase in atmospheric moisture and instability over the monsoon region during the past 60 years in the downscaled reanalysis is consistent with observations from radiosonde sounding data in the region (Table 1), as reported by Lahmers et al. (2016). As this prior work has already asserted, these modeled MUCAPE and PW changes are likely not artifacts of the global reanalysis that are appearing as a result of changes in instrument observing systems, principally the introduction of satellite data into the data assimilation after the late 1970s. We note that similar long-term increases in atmospheric moisture from radiosonde data within the western United States have also been documented in previous work by Durre et al. (2009). In the next sections we will test our hypothesis that these observed increases in atmospheric moisture and instability are more conducive to heavier convective precipitation during the modeled severe weather event days.
Differences (present day − historical past) of means for (a) MUCAPE (J kg−1) and (b) column-integrated PW (mm) from the 35-km dynamically downscaled NCEP reanalysis. The means of daily maximum MUCAPE and PW gridded data of the late period (1980–2010) were subtracted from the mean of the early period (1950–80). This figure is adapted from Fig. 31 of Castro (2017).
Citation: Journal of Applied Meteorology and Climatology 56, 9; 10.1175/JAMC-D-16-0358.1
Correlation coefficients and t statistics (in parentheses) for annual-average CAPE and PW (15 Jun–15 Sep) long-term trends at radiosonde sites in the Southwest. Boldface font denotes statistically significant values at the 90th percentile. [This table is adapted from Lahmers et al. (2016).]
5. Severe weather simulations during the period of the stage-IV product
To verify the performance of the WRF severe weather event simulations on the CPM grid prior to any evaluation of long-term trends in precipitation intensity, we compare model-simulated precipitation with the stage-IV combined radar–gauge observed precipitation product during the 9-yr period of 2002–10.
a. WRF-simulated daily precipitation versus stage IV during 2002–10
The model-simulated daily average precipitation (over 24 h of the convective day) and the corresponding stage-IV precipitation for all thermodynamically favorable severe weather event days within the 35-km downscaled NCEP reanalysis are shown in Fig. 3. The stage-IV data show that the thermodynamically identified severe weather event days have widespread precipitation over all of the Southwest. The average observed daily precipitation is a maximum (>5 mm) over the highest elevations because the mountains are the focal point for convective initiation. For example, the highest average observed precipitation in Arizona occurs over the Mogollon Rim (highlighted in the figure), which roughly bisects the state from the southeast corner to the northwest corner. In general, the CPM severe weather event simulations exhibit similar behavior in terms of capturing the terrain dependence of monsoon precipitation and show precipitation occurring throughout the Southwest. There is an underestimation of precipitation in the CPM simulations that is on the order of 1–2 mm day−1 (Fig. 3a). In contrast, similar to the WRF simulations of Tripathi and Dominguez (2013), there is a widespread area of overestimation of precipitation in the coarser-resolution simulation, in general on the order of 3–5 mm day−1 and even higher within mountainous areas (Fig. 3d). Overall, the CPM yields precipitation amounts that better correspond to the stage-IV product than the equivalent coarse-resolution model grid, notwithstanding the relatively higher uncertainty in stage-IV precipitation in mountainous regions that was mentioned earlier.
Composite daily means of precipitation (mm day−1) for all selected severe weather event days during 2002–10 of (a) the CPM (2.5 km) simulation, (b) stage-IV observations, and (c) the coarse-resolution (35 km) simulation and (d) the difference of the CPM simulation minus the coarse-resolution simulation.
Citation: Journal of Applied Meteorology and Climatology 56, 9; 10.1175/JAMC-D-16-0358.1
b. Diurnal cycle of convection
The CPM simulations are also able to reasonably simulate the diurnal cycle of convection in comparison with stage-IV data (calculated as precipitation averaged over 6-h time periods in Fig. 4). There is a maximum in precipitation that is centered over areas of high terrain during the afternoon [1100–1700 local time (LT)] and greater amounts of precipitation off the terrain of the Mogollon Rim during early evening (1700–2300 LT). To further demonstrate the value added of the CPM in representing the propagation of monsoon convection in association with the diurnal cycle (e.g., Prein et al. 2015), Fig. 5 shows the timing of maximum precipitation in the CPM and coarse-resolution simulations and stage-IV data in 6-hourly blocks. Similar to stage IV, the CPM exhibits a westward propagation of precipitation off the Mogollon Rim in late afternoon and into the nighttime hours (1700–0500 LT), reflecting MCS-type convection (e.g., Cassell et al. 2017, manuscript submitted to Mon. Wea. Rev.). The coarse-resolution simulation (incorrectly) shows all convective precipitation to occur with the maximum heating of the day (1100–1700 LT). Evolution of rainfall in New Mexico is also more reasonably captured in the CPM, in terms of pattern and intensity at selected peak hours from 1100 to 2300 LT.
Peak-hour composite 6-hourly means of precipitation (mm h−1) for all selected severe weather event days during 2002–10 of (a),(c) CPM (2.5 km) simulations and (b),(d) stage-IV observations for (top) 1100–1700 and (bottom) 1700–2300 LT.
Citation: Journal of Applied Meteorology and Climatology 56, 9; 10.1175/JAMC-D-16-0358.1
Period of peak rainfall in 6-h time intervals as indicated by color-coded periods for severe weather event days: (a) CPM (2.5 km) simulations, (b) stage-IV data, and (c) coarse-resolution (35 km) simulations.
Citation: Journal of Applied Meteorology and Climatology 56, 9; 10.1175/JAMC-D-16-0358.1
6. Long-term changes in monsoon precipitation in the Southwest
Long-term changes in monsoon precipitation are evaluated in this section from the perspective of both observations and results of the CPM severe weather event simulations. In the presentation of these results, we attempt to distinguish between the mean changes in precipitation and the changes in precipitation extremes in the tails of the distribution, evaluated using the POT GP technique.
a. Broadening and flattening of the distribution of daily monsoon precipitation
The histogram of NCDC COOP station extreme precipitation data for Phoenix (PHX) and Flagstaff (FLG), Arizona, is shown in Fig. 6. The red and blue lines on the figure are the POT GP distributions fitted to the right tail of the PDF for precipitation events above the 90th percentile. The fitted distributions satisfy the chi-square goodness-of-fit test at significance level α = 0.1. By “extreme” in reference to the plotting of the histogram in this figure, we are showing only that portion of the histogram of daily precipitation data that is above the 90th percentile and to which the GP distribution is applied to generate a theoretical distribution fit. The point at which the solid GP distribution curves abruptly end or become vertical lines toward the y axis defines the 90th percentile. So, for example, at Phoenix, the 90th percentile in the COOP station data is approximately 3 mm day−1. Using the COOP data, both Phoenix and Flagstaff have experienced an increase in precipitation extremes during the present period 1991–2010 as compared with the historical period 1951–70 (Fig. 6, left panels). The increase in extremes in monsoon precipitation in Phoenix can be interpreted as a broadening and flattening of the daily precipitation distribution. The differences are statistically significant with bootstrapping at a significance level α = 0.01. The right panels of Fig. 6 show the same analysis performed for the CPM-simulated severe weather events. Although CPM-simulated precipitation exhibits a dry bias, as discussed earlier with reference to comparison with the stage-IV product, it shows basically the same type of broadening and flattening of the distribution as the observed station data. The equivalent results are shown for Maurer and Livneh data in Fig. 7 and also confirm this behavior, strongly suggesting that the changes in daily precipitation distributions are likely not an artifact of any local-data quality-control issues at these particular COOP sites (i.e., changes in physical location of the station or instrumentation system; missing data).
Probability distributions of daily precipitation extremes in (top) PHX and (bottom) FLG of (left) NCDC COOP and (right) convective-permitting simulations. Blue and yellow bars are station histograms. Light blue and red lines are POT GP distributions fitted into the right tail of the PDF with events on the right of 90th percentile. The value of the 90th percentile is indicated when the fitted distribution curves end and/or become vertical as they approach the y axis starting from the right side of the plot.
Citation: Journal of Applied Meteorology and Climatology 56, 9; 10.1175/JAMC-D-16-0358.1
As in Fig. 6 for the Phoenix and Flagstaff locations, but using (left) Maurer and (right) Livneh gridded precipitation data.
Citation: Journal of Applied Meteorology and Climatology 56, 9; 10.1175/JAMC-D-16-0358.1
The idea of a broadening and flattening of the daily precipitation distribution also holds over the entire Southwest, in the context of comparing gridded precipitation products with the simulated severe weather events within the CPM domain. The present-day minus historical-past change in mean daily CPC and COOP observed precipitation for all days during the monsoon months of July and August is shown in Figs. 8a and 8c, in terms of the absolute change. The corresponding results for Maurer and Livneh data are shown in Figs. 9a and 9c. The results from the individual COOP stations with sufficient long-term records in the Southwest are included in Table 2. From the observational standpoint, mean daily monsoon precipitation has decreased as a whole in Arizona in recent decades. The largest absolute precipitation decreases in all of the observed gridded precipitation products occur over the Mogollon Rim (1 mm day−1, or greater than 30%). Decreases also occur in the Colorado River Valley and over parts of southwestern Arizona, where the more infrequent, organized convection accounts for a greater proportion of monsoon precipitation (e.g., Castro et al. 2007). The data from the COOP sites also show decreasing mean precipitation over most of Arizona. The COOP and CPC data also show significant increases in mean precipitation in western New Mexico (west of the Continental Divide), but these are not apparent in the Livneh and Maurer data. These results are similar to our previous analyses of long-term changes in monsoon precipitation using CPC data, as reported in Chang et al. (2015).
Significant changes (present day − historical past) in CPC (a) mean and (b) extreme precipitation (mm day−1) and in the (c) mean and (d) extremes of NCDC COOP precipitation data (mm day−1) for Southwest stations. In (c) and (d), stations at which changes are statistically significant are circled, with the color of the circle corresponding to the color keys for (a) and (b). Statistical significance is computed at the 0.01 level.
Citation: Journal of Applied Meteorology and Climatology 56, 9; 10.1175/JAMC-D-16-0358.1
As in Figs. 8a and 8b, but for (a),(b) Maurer and (c),(d) Livneh gridded precipitation data.
Citation: Journal of Applied Meteorology and Climatology 56, 9; 10.1175/JAMC-D-16-0358.1
Changes in mean and extreme precipitation (mm day−1) trends (present day − historical past) at NCDC COOP stations that have daily records back to the 1950s. Boldface font denotes statistical significance at the 0.01 level.
The corresponding pattern of changes in extreme monsoon precipitation is very different than that of the changes in mean precipitation. We characterize the changes in extreme precipitation above the 90th percentile, considering only the limited subset of severe weather event days as the entire sample. Severe weather event days are defined in different ways depending on the particular dataset considered: 1) in the model data, they are defined as those days that exceed the 80th percentile for atmospheric instability and moisture criteria and 2) in the observational data, they are defined as the top 20% of observed daily precipitation events. We characterize the changes in extreme precipitation above the 90th percentile, considering the entire modeled or observed sample size, respectively, as the limited set of modeled severe weather event days or top 20% of the observed daily precipitation events. The CPC data (Fig. 8b) would seem to indicate that observed extreme-event monsoon precipitation is, at least, not decreasing, but a much clearer and more physically revealing picture emerges when the Maurer and Livneh data are used (Figs. 9b,d). Both of these gridded data sources show statistically significant increases in extreme precipitation centered in the western half of Arizona and the Colorado River Valley, where organized MCS-type convection is the more dominant precipitation mechanism, as mentioned before. The corresponding changes in COOP-station precipitation data for extremes (Fig. 8d) match better the Livneh and Maurer data in this same area than the CPC data. The largest and most statistically significant increases in extreme precipitation in the Southwest occur at the COOP stations located in the southwestern portion of Arizona, at Yuma (30 mm) and Alamos (45 mm). At Phoenix, the increase in extreme precipitation (14 mm) and decrease in mean precipitation (−0.1 mm) support the conceptual idea of a broadening and flattening of the precipitation distribution.
The simulated precipitation changes from the present day to the historical past are shown in Fig. 10 for the severe weather event simulations at the coarse-resolution of 35 km (Figs. 10a,b) and the CPM grid spacing (Figs. 10c,d), for the significant changes in mean precipitation (Figs. 10a,c) and extreme precipitation (Figs. 10b,d). As in the observations, both the coarse-resolution and CPM simulations show a decrease in mean precipitation for the simulated severe weather events between the periods, with decreases maximized in the vicinity of the Mogollon Rim area of Arizona. The coarse-resolution and CPM simulations also show overall statistically significant increases in extreme precipitation in the Southwest, in contrast to the decreases in mean precipitation. The exact geographic locations at which the modeled extreme precipitation is becoming more intense and the spatial extent of the increases are markedly different between the two simulations, however. The CPM simulations show a relatively more coherent and larger geographic area over which extreme precipitation is becoming more intense. The most dramatic increases in Arizona occur in the western half of state and in the Colorado River Valley, to the south and west of the Mogollon Rim, similar to the Livneh and Maurer data. These areas also correspond to where the greatest increases in precipitable water are in the downscaled reanalysis (Fig. 2b). The significant increases in extreme monsoon precipitation in the CPM simulations are on the order of 10 mm day−1 (approximately 0.5 in.) or greater. By contrast, the coarser-resolution model simulations do not capture the correct geographic location of where extreme precipitation is increasing; their increases are centered more directly over Mogollon Rim and not to the south and west of this mountain range.
As in Figs. 8a and 8b, but for (a),(b) WRF coarse-resolution (35 km) and (c),(d) CPM (2.5 km) simulations of severe weather event days.
Citation: Journal of Applied Meteorology and Climatology 56, 9; 10.1175/JAMC-D-16-0358.1
To more quantitatively establish the level of correspondence between the precipitation changes in model simulations and observations, we consider the spatial pattern correlations and maps of the precipitation-change coherency metric. The spatial pattern correlation results are shown in Table 3 for the Livneh and Maurer data, using all points in the CPM domain. Although both coarse-resolution and CPM simulations show a decrease in mean precipitation, the pattern correlation with observed precipitation changes is negative. In contrast, the pattern correlation for changes in extremes is positive and statistically significant for both coarse and CPM simulations. Using the more spatially resolved Livneh dataset, the correlation coefficient increases from 0.56 to 0.69 from the coarse-resolution to CPM simulations. The results for the precipitation-change coherency metric are shown for coarse-resolution and CPM simulations, respectively, in Figs. 11 and 12. In both simulations, the geographic area of highest correspondence of observed precipitation changes to model results for both means and extremes is the western half of Arizona and the Colorado River Valley, especially for CPM-simulated changes in precipitation extremes.
Pattern correlation coefficients for model-simulated changes in means and extremes (present day − historical period) to Maurer and Livneh gridded observed precipitation products, considering every grid point within the CPM simulation domain. Boldface values indicate statistical significance at the 0.01 level.
Coherence metric for evaluating correspondence of long-term model-simulated changes in precipitation to equivalent changes in gridded observational precipitation data for (a),(b) Maurer and (c),(d) Livneh data. Results for coarse-resolution (35 km) model-simulated precipitation are shown for changes in (left) means and (right) extremes. A value of 1 indicates a perfect correspondence of the model-simulated result to the equivalent precipitation change from the source of observational data.
Citation: Journal of Applied Meteorology and Climatology 56, 9; 10.1175/JAMC-D-16-0358.1
As in Fig. 11, but for CPM (2.5 km) simulations.
Citation: Journal of Applied Meteorology and Climatology 56, 9; 10.1175/JAMC-D-16-0358.1
b. Change in model-simulated propagation of convection in conjunction with the diurnal cycle
Atmospheric models are generally challenged to represent the diurnal cycle of convective precipitation during the warm season (e.g., Trenberth et al. 2003). In the Southwest, the diurnal cycle of monsoon thunderstorms is intimately linked to mountain–valley circulations and propagation of thunderstorms off the terrain when upper-level winds are from a favorable direction (easterly in Arizona). To evaluate how the diurnal cycle of precipitation during severe weather event days has changed in the last 60 years, we use modeled results along with selected NCDC COOP observing stations. Only a very limited number of COOP stations in the Southwest have sufficient hourly precipitation records since the middle of the twentieth century to make such a comparison. Not surprising is that they are located in the largest cities in the southwestern U.S. region: PHX, Tucson (TUS), FLG, and Albuquerque (ABQ). PHX (337 m) is chosen in these comparisons to represent low-elevation stations because of the location downwind of mountain ranges, and FLG (2139 m) is chosen to represent higher-elevation stations in close proximity to Phoenix.
The diurnal cycle for the two represented stations is shown in Fig. 13. The horizontal axis in these figures is local time [mountain standard time (MST)], and 18 h of a convective day are displayed, starting at 1100 LT and ending at 0500 LT the next day. Hourly precipitation is accumulated from the previous hour to the time plotted. The vertical axis is plotted in millimeters per hour, albeit with different scaling for model and observations for two reasons: 1) we are more interested in whether the model simulation can reproduce the pattern of the diurnal cycle of precipitation and not necessarily the exact magnitude of precipitation at each hour because of the known negative (toward lower values) precipitation biases of the CPM as discussed earlier and 2) hourly observations for the historical past are not as reliable as those of the present day. These periods are denoted as blue and red lines, respectively, on the figures with model-simulation results.
Mean diurnal cycles in (top) Phoenix and (bottom) Flagstaff of (left) CPM (2.5 km) simulations and (right) NCDC COOP precipitation (mm h−1), given in local time (MST). Blue and red lines represent historical past and present day, respectively.
Citation: Journal of Applied Meteorology and Climatology 56, 9; 10.1175/JAMC-D-16-0358.1
We do not explicitly assess the statistical significance of the hourly differences between the two time periods here, although we do point out some physically interesting results. The observed severe weather event day precipitation evolution over Phoenix shows that most of the precipitation at this station occurs during the late-evening hours (2200–0200 LT). The precipitation maximum at this time reflects the passage of more-organized, westward-propagating convection through the city at night that develops during the previous day over the mountains, most predominantly over the Mogollon Rim to the north and east of the city. During the present-day period, this late-evening precipitation maximum (on the order of 0.10 mm h−1) shifts about 1–2 h earlier. Note that there is high variability in precipitation rate hour by hour, which may be an artifact of the sample size. The equivalent model results for PHX in the severe weather event simulations similarly show a nocturnal maximum in precipitation (at 2200 LT) in the historical past and a shifting of this maximum to occur approximately 5 h earlier (at 1700 LT) in the present day. Therefore, a more favorable thermodynamic environment in the present day seems to facilitate having thunderstorms occur in Phoenix earlier during the evening.
Precipitation at higher-elevation stations is more due to locally forced airmass-type monsoon thunderstorms than to organized convection. The peak intensity of observed precipitation at Flagstaff is during the early to midafternoon (1400–1500 LT). The change in timing of the maxima of CPM-simulated and observed precipitation is in agreement. They both indicate that peak rainfall hours are extended 1–2 h more to 1600–1700 LT. Figure 13 basically implies that there is a trend of convective precipitation occurring earlier in the valleys and lasting longer over the mountains.
The differences in CPM-simulated precipitation for the most extreme precipitation events between the present day and historical past over the course of the diurnal cycle are revealed every 6 h in Fig. 14 (1100–2300). This 12-h window is when the majority of convection occurs. The largest increases in CPM-simulated precipitation occur over the southwestern desert areas of Arizona in the middle of the day and in the afternoon (1100–1700 LT), when airmass-type thunderstorms are developing in a thermodynamically favorable environment. In contrast, the mountainous areas experience the largest CPM-simulated increases in precipitation in the late afternoon and evening (1700–2300 LT), when MCSs are most likely to occur. In the late evening to early morning (2300–0500 LT) there is little change in extreme precipitation (not shown), as organized monsoon convection typically dissipates by this time anyway. Thus, the long-term CPM-simulated significant changes in the diurnal cycle of precipitation also support the idea of earlier initiation of convection in lowland areas and greater persistence of convection over the mountains.
Significant changes (present day − historical past) in CPM (2.5 km) simulated extreme precipitation [mm (6 h−1)] every 6 h (a) from 1100 to 1700 LT and (b) from 1700 to 2300 LT. Statistical significance is computed at the 0.01 level.
Citation: Journal of Applied Meteorology and Climatology 56, 9; 10.1175/JAMC-D-16-0358.1
c. Changes in downdraft in associated with severe weather days
Another CPM-simulated variable that we consider in the context of extreme monsoon weather is downdraft velocity, measured by the metric of downdraft convective available potential energy (DCAPE). There are two reasons why DCAPE is of interest in reference to our study objectives. First, downdraft outflows help to maintain existing convection and trigger new convection via cold-pool dynamics. Convective downdrafts hitting the surface can provide a lifting mechanism to trigger new convection, as the authors have discussed in Lahmers et al. (2016). Second, wind gusts from downdraft outflows are associated with the severe weather hazards of microbursts and haboobs (e.g., Raman et al. 2014).
Significant changes (present day − historical past) in CPM (2.5 km) simulated (a) mean and (b) extremes of downdraft wind speed (m s−1), as computed from DCAPE. Statistical significance is computed at the 0.01 level.
Citation: Journal of Applied Meteorology and Climatology 56, 9; 10.1175/JAMC-D-16-0358.1
7. Discussion and conclusions
The overall objective of this study is to evaluate long-term changes in precipitation intensity during the North American monsoon in the southwestern United States, through the use of CPM simulations of objectively identified severe weather events during periods that represent the historical past (1950–70) and the present-day (1991–2010). The severe weather event CPM simulations appear to reasonably represent the diurnal cycle of convective precipitation during the period of the stage-IV product, in terms of the development of precipitation over the highest terrain during the day and convective organization and propagation into the evening hours. The comparisons with hourly precipitation data from NCDC COOP stations also show that the model simulations effectively capture the differences in the timing of convective precipitation during the monsoon in relation to elevation. CPM-simulated severe weather event precipitation tends to be slightly underestimated in comparison with the stage-IV product but is an improvement from the coarse-resolution simulations.
There has been an observed, statistically significant long-term increase in atmospheric moisture and instability in the Southwest over the past 60 years per trends in observed radiosonde data, which is realized in the long-term regional climate simulation that provides the boundary forcing to the CPM. Mean daily monsoon precipitation in the Southwest has generally decreased from the analysis of observations, whereas the most extreme monsoon precipitation has become more intense. This may be interpreted statistically as a broadening and flattening of the daily precipitation distribution during the period of 1950–2010. We observe a similar increasing intensity of extreme monsoon precipitation in the WRF severe weather event simulations, with the CPM simulations better resembling the equivalent changes in the Maurer and Livneh observed daily precipitation products.
In the context of the CPM simulations, the changes in monsoon precipitation intensity occur in association with the diurnal cycle of convection. Subsequent to their initiation over relatively high terrain, the simulated monsoon thunderstorms are able propagate sooner to the lower elevation in the more favorable thermodynamic environment of the present day when compared with the historical past. Extreme monsoon precipitation in the CPM is thus tending to last longer and be more intense in the present day. The largest observed and CPM-simulated increases in extreme-event precipitation occur in central and southwestern Arizona, focused on the afternoon and evening hours, in the area where MCSs account for a greater proportion of total monsoon precipitation. The downdraft winds within the most extreme convective events are also increasing, exacerbating the severe weather hazards of microbursts and dust storms (haboobs). In summary, we conclude that a more favorable thermodynamic environment during the last 30 years within the southwestern United States is facilitating stronger organized monsoon convection and that these types of changes can be much better realized within the context of a CPM. Future work will apply the same methodological approach to dynamically downscaled models of global climate change projection so as to more explicitly attribute the increases in monsoon precipitation intensity to anthropogenic global climate change.
There are caveats to the event-based CPM approach that is used in this study. Using the thermodynamic criteria to select the modeled severe weather event days will exclude severe weather events that occur in environments with more marginal instability or moisture (Mazon et al. 2016)—for example, heavy rainfall events that occur with tropical cyclone remnants (low instability) or dry microbursts with high-based convection (low moisture). Assessing the possible influence of slowly evolving land-surface feedback processes on monsoon precipitation (e.g., Hu and Dominguez 2015), which may also contribute to changes in precipitation extremes, would require RCM-type simulations on CPM scales.
Acknowledgments
This work was principally supported by the Strategic Environmental Research and Development Program (SERDP; Project RC-2205) through the U.S. Departments of Defense and Energy and the U.S. Environmental Protection Agency. Additional support was provided by UNAM-PAPIIT Projects IA103916 and IA100916; the Consortium for Arizona–Mexico Arid Environments (CAZMEX), with funding from the Consejo Nacional de Ciencia y Technología de México and The University of Arizona; and the University of Arizona Transboundary Aquifer Assessment Program (TAAP), authorized by Public Law 109-448, along with the University of Arizona Technology and Research Initiative Fund (TRIF). The comments from three anonymous reviewers substantially improved the quality of the manuscript. Various scientific materials and text in this paper were taken from the Ph.D. thesis of the first author, which can be found online (http://hdl.handle.net/10150/595660), and the final SERDP RC-2205 project report (Castro 2017).
REFERENCES
Adams, D. K., and A. C. Comrie, 1997: The North American monsoon. Bull. Amer. Meteor. Soc., 78, 2197–2213, doi:10.1175/1520-0477(1997)078<2197:TNAM>2.0.CO;2.
Adams, D. K., C. Minjarez, Y. Serra, A. Quintanar, L. Alatorre, A. Granados, E. Vázquez, and J. Braun, 2014: Mexican GPS tracks North American monsoon convection. Eos, Trans. Amer. Geophys. Union, 95, 61–62, doi:10.1002/2014EO070001.
Anderson, B. T., J. Wang, G. Salvucci, S. Gopal, and S. Islam, 2010: Observed trends in summertime precipitation over the southwestern United States. J. Climate, 23, 1937–1944, doi:10.1175/2009JCLI3317.1.
Archer, C. L., and K. Caldeira, 2008: Historical trends in the jet streams. Geophys. Res. Lett., 35, L08803, doi:10.1029/2008GL033614.
Bukovsky, M. S., D. J. Gochis, and L. O. Mearns, 2013: Towards assessing NARCCAP regional climate model credibility for the North American monsoon: Current climate simulations. J. Climate, 26, 8802–8826, doi:10.1175/JCLI-D-12-00538.1.
Bukovsky, M. S., C. M. Carrillo, D. J. Gochis, D. M. Hammerling, R. R. McCrary, and L. O. Mearns, 2015: Toward assessing NARCCAP regional climate model credibility for the North American Monsoon: Future climate simulations. J. Climate, 28, 6707–6728, doi:10.1175/JCLI-D-14-00695.1.
Castro, C. L., 2017: Assessing climate change impacts for DoD installations in the southwest United States during the warm season. Strategic Environmental Research and Development Program Final Rep. RC-2205, 113 pp. [Available online at https://www.serdp-estcp.org/Program-Areas/Resource-Conservation-and-Resiliency/Infrastructure-Resiliency/Vulnerability-and-Impact-Assessment/RC-2205.]
Castro, C. L., R. A. Pielke Sr., and J. O. Adegoke, 2007: Investigation of the summer climate of the contiguous United States and Mexico using the Regional Atmospheric Modeling System (RAMS). Part I: Model climatology (1950–2002). J. Climate, 20, 3866–3887, doi:10.1175/JCLI4211.1.
Castro, C. L., H.-I. Chang, F. Dominguez, C. Carrillo, J. Kyung-Schemm, and H. H.-M. Juang, 2012: Can a regional climate model improve warm season forecasts in North America? J. Climate, 25, 8212–8237, doi:10.1175/JCLI-D-11-00441.1.
Chang, H.-I., C. L. Castro, C. M. Carrillo, and F. Dominguez, 2015: The more extreme nature of U.S. warm season climate in the recent observational record and two “well-performing” dynamically downscaled CMIP3 models. J. Geophys. Res. Atmos., 120, 8244–8263, doi:10.1002/2015JD023333.
Craven, J. P., R. E. Jewell, and H. E. Brooks, 2002: Comparison between observed convective cloud-base heights and lifting condensation level for two different lifted parcels. Wea. Forecasting, 17, 885–890, doi:10.1175/1520-0434(2002)017<0885:CBOCCB>2.0.CO;2.
Dominguez, F., E. Rivera, D. P. Lettenmaier, and C. L. Castro, 2012: Changes in winter precipitation extremes for the western United States under a warmer climate as simulated by regional climate models. Geophys. Res. Lett., 39, L05803, doi:10.1029/2011GL050762.
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, doi:10.1175/1520-0469(1989)046<3077:NSOCOD>2.0.CO;2.
Durre, I., C. N. Williams Jr., X. Yin, and R. S. Vose, 2009: Radiosonde-based trends in precipitable water over the Northern Hemisphere: An update. J. Geophys. Res., 114, D015 112, doi:10.1029/2008JD010989.
Garfin, G., A. Jardine, R. Merideth, M. Black, and S. Leroy, Eds., 2013: Assessment of Climate Change in the Southwest United States: A Report Prepared for the National Climate Change Assessment. Island Press, 531 pp.
Geil, K., Y. L. Serra, and X. Zeng, 2013: Assessment of CMIP5 model simulations of the North American monsoon system. J. Climate, 26, 8787–8801, doi:10.1175/JCLI-D-13-00044.1.
Groisman, P. Ya., R. W. Knight, D. R. Easterling, T. R. Karl, G. C. Hegerl, and V. N. Razuvaev, 2005: Trends in intense precipitation in the climate record. J. Climate, 18, 1326–1350, doi:10.1175/JCLI3339.1.
Grossman-Clarke, S., J. A. Zehnder, T. Loridan, and C. S. B. Grimmond, 2010: Contribution of land use changes to near surface air temperatures during recent summer extreme heat events in the Phoenix metropolitan area. J. Appl. Meteor. Climatol., 49, 1649–1664, doi:10.1175/2010JAMC2362.1.
Higgins, R. W., J. E. Janowiak, and Y.-P. Yao, 1996: A gridded hourly precipitation data base for the United States (1963–1993). NCEP/Climate Prediction Center Atlas 1, NOAA NWS, 47 pp. [Available online at http://www.cpc.ncep.noaa.gov/research_papers/ncep_cpc_atlas/1/cover.html.]
Hu, H., and F. Dominguez, 2015: Evaluation of oceanic and terrestrial sources of moisture for the North American monsoon using numerical models and precipitation stable isotopes. J. Hydrometeor., 16, 19–35, doi:10.1175/JHM-D-14-0073.1.
IPCC, 2013: Climate Change 2013: The Physical Science Basis. Cambridge University Press, 1535 pp., doi:10.1017/CBO9781107415324.
Janjić, Z. I., 1990: The step-mountain coordinate: Physical package. Mon. Wea. Rev., 118, 1429–1443, doi:10.1175/1520-0493(1990)118<1429:TSMCPP>2.0.CO;2.
Janjić, Z. I., 1996: The surface layer in the NCEP Eta Model. 11th Conf. on Numerical Weather Prediction, Norfolk, VA, Amer. Meteor. Soc., 354–355.
Janjić, Z. I., 2001: Nonsingular implementation of the Mellor-Yamada level 2.5 scheme in the NCEP Meso Model. NCEP Office Note 437, 61 pp. [Available online at http://www.emc.ncep.noaa.gov/officenotes/newernotes/on437.pdf.]
Kain, J. S., 2004: The Kain–Fritsch convective parameterization: An update. J. Appl. Meteor., 43, 170–181, doi:10.1175/1520-0450(2004)043<0170:TKCPAU>2.0.CO;2.
Kain, J. S., and J. M. Fritsch, 1993: Convective parameterization for mesoscale models: The Kain–Fritsch scheme. The Representation of Cumulus Convection in Numerical Models, Meteor. Monogr., No. 46, Amer. Meteor. Soc., 165–170.
Kalnay, E., and Coauthors, 1996: The NCEP/NCAR 40-Year Reanalysis Project. Bull. Amer. Meteor. Soc., 77, 437–471, doi:10.1175/1520-0477(1996)077<0437:TNYRP>2.0.CO;2.
Karl, T. R., and R. W. Knight, 1998: Secular trends in the precipitation amount, frequency, and intensity in the United States. Bull. Amer. Meteor. Soc., 79, 231–241, doi:10.1175/1520-0477(1998)079<0231:STOPAF>2.0.CO;2.
Karl, T. R., and K. E. Trenberth, 2003: Modern climate change. Science, 302, 1719–1723, doi:10.1126/science.1090228.
Katz, R. W., M. B. Parlange, and P. Naveau, 2002: Statistics of extremes in hydrology. Adv. Water Resour., 25, 1287–1304, doi:10.1016/S0309-1708(02)00056-8.
Kendon, E., and Coauthors, 2017: Do convection-permitting regional climate models improve projections of future precipitation change? Bull. Amer. Meteor. Soc., 98, 79 –93, doi:10.1175/BAMS-D-15-0004.1.
Kharin, V., and F. Zwiers, 2005: Estimating extremes in transient climate change simulations. J. Climate, 18, 1156–1173, doi:10.1175/JCLI3320.1.
Kunkel, K. E., and Coauthors, 2013: Monitoring and understanding trends in extreme storms: State of knowledge. Bull. Amer. Meteor. Soc., 94, 499–514, doi:10.1175/BAMS-D-11-00262.1.
Lahmers, T. M., C. L. Castro, D. K. Adams, Y. L. Serra, J. J. Brost, and T. Luong, 2016: Long-term changes in the climatology of transient inverted troughs over the North American monsoon region and their effects on precipitation. J. Climate, 29, 6037–6064, doi:10.1175/JCLI-D-15-0726.1.
Livneh, B., E. A. Rosenberg, C. Lin, B. Nijssen, V. Mishra, K. M. Andreadis, E. P. Maurer, and D. P. Lettenmaier, 2013: A long-term hydrologically based dataset of land surface fluxes and states for the conterminous United States: Update and extensions. J. Climate, 26, 9384–9392, doi:10.1175/JCLI-D-12-00508.1.
Lu, J., C. Deser, and T. Reichler, 2009: Cause of the widening of the tropical belt since 1958. Geophys. Res. Lett., 36, L03803, doi:10.1029/2009GL038880.
Maddox, R. A., D. M. McCollum, and K. W. Howard, 1995: Large-scale patterns associated with severe summertime thunderstorms over central Arizona. Wea. Forecasting, 10, 763–778, doi:10.1175/1520-0434(1995)010<0763:LSPAWS>2.0.CO;2.
Maurer, E. P., A. W. Wood, J. C. Adam, D. P. Lettenmaier, and B. Nijssen, 2002: A long-term hydrologically based dataset of land surface fluxes and states for the conterminous United States. J. Climate, 15, 3237–3251, doi:10.1175/1520-0442(2002)015<3237:ALTHBD>2.0.CO;2.
Mazon, J. J., C. L. Castro, D. K. Adams, H. Chang, C. M. Carrillo, and J. J. Brost, 2016: Objective climatological analysis of extreme weather events in Arizona during the North American monsoon. J. Appl. Meteor. Climatol., 55, 2431–2450, doi:10.1175/JAMC-D-16-0075.1.
Meehl, G. A., F. Zwiers, J. Evans, T. Knutson, L. Mearns, and P. Whetton, 2000: Trends in extreme weather and climate events: Issues related to modeling extremes in projections of future climate change. Bull. Amer. Meteor. Soc., 81, 427–436, doi:10.1175/1520-0477(2000)081<0427:TIEWAC>2.3.CO;2.
Min, S.-K., X. Zhang, F. W. Zwiers, P. Friederichs, and A. Hense, 2009: Signal detectability in extreme precipitation changes assessed from twentieth century climate simulations. Climate Dyn., 32, 95–111, doi:10.1007/s00382-008-0376-8.
Min, S.-K., X. Zhang, F. W. Zwiers, and G. C. Hegerl, 2011: Human contribution to more-intense precipitation extremes. Nature, 470, 378–381, doi:10.1038/nature09763.
Minjarez-Sosa, C. M., C. L. Castro, K. L. Cummins, J. Waissmann, and D. K. Adams, 2017: An improved QPE over complex terrain employing cloud-to-ground lightning occurrences. J. Appl. Meteor. Climatol., 56, 2489–2507, doi:10.1175/JAMC-D-16-0097.1.
Mlawer, E. J., S. J. Taubman, P. D. Brown, M. J. Iacono, and S. A. Clough, 1997: Radiative transfer for inhomogeneous atmospheres: RRTM, a validated correlated-k model for the longwave. J. Geophys. Res., 102, 16 663–16 682, doi:10.1029/97JD00237.
Niu, G.-Y., and Coauthors, 2011: The community Noah land surface model with multiparameterization options (Noah-MP): 1. Model description and evaluation with local-scale measurements. J. Geophys. Res., 116, D12109, doi:10.1029/2010JD015139.
Petrie, M., S. Collins, D. Gutzler, and D. Moore, 2014: Regional trends and local variability in monsoon precipitation in the northern Chihuahuan Desert, USA. J. Arid Environ., 103, 63–70, doi:10.1016/j.jaridenv.2014.01.005.
Prein, A. F., and Coauthors, 2015: A review on regional convection-permitting climate modeling: Demonstrations, prospects, and challenges. Rev. Geophys., 53, 323–361, doi:10.1002/2014RG000475.
Raman, A., A. F. Arellano, and J. J. Brost, 2014: Revisiting haboobs in the southwestern United States: An observational case study of the 5 July 2011 Phoenix dust storm. Atmos. Environ., 89, 179–188, doi:10.1016/j.atmosenv.2014.02.026.
Rivera, E., F. Dominguez, and C. L. Castro, 2014: Atmospheric rivers and extreme cool season precipitation events in the Verde River basin of Arizona. J. Hydrometeor., 15, 813–829, doi:10.1175/JHM-D-12-0189.1.
Santer, B. D., and Coauthors, 2007: Identification of human-induced changes in atmospheric moisture content. Proc. Natl. Acad. Sci. USA, 104, 15 248–15 253, doi:10.1073/pnas.0702872104.
Seidel, D. J., Q. Fu, W. J. Randel, and R. J. Reichler, 2008: Widening of the tropical belt in a changing climate. Nat. Geosci., 1, 21–24, doi:10.1038/ngeo.2007.38.
Skamarock, W. C., and Coauthors, 2008: A description of the Advanced Research WRF version 3. NCAR Tech. Note NCAR/TN-475+STR, 113 pp., doi:10.5065/D68S4MVH.
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, doi:10.1175/1520-0493(2004)132<0519:EFOWPU>2.0.CO;2.
Trenberth, K., A. Dai, R. Rasmussen, and D. Parsons, 2003: The changing character of precipitation. Bull. Amer. Meteor. Soc., 84, 1205–1217, doi:10.1175/BAMS-84-9-1205.
Tripathi, O. P., and F. Dominguez, 2013: Effects of spatial resolution in the simulation of daily and subdaily precipitation in the southwestern US. J. Geophys. Res. Atmos., 118, 7591–7605, doi:10.1002/jgrd.50590.
Truong, N. M., T. T. Tien, R. A. Pielke Sr., C. L. Castro, and G. Leoncini, 2009: A modified Kain–Fritsch scheme and its application for simulation of an extreme precipitation event in Vietnam. Mon. Wea. Rev., 137, 766–789, doi:10.1175/2008MWR2434.1.
Willett, K. M., N. P. Gillett, P. D. Jones, and P. W. Thorne, 2007: Attribution of observed surface humidity changes to human influence. Nature, 449, 710–712, doi:10.1038/nature06207.
