SEPTEMBER 2017 J I N G E T A L . 2541

Evaluating Seasonal Orographic in the Interior Using Gauge Data, Gridded Precipitation Estimates, and a Regional Simulation

XIAOQIN JING AND BART GEERTS Department of Atmospheric Science, University of Wyoming, Laramie, Wyoming

YONGGANG WANG Department of Geosciences, Texas Tech University, Lubbock, Texas

CHANGHAI LIU National Center for Atmospheric Research, Boulder, Colorado

(Manuscript received 31 March 2017, in final form 12 July 2017)

ABSTRACT

There are several high-resolution (1–12 km) gridded precipitation datasets covering the interior western United States. This study cross validates seasonal orographic precipitation estimates from the Snowpack Telemetry (SNOTEL) network; the national hourly multisensor precipitation analysis Stage IV dataset (NCEP IV); four gauge-driven gridded datasets; and a 10-yr, 4-km, convection-permitting Weather Research and Forecasting (WRF) Model simulation. The NCEP IV dataset, which uses the NEXRAD network and precipitation gauges, is challenged in this region because of blockage and lack of low-level radar coverage in complex terrain. The gauge-driven gridded datasets, which statistically interpolate gauge measurements over complex terrain to better estimate orographic precipitation, are challenged by the highly heterogeneous, weather-dependent nature of precipitation in complex terrain at scales finer than can be resolved by the gauge network, such as the SNOTEL network. Gauge-driven gridded precipitation estimates disagree in areas where SNOTEL gauges are sparse, especially at higher elevations. The WRF simulation captures wintertime orographic precipitation distribution and amount well, and biases over specific mountain ranges are identical to those in an independent WRF simulation, suggesting that these biases are at least partly due to errors in the snowfall measurements or the gridding of these measurements. The substantial disagreement between WRF and the gridded datasets over some mountains may motivate reevaluation of some gauge records and in- stallation of new SNOTEL gauges in regions marked by large discrepancies between modeled and gauge- driven precipitation estimates.

1. Introduction hydrology, agriculture, forestry, and ecology (Mote et al. 2005; Bales et al. 2006; Ebert et al. 2007; Barnett et al. The interior western United States (IWUS)1 is mostly 2008; Rasmussen et al. 2011). Most of the precipitation arid, but is also home to the headwaters of several major over the IWUS falls as over its mountains (Daly river systems, for example, the Colorado, Missouri, and et al. 1994). QPE is especially challenging in complex Snake Rivers (Woodhouse 2004). There is much interest terrain (e.g., Liu et al. 2011), and gauge-based snowfall in quantitative precipitation estimation (QPE) in this rate estimation is more uncertain than rain rate estima- region, in a range of disciplines including , tion (Rasmussen et al. 2012). In the mountainous IWUS, the Snowpack Telemetry (SNOTEL) network, operated by the Natural Resources Conservation Service (NRCS), 1 In this study, IWUS includes south Montana, east Idaho, has been used as a reference in many studies to evaluate Wyoming, Utah, Colorado, north Arizona, and north New Mexico. model output (e.g., Liu et al. 2011; Gutmann et al. 2012) and serves as the basis for gridded precipitation estimates Corresponding author: Xiaoqin Jing, [email protected] over mountains. However, SNOTEL only provides point

DOI: 10.1175/JHM-D-17-0056.1 Ó 2017 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses). Unauthenticated | Downloaded 10/04/21 04:28 AM UTC 2542 JOURNAL OF HYDROMETEOROLOGY VOLUME 18 measurements of precipitation, and the gauge density developed to study the precipitation climatology (e.g., is low compared to that in highly populated or agricul- Lin and Mitchell 2005; Hou et al. 2014). Datasets de- tural regions. Several different techniques have been veloped using ground-based scanning weather radars, developed to provide more complete precipitation dis- such as the National Centers for Environmental Pre- tribution maps, including ‘‘terrain aware’’ interpolation diction hourly multisensor precipitation analysis Stage IV techniques using gauge measurements as forcing data (NCEP IV) dataset (Lin and Mitchell 2005), are quite (e.g., Daly et al. 1994), space-based and ground-based suitable to study the precipitation distribution at high remote sensing retrievals (e.g., Lin and Mitchell 2005), spatial resolution (4 km) over relatively flat terrain, such and numerical model simulations (Liu et al. 2017). But as in the central and eastern United States. However, the the relative performances of different techniques in QPE operational network is challenged over the are not well understood, so a cross validation of different complex terrain environment of the western United precipitation datasets is necessary. States because of blockage by the first range of mountains The terrain-aware interpolation techniques have been and the inability to capture the low-level orographic widely used to study the precipitation distribution over precipitation growth zone if the lowest unblocked beam is IWUS (e.g., Daly et al. 1994; Thornton et al. 1997; Xia et al. high above the surface, which is common (e.g., Fulton 2012; Newman et al. 2015a). Daly et al. (1994) developed et al. 1998; Maddox et al. 2002; Lin and Mitchell 2005; Lin the Parameter-Elevation Regressions on Independent and Hou 2012; Smalley et al. 2014). A radar network Slopes Model (PRISM) to produce a terrain-sensitive much denser than that currently in operation in the gridded precipitation dataset, which remains widely used. IWUS would be needed to achieve NCEP IV pre- This model estimates the precipitation in areas without cipitation accuracies comparable to those over the cen- gauges using physically informed statistical relations tral/eastern United States. Space-based remote sensing between terrain and gauge precipitation. Several other has the advantage of vertical incidence and thus no gauge-driven datasets have been developed using different blockage issues. Precipitation estimates using passive statistical downscaling techniques, including Daymet remote sensing techniques cannot reveal the variable (Thornton et al. 1997), North American Land Data As- precipitation distribution over complex terrain (Ebert similation System Stage II (NLDAS II; Xia et al. 2012), et al. 2007). Active space-based radar measurements, and the continental United States ensemble gridded da- such as Global Precipitation Measurement (GPM, from tasets (CUSEG; Newman et al. 2015a). 2014 to present; Hou et al. 2014), are inadequate to study The uncertainties of the gauge-driven gridded datasets the finescale quantitative precipitation climatology now have been discussed in several previous studies (e.g., Daly because of lack of overpasses, but they could be useful in et al. 2008; Gutmann et al. 2012). For example, Daly et al. the future. (2008) tried to estimate the error of PRISM using a data Recently, numerical weather prediction (NWP) denial cross-validation method. In this method, the models with high resolution (,6 km) have been used to PRISM regression with a single gauge removed is com- study the precipitation climatology over complex ter- pared to that with all gauges in the vicinity of the gauge rain. Ikeda et al. (2010) showed that the Weather Re- site. They show that the resulting difference in annual search and Forecasting (WRF) Model at a grid spacing precipitation estimates is 20%–30% over mountains. smaller than 6 km well captures the seasonal snowfall in Daly et al. (2008) also shows PRISM precipitation has a the Colorado Rockies; the difference of cold-season larger uncertainty in winter than in summer. Gutmann precipitation between the model output and SNOTEL is et al. (2012) compared the winter precipitation estimates within 20% for 71% of the SNOTEL sites. Liu et al. from PRISM against a SNOTEL site in the eastern San (2011) pointed out that this performance is highly sen- Juan Mountains at Moon Pass. The SNOTEL gauge was sitive to the choice of cloud microphysics parameteri- installed in October 2008 and was not applied in the zation. Rasmussen et al. (2011, 2014) further confirmed PRISM used in their study. The comparison indicates that WRF, at a resolution of 4 km, captures the cold- PRISM significantly overestimates the winter pre- season precipitation distribution and amount over the cipitation (600 mm) compared to SNOTEL (232 mm) at Colorado headwaters region well, with a bias of 10%–15% that point. Recently, Henn et al. (2017) compared the compared to SNOTEL measurements. Because of precipitation trend estimates from gauge-driven gridded the good performance in simulating orographic precipi- datasets against the streamflow observations in Sierra tation over complex terrain, high-resolution WRF Nevada and suggests the gauge-driven gridded datasets simulations have been used to assess changes of oro- may have substantial uncertainty at high elevations. graphic precipitation in a changing global climate. For Other than gauge-driven gridded datasets, space-based instance, Rasmussen et al. (2011, 2014) analyzed the and ground-based remote sensing techniques have been hydrological cycle in the Colorado headwaters region

Unauthenticated | Downloaded 10/04/21 04:28 AM UTC SEPTEMBER 2017 J I N G E T A L . 2543 using high-resolution WRF simulations and explored its sensitivity to climate change using a pseudo–global warming approach. Liu et al. (2017) recently extended the work to the continental United States (hereafter CONUS WRF). Gridded precipitation datasets developed using differ- ent techniques have been widely used for various purposes because of their completeness (Lundquist et al. 2015), but their accuracy is not well known. Biases in the gridded precipitation datasets have impacts on the evaluation of hydrologic and climate model simulations and would in- fluence the accuracy of model simulations if they are used as forcing data. Some previous studies have compared different precipitation datasets (e.g., Ebert et al. 2007), but few have specially focused on the mountainous IWUS, and even fewer used high-resolution datasets (Henn et al. 2017). This study aims to cross validate high-resolution precipitation datasets currently available in the IWUS and a 10-yr continuous convection-permitting WRF sim- ulation at 4-km horizontal resolution (hereafter IWUS WRF). The focus is on orographic precipitation across seasons. The model output is compared with gauge data (SNOTEL), a radar-based dataset (NCEP IV), and four gauge-driven gridded datasets (PRISM, Daymet, NLDAS II, CUSEG). IWUS WRF is also compared with CONUS WRF to examine the impact of different driven datasets and boundaries on simulations. The paper is organized as follows. Section 2 describes the WRF Model configuration and introduces the observational datasets. The model output and datasets are compared and statistically ana- lyzed in section 3. Discussion and conclusions are given in sections 4 and 5, respectively.

2. Description of the WRF configuration and observational datasets a. WRF configuration FIG. 1. Topography of (a) the WRF Model domain and (b) the study domain. In (b), the three red boxes delineate subdomains A 10-yr continuous simulation is conducted using the referred to as the Greater Yellowstone (upper), Utah (lower left), WRF Model version 3.7.1 to study the precipitation and and Colorado (lower right). The dark gray circles indicate the lo- snowpack in the IWUS. The Climate Forecast System cations of SNOTEL sites and the gray line is the 2-km height MSL contour. Several mountain ranges are marked in (b). GV, WRR, Reanalysis (CFSR; Saha et al. 2010) is used to provide the and SRR are abbreviations for Gros Ventre, River Range initial and lateral boundary conditions. The non-nested and Salt River Range, respectively. model domain, shown in Fig. 1a,has4203 410 grid points at 4-km horizontal resolution and 51 vertical levels, with a high layer density close to the ground. The simulation examine the impact of different input datasets and runs from October 2001 to February 2012, and the output boundaries. The selection of physical schemes follows from March 2002 to February 2012 is used to study the CONUS WRF: the Rapid Radiative Transfer Model for precipitation patterns in the IWUS. This is different from General Circulation Models (RRTMG) shortwave and CONUS WRF, which covers the continental United longwave radiation scheme (Iacono et al. 2008), the re- States and runs from October 2000 to September 2013 vised Monin–Obukhov surface layer scheme (Jimenez using ERA-Interim reanalysis data as input. Comparison et al. 2012), the Noah-MP land surface scheme (Niu et al. between IWUS WRF and CONUS WRF will be made to 2011 and Yang et al. 2011), the Yonsei University (YSU)

Unauthenticated | Downloaded 10/04/21 04:28 AM UTC 2544 JOURNAL OF HYDROMETEOROLOGY VOLUME 18 planetary boundary layer (PBL) scheme (Hong and Pan In this study, we also use daily SNOTEL gauge measure- 1996), and the Thompson cloud microphysics scheme ments as the reference to evaluate the WRF simulation (Thompson et al. 2004). Convection is not parameterized. A and the various gridded datasets. 30-yr (1981–2011) retrospective simulation is being con- The PRISM dataset (PRISM Climate Group 2016)is ducted based on the same configuration, as a basis to study based on a statistical model developed to interpolate changes in orographic precipitation in a changing climate climate elements to finescale complex terrain. We use (Wang et al. 2017, manuscript submitted to Int. J. Climatol.). the publically available monthly data at 4-km resolu- This study focuses on complex terrain in a subregion tion (Table 1), although higher-resolution PRISM data of the model domain, here called the ‘‘study domain,’’ are available commercially. Basically, PRISM uses a shown in Fig. 1a. We will distinguish three main moun- precipitation-elevation regression and takes into account tainous regions within this study domain (Fig. 1b): the other factors such as terrain slope, coastal proximity, and ‘‘Greater Yellowstone’’ region, the central Utah range topographic facet orientation to estimate the precipita- (called ‘‘Utah’’), and the central Rocky Mountain re- tion at each digital elevation model grid point (Daly et al. gion (called ‘‘Colorado’’). Most of the precipitation in 1994, 2002, 2008). Precipitation data are based on various IWUS falls in areas 2 km above mean sea level (MSL), gauge networks. The National Weather Service Cooper- contoured in Fig. 1b. The SNOTEL sites shown in ative Observer Program (COOP) gauge network (Daly Fig. 1b are those in operation between March 2002 and et al. 2007) and SNOTEL are the two main sources of February 2012. Data from SNOTEL sites installed after precipitation data used in PRISM. The density of the March 2002 are not used in this study. COOP station network relates to population density and farming intensity. Towns and agricultural areas occupy b. Datasets description the lower-elevation plains and valleys in the IWUS. The Other than the WRF simulation, we compare six SNOTEL data largely control PRISM precipitation in precipitation datasets in this study: SNOTEL, PRISM, the mountains. The PRISM precipitation climatology in Daymet, NLDAS II, CUSEG, and NCEP IV. Basic in- the United States has been used as a reference in many formation for the six datasets is summarized in Table 1. studies (e.g., Liu et al. 2017). Uncertainty of PRISM pre- SNOTEL (NRCS 1977) is an important precipitation cipitation can be attributed to the quality of gauge data, network across the IWUS mountains (Doesken and especially for snow, the precipitation-elevation regression, Schaefer 1987; Serreze et al. 1999). A standard sensor the density of gauge stations, the complexity of terrain, and configuration includes a pressure-sensing snow pillow (to assumptions made in the model (Daly et al. 2008). estimate snow water equivalent), a storage precipitation Daymet (Thornton et al. 2014), NLDAS II (NCEP gauge, and an air sensor. The daily difference 2014), and CUSEG (Newman et al. 2015b) are all gauge- in snow water equivalent can be used as a measure of daily driven gridded datasets, similar to PRISM, but they snowfall accumulation. We do not use those data in this use different statistical interpolation techniques. In study because we are interested in all precipitation (not Daymet, a terrain-dependent spatial convolution of a just snow) and because of uncertainties related to snow truncated Gaussian weighting filter is used to interpolate drift and sublimation. Here we use daily precipitation es- gauge data to 1-km resolution grids (Thornton et al. timates from the storage gauges. These gauges have a wide 1997). In NLDAS II, the Climate Prediction Center orifice and a wind shield to maximize catch efficiency. The (CPC) gauge-based daily precipitation dataset at 0.1258 SNOTEL sites typically are in small openings in forests to resolution is used as the initial input. PRISM is used minimize snow drifts and thus to optimize the estimation to adjust the magnitude, and NCEP Stage II data are of the seasonal evolution of snow water equivalent. then used to temporally disaggregate the dataset to As a result, the SNOTEL site density is not uniform with hourly scale (Xia et al. 2012). CUSEG is a 100-member elevation across a mountain range (Fig. 1b): there are al- ensemble daily precipitation dataset. A probabilistic in- most no SNOTEL sites above the tree line, which repre- terpolation is used to interpolate gauge data to 0.1258 sents large mountain areas in the Wind River Range or resolution grids. Terrain impacts (e.g., elevation and San Juan Mountains, for instance. On the other hand, slope) are included (Newman et al. 2015a). The ensemble SNOTEL sites are found only near mountain tops in drier, approach accounts for the probability density function lower-elevation regions, for example, in Utah west of the (PDF) of uncertainties due to spatial undersampling and Wasatch Range (Fig. 1b). The SNOTEL sensors record subsequent extrapolation, projecting irregular measure- data every 15 min; hourly and daily data are archived. ments onto a grid, and also implicitly accounts for random SNOTEL has been widely used as the ‘‘truth’’ to evaluate measurement errors based on the mean monthly pre- other precipitation datasets and model simulations (e.g., cipitation and PDF of daily precipitation estimates from Daly et al. 2008; Rasmussen et al. 2011; Liu et al. 2011). gauges (Clark and Slater 2006).

Unauthenticated | Downloaded 10/04/21 04:28 AM UTC SEPTEMBER 2017 J I N G E T A L . 2545

TABLE 1. Information about WRF and the precipitation datasets used in this study.

Spatial Temporal resolution resolution Duration Type Reference SNOTEL Single points Daily 1979–present Gauge Serreze et al. (1999) PRISM 4 km Monthly 1981–present Gauge and statistical model Daly et al. (1994) Daymet 1 km Daily 1980–2015 Gauge and statistical model Thornton et al. (1997) NLDAS II 0.1258 (;12 km) Hourly 1979–present Gauge, ground-based radar Xia et al. (2012) and statistical model CUSEG 0.1258 (;12 km) Daily 1980–2012 Gauge and statistical model Newman et al. (2015a) NCEP IV 4 km Hourly 2002–present Ground-based radar and gauge Lin and Mitchell (2005) WRF 4 km Hourly March 2002–February 2012 NWP model

The NCEP IV dataset (NCEP 2001) uses merged WRF estimates more precipitation than the other datasets surface radar and products to produce hourly over the Wind River Range and some plains areas, es- precipitation at 4-km resolution (Lin and Mitchell pecially over the high plains of eastern Wyoming; in fact, 2005). The finescale distribution of precipitation is the jWRF 2 SNOTELj difference in this region tends to controlled by the hourly radar rainfall products from exceed the mean absolute difference between gauge- about 140 Weather Surveillance Radar-1988 Doppler driven gridded datasets (as shown by the hatches in (WSR-88D) radars, without direct adjustment for ter- Fig. 2b). This bias is most pronounced in spring and rain (Lin and Mitchell 2005). The radar-estimated summer. Overall, WRF estimates 55% of the annual precipitation is then merged with about 3000 hourly precipitation falls on mountain areas or high plains above gauge reports using the methodology developed by Seo 2000 m, and 27% falls on mountain ranges above 2500 m. (1998). This dataset has been widely used to evaluate NCEP IV underestimates the annual precipitation in other precipitation products using remote sensing most of the areas, especially over the mountains, although techniques (e.g., Gourley et al. 2010; Tesfagiorgis et al. it does capture the general pattern of orographic pre- 2011; Lin and Hou 2012). cipitation (Fig. 2g). NCEP IV orographic precipitation comes closest to other datasets within ;100 km range of WSR-88D radars, for example, in the Wasatch Range, 3. Results but is a gross underestimation over remote plains and radar-blocked mountains, for example, the Bighorn or a. Spatial precipitation distribution the La Sal Ranges. The annual mean precipitation maps over the IWUS, The annual-mean data in Fig. 2 are partitioned in four as estimated by SNOTEL, WRF, PRISM, Daymet, seasons in Fig. 3, since precipitation processes are highly NLDAS II, CUSEG, and NCEP IV over the decade dependent on the time of year. Among the four gauge- from March 2002 to February 2012, are shown in Fig. 2. driven gridded datasets, only PRISM is shown in this The terrain is gray-shaded in Fig. 2 to highlight the figure; Daymet, NLDAS II, and CUSEG provide similar mountain ranges and to emphasize the orographic na- patterns compared to PRISM. The mountains west of the ture of precipitation in the IWUS. Not surprisingly, continental divide receive most of their precipitation in PRISM, Daymet, NLDAS II, and CUSEG appear to be winter and experience dry summers, while the high plains quite consistent with SNOTEL (Figs. 2c–f). Relatively are wetter in spring and summer than in fall and winter. high-resolution datasets (PRISM and Daymet) show During all seasons PRISM,Daymet,NLDASII,and finer texture of terrain-relative precipitation distribu- CUSEG are quite consistent with SNOTEL over moun- tion than coarser-resolution datasets (NLDAS II and tain areas, which is no surprise. In areas without gauges, CUSEG) and can resolve orographic precipitation over their accuracy is unknown. The WSR-88D network can relatively small mountains (e.g., the Salt River, Wyom- hardly detect shallow orographic storms (Smalley et al. ing, and La Sal Ranges; see Fig. 1b for the location of 2014), which prevail in winter (Fig. 3d4). Thus, NCEP IV these mountain ranges). seriously underestimates winter precipitation over The WRF precipitation distribution compares well mountains remote from any radar sites, such as the Big- against SNOTEL (Figs. 2a,b) and shows similar patterns horn and Park Ranges (Fig. 3d4). During summer, most against PRISM, Daymet, NLDAS II, and CUSEG. But precipitation is generated through deep convection, which WRF appears to underestimate the annual precipitation is detected over a greater range by radars (Fig. 3d2). In over the Wasatch, which is relatively narrow and steep. the high plains as well as in the adjacent mountains in

Unauthenticated | Downloaded 10/04/21 04:28 AM UTC 2546 JOURNAL OF HYDROMETEOROLOGY VOLUME 18

FIG. 2. Mean annual precipitation maps from March 2002 to February 2012 estimated by (a) SNOTEL, (b) WRF, (c) PRISM, (d) Daymet, (e) NLDAS II, (f) CUSEG, and (g) NCEP IV. The hatched areas depict regions where the absolute difference between WRF and PRISM exceeds the absolute mean difference between PRISM and any other gauge-driven dataset (Daymet, NLDAS II, and the 100 CUSEG datasets; 103 in total). The stars in (g) show the location of WSR-88D radars, and the circles show the 100-km range ring.

Unauthenticated | Downloaded 10/04/21 04:28 AM UTC SEPTEMBER 2017 J I N G E T A L . 2547

FIG. 3. Mean seasonal precipitation maps estimated by (a1)–(a4) SNOTEL, (b1)–(b4) WRF, (c1)–(c4) PRISM, and (d1)–(d4) NCEP IV. Panels from left to right are precipitation maps in spring (MAM), summer (JJA), fall (SON), and winter (DJF). Hatched areas are defined as in Fig. 2.

Unauthenticated | Downloaded 10/04/21 04:28 AM UTC 2548 JOURNAL OF HYDROMETEOROLOGY VOLUME 18

Wyoming and Colorado, WRF appears to estimate more Wasatch Range, while the positive biases in the lee may be spring and summer precipitation than SNOTEL, mainly in due to SNOTEL snowfall underestimates in the presence spring (MAM) over the high plains and in summer (JJA) of strong surface (Rasmussen et al. 2012), which are over the mountains. This may relate to the inability of the common in the lee foothills. Figure 5f shows the WRF 2 model with a resolution of 4 km to accurately capture SNOTEL bias of monthly precipitation in winter as a convective development and upscale growth (Rasmussen function of wind speed; this is the WRF-derived prevailing et al. 2014). In addition, our WRF simulation uses the same surface wind during precipitation events. The figure sug- configuration in all seasons, which may not be optimal gests the mean difference between WRF and SNOTEL is 2 (e.g., Ebert et al. 2007). very small. For strong winds (.20 m s 1), WRF estimates slightly more snowfall than SNOTEL, and strong winds b. Evaluation of WRF simulation are more common at lee gauges than at upwind gauges. The good performance of high-resolution WRF in ThismaybebecauseSNOTELunderestimates the snow- simulating orographic precipitation, especially winter- fall due to strong surface wind (Rasmussen et al. 2012), but time precipitation, is evident in previous studies (section more studies are needed in the future to better understand 1). In this section, we provide an additional evaluation of the uncertainty of snow gauges and WRF. the WRF simulation to discuss the terrain-related dif- There is no clear pattern either in the bias or in the ferences between WRF and SNOTEL. In addition, correlation coefficient; neither parameter relates with ter- a comparison between IWUS WRF and CONUS WRF rain elevation (Figs. 5c,d). The mean bias is close to 0, and examines the impact of different driven datasets. the standard deviation is smaller than 40 mm, irrespective Moreover, a comparison between simulations at 4- and of elevation, but the precipitation bias (WRF 2 SNOTEL) 1.33-km resolutions is made to examine the impact of is correlated to the height difference between WRF boundaries (i.e., domain size) and resolutions. (model terrain height) and SNOTEL (actual height) Scatterplots of seasonal precipitation estimated by (Fig. 5e). This reflects a resolution-related weakness in WRF against SNOTEL in the entire study domain are the model: at a resolution of 4 km, WRF underestimates shown in Fig. 4. The data are distributed well on both sides (overestimates) precipitation at SNOTEL sites located of the 1:1 lines during spring and winter, whereas in on a local mountain (in a local valley or terrain concav- summer WRF overestimates the precipitation by 30 mm ity). The monthly precipitation bias between WRF and and in fall WRF underestimates the precipitation by SNOTEL (Figs. 5c,e) indicates WRF may have larger 32 mm on average. The WRF root-mean-square bias uncertainty than the mean bias at an individual SNOTEL (RMSB) is larger than 55 mm in all the seasons. The small site; this is probably because the same model configura- mean biases and large RMSBs in spring and winter reflect tion is used for the 10-yr simulation, which is not optimal. both uncertainties in gauge measurement and small-scale A simulation with a resolution finer than 4 km may terrain effects that are unresolved in WRF. better resolve the precipitation over complex terrain, To better understand the difference of winter pre- but the areal-mean difference across a mountain range is cipitation between SNOTEL and WRF, we map out the small. This has been shown before over the Colorado mean bias and correlation coefficient between WRF and mountains (Ikeda et al. 2010) and is confirmed by a SNOTEL for the winter months (Figs. 5a,b). The back- comparison of winter precipitation over Greater Yel- ground color field in Fig. 5a represents the WRF-resolved lowstone between simulations with 4- and 1.33-km res- terrain slope in the direction of the prevailing wind. This olutions (Fig. 6). The 1.33-km resolution simulation is is a weighted-average surface wind, the weighting being driven by hourly 4-km resolution WRF outputs and is precipitation rate, using hourly data over 10 winters. The run from December 2007 to February 2008. At higher slope for each grid point is calculated as an interpolated resolution (Fig. 6b), WRF resolves a finer texture of height difference over an along-wind distance corre- terrain-relative precipitation distribution than the 4-km sponding to twice the grid spacing. A positive (negative) simulation (Fig. 6a) and is slightly more consistent with slope is referred to as the windward (lee) side. For 96% of SNOTEL in terms of correlation coefficient and RMSB the SNOTEL sites, the mean monthly winter precipitation (Table 2). The 1.33-km simulation estimates 0.77 mm 2 bias is within 40 mm month 1, and the correlation (1.3%) more winter precipitation than the 4-km simu- coefficient is greater than 0.8, but there are some outliers. lation. The correlation coefficient, RMSB, and root- Many negative biases are observed on the windward side mean-square percent bias (RMSPB) between the two while positive biases are often observed on the lee side. runs are 0.99, 14.4 mm, and 8.7%, respectively. Over The negative biases on the windward side may be some mountains (e.g., Gros Ventre, Salt River Range), because a 4-km WRF resolution is not fine enough to re- the difference locally is as large as 50 mm, which is still solve some steep windward slopes, for example, in the small compared to the difference between simulation

Unauthenticated | Downloaded 10/04/21 04:28 AM UTC SEPTEMBER 2017 J I N G E T A L . 2549

FIG. 4. Scatterplots of seasonal precipitation estimated by WRF against SNOTEL at all SNOTEL sites in MAM, JJA, SON, and DJF. The black dotted lines are the 1:1 lines. The correlation coefficient, RMSB, and mean bias are shown in each panel. and gauge-driven gridded datasets and the uncertainties areas with low gauge density). The sensitivity study by Liu of gauge-driven gridded datasets (shown later). et al. (2011) shows that the microphysics scheme has the The winter precipitation difference between IWUS most significant impact on the simulations. The micro- WRF and PRISM (IWUS WRF 2 PRISM) is shown in physics scheme used in this study, the Thompson scheme, Fig. 7a. Especially within the Greater Yellowstone area, captures the characteristics of winter orographic pre- WRF estimates more winter precipitation over some cipitation in IWUS best (Liu et al. 2011). Other physical mountains (Wind River Range, Gros Ventre, Big Horn, schemes (e.g., land surface, PBL, radiation) have minor and Absaroka) than PRISM, while for others, such as the impacts on the model results (Liu et al. 2011). However, it Teton Range, WRF estimates less precipitation. In the is not known whether different driven datasets (boundary Utah and Colorado subdomains, the agreement generally conditions) impact the results. Therefore, we compare the is better. WRF performance differences between nearby CONUS WRF against PRISM (Fig. 7b). CONUS WRF mountain ranges may be due to three types of un- uses a different driver (ERA-Interim reanalysis data), a certainties: 1) model related (uncertainties related to very different domain (exceeding the continental United WRF physics and boundary conditions), 2) measurement States), and a different series of years (October 2000– related (often related to local vegetation and finescale September 2013) (Liu et al. 2017). Yet, the comparison terrain factors around the SNOTEL sites), and 3) PRISM with PRISM reveals a remarkably similar pattern of winter algorithm related (uncertainties due to extrapolation in snowfall biases, suggesting that uncertainties of the second

Unauthenticated | Downloaded 10/04/21 04:28 AM UTC 2550 JOURNAL OF HYDROMETEOROLOGY VOLUME 18

FIG. 5. (a) Mean monthly precipitation difference and (b) correlation coefficients in DJF between WRF and SNOTEL. The background in (a) is the terrain slope in the direction of prevailing wind during precipitation events. (c) Monthly winter precipitation difference (WRF 2 SNOTEL) and (d) correlation coefficients between WRF and SNOTEL as a function of elevation. Gray dots in (c) represent the monthly precipitation difference (WRF 2 SNOTEL) at all SNOTEL sites in DJF from March 2002 to February 2012, and gray dots in (d) represent the correlation coefficients of monthly precipitation in winter from March 2002 to February 2012 between WRF and SNOTEL. The black dots are the mean values at different elevation intervals, and the black bars indicate 61 standard deviation. (e),(f) As in (c), but for the WRF 2 SNOTEL difference as functions of elevation bias and prevailing wind speed. Brown and green dots in (f) represent the upwind and lee SNOTEL gauges, respectively. and third type (i.e., uncertainties in the gauge-driven data. In this section, we examine the differences be- gridded dataset) matter more, which points to errors in tween different gauge-driven datasets and try to better the measurements or the gridding of these measurements. understand their uncertainties. The maximum absolute bias between any two of c. Intercomparison of gauge-driven gridded PRISM, Daymet, NLDAS II, and the 100 CUSEG precipitation datasets datasets (103 datasets in total, giving a total of 5253 Analyses in section 3a suggest in gauge-deprived pairs) at any grid point in the study domain is shown in areas, such as the Gros Ventre Range, the Elkhead Fig. 8.2 Bilinear interpolation is used to project different Range, the eastern San Juan Mountains (all labeled in Fig. 1b), the Yellowstone National Park area (northwest corner of Wyoming), and the Great Salt Lake, the ac- 2 The same analysis using PRISM, Daymet, NLDAS II, and just curacy of the precipitation estimated by different gauge- the mean CUSEG datasets (four in total, or six pairs) produces driven gridded datasets is questionable given the lack of similar information.

Unauthenticated | Downloaded 10/04/21 04:28 AM UTC SEPTEMBER 2017 J I N G E T A L . 2551

FIG. 6. Maps of wintertime precipitation in Greater Yellowstone from December 2007 to February 2008 esti- mated by the IWUS WRF simulation at (a) 4- and (b) 1.33-km resolution, as well as (c) their difference and (d) percent difference. datasets onto a common 4-km grid. The difference be- Front Range in Colorado), the difference between the tween different datasets generally is larger over the datasets is relatively small. The differences may also be mountains than the plains in all seasons, not in a percent related to the different resolutions of gauge-driven sense (not shown), but certainly in an absolute sense gridded datasets, as shown by the relatively large un- (Fig. 8). This likely is due to different statistical relations certainty over the Wasatch Range, which is narrow and between precipitation and terrain used for the different steep (Fig. 8d). The differences between gridded data- datasets. In areas with lots of SNOTEL sites (e.g., the sets are relatively small in summer (0–150 mm) and

Unauthenticated | Downloaded 10/04/21 04:28 AM UTC 2552 JOURNAL OF HYDROMETEOROLOGY VOLUME 18

TABLE 2. Correlation coefficient, mean bias, and RMSB between WRF simulations and SNOTEL in Greater Yellowstone.

Correlation coefficient Mean bias (mm) RMSB (mm)

WRF4km vs SNOTEL 0.73 24.93 88.55 WRF1.33km vs SNOTEL 0.81 26.33 75.65

relatively large in winter and spring (0–250 mm; DJF and since they are anchored to these measurements, but their MAM), which is the wettest period over the mountains. agreement deteriorates in areas where gauge sites are The relative performances of the four gauge-driven sparse. The accuracy of the gridded datasets at higher el- gridded datasets are assessed in Fig. 9, which compares evations is less certain, not just due to relative gauge gauge-driven gridded datasets (PRISM, Daymet, NLDAS sparsity in the mountains, but also due to SNOTEL data II, or CUSEG) against their mean. The upper and lower uncertainty related to the higher uncertainty of the mea- panels show the summer and winter precipitation difference surement of snow versus rain, the higher fraction of snow maps, respectively. All four datasets have a correlation co- versus rain at higher elevation, the higher uncertainty of efficient higher than 0.9 with the mean, because the devel- the snowfall measurement under stronger wind, and the opment of the datasets is strongly terrain related. PRISM is typically higher wind speed at higher elevations. the most consistent with the mean for summer precipitation, d. Comparison of winter precipitation between and Daymet is the most consistent with the mean for winter gridded datasets and WRF precipitation. NLDAS II estimates the most summer pre- cipitation and the least winter precipitation, while CUSEG Given these possible data-related uncertainties, we estimates the least summer precipitation and the most now assume WRF to be the reference to evaluate grid- winter precipitation. All the four gauge-driven gridded da- ded observational datasets. Maps of the mean winter- tasets have larger differences against the mean for winter time precipitation absolute and relative difference precipitation than summer precipitation. between gauge-driven gridded datasets and WRF are In short, gauge-driven gridded datasets using statistical shown in the upper and lower panels of Fig. 10, re- interpolation techniques perform best at SNOTEL sites, spectively. The correlation coefficient r, RMSB, and

FIG. 7. Difference of winter precipitation between (a) IWUS WRF and PRISM and (b) CONUS WRF and PRISM.

Unauthenticated | Downloaded 10/04/21 04:28 AM UTC SEPTEMBER 2017 J I N G E T A L . 2553

FIG. 8. Seasonal precipitation estimation uncertainty evident from gauge-driven gridded datasets in MAM, JJA, SON, and DJF. The uncertainty is defined as the maximum absolute difference between any two of PRISM, Daymet, NLDAS II, and the 100 CUSEG datasets (103 in total) for each grid box. The blue (black) dots indicate the locations of SNOTEL (other) gauges.

RMSPB of precipitation at locations 2 km MSL are Of all gauge-driven gridded datasets, PRISM and Daymet shown. Large differences are evident in some moun- compare better against WRF in the study domain than tains, especially where SNOTEL sites are scarce (e.g., in NLDAS II and CUSEG. PRISM and Daymet signifi- the Wind River Range, above the tree line; Figs. 10a–d). cantly underestimate the precipitation over the Gros

Unauthenticated | Downloaded 10/04/21 04:28 AM UTC 2554 JOURNAL OF HYDROMETEOROLOGY VOLUME 18

FIG. 9. Difference of (a)–(d) summer and (e)–(h) winter precipitation between a gauge-driven dataset (PRISM, Daymet, NLDAS II, or CUSEG) and the mean of the four datasets. The correlation coefficient, mean bias, and RMSB (mm) of precipitation for grid points 2 km MSL are shown.

Ventre and Wind River Ranges, NLDAS II has the as in some high-elevation places, indicating uncertainties largest RMSB and has relatively larger uncertainties in the statistical interpolation techniques. Here the gauge over Greater Yellowstone than Utah and Colorado, and density is defined as the number of gauges (including all CUSEG overestimates the precipitation over many of gauges plotted in Fig. 8, not just SNOTEL gauges) the mountains but underestimates the precipitation over within a range of 30 km for each gauge site. some mountains (e.g., Gros Ventre and Wind River Since the algorithms behind the gauge-driven gridded Ranges). Since the four datasets are all SNOTEL based, datasets lack detailed physical processes, we examine the the pattern of the bias maps is generally consistent with precipitation bias at 3 km MSL between the gridded da- the difference between WRF and SNOTEL (Fig. 5a). tasets against WRF as a function of wind speed (Fig. 11c). To better understand the difference between gauge- As defined in section 3b, this is the WRF-derived prevail- driven gridded datasets and WRF, we analyze the stan- ing surface wind during precipitation events. The gauge- dard deviations of wintertime precipitation bias (again driven gridded datasets slightly overestimate precipitation 2 assuming WRF to be the truth), estimated by all gauge- when the surface wind is weaker than 16 m s 1 and un- 2 based gridded datasets, as shown in Fig. 11a. The circles derestimate precipitation in winds exceeding 24 m s 1.So represent the bias standard deviation for different ele- in mountain areas where strong winds are common (e.g., vation intervals, and the blue histogram represents the Wind River Range), all gauge-driven gridded datasets may amount of grids in the study domain. As seen from the underestimate the wintertime precipitation. figure, the bias standard deviations increase with terrain height (Fig. 11a), suggesting that these observational 4. Discussion datasets have larger uncertainties at higher elevations, consistent with Fig. 8. Indeed, the correlation coefficient Several high-resolution gridded historical precipitation between any of the gauge-driven gridded datasets and datasets have been developed, using different interpola- WRF decreases with decreasing gauge density (Fig. 11b), tion techniques, and they arewidelyusedforarangeof

Unauthenticated | Downloaded 10/04/21 04:28 AM UTC SEPTEMBER 2017 J I N G E T A L . 2555

FIG. 10. DJF precipitation (a)–(d) bias and (e)–(h) percent bias between gridded observational datasets and WRF. Here WRF is considered to be the truth. The correlation coefficient and RMSB (mm) of precipitation for grid points 2 km MSL are shown in (a)–(d) and RMSPB (%) of precipitation at 2 km MSL is shown in (e)–(h). purposes, for example, hydrometeorology and ecology gridded observations, areas where either the SNOTEL analyses, forecasting and agricultural operations, and hy- measurements are of questionable quality or where gauges drology and land surface modeling (e.g., Bales et al. 2006; are lacking. The 4-km convection-permitting IWUS WRF Lundquist et al. 2015; Henn et al. 2017). Our analysis shows regional climate simulation nicely captures cold-season that gauge-driven gridded datasets have relatively large orographic precipitation in the IWUS, consistent with uncertainties over mountains inadequately covered with previous simulations (Ikeda et al. 2010; Rasmussen et al. gauge sites, especially in winter. Some previous studies 2011). Comparison between WRF and the four gauge- using different methods have also suggested that gauge- driven datasets suggest PRISM and Daymet are more driven gridded datasets may have large uncertainties in consistent with WRF for winter precipitation than NLDAS wintertime precipitation estimation over mountains in the II and CUSEG. There are remarkable similarities in the IWUS (e.g., Daly et al. 2008; Gutmann et al. 2012; Henn cold-season precipitation biases over specific mountain et al. 2017), but it is not known over which particular ranges (comparing against PRISM) between the IWUS mountains the largest uncertainties are found and which WRFandtheCONUSWRF(Liu et al. 2017) simulations dataset performs the best. In this study we argue that high- (Fig. 7), notwithstanding the fact that these simulations resolution model simulations have become a useful tech- use different domains, different time periods, and different nique to evaluate and improve gauge-driven gridded model physics. This indicates that these biases are at least precipitation estimates. Such simulations are useful also to partly due to uncertainties in the measurement of snowfall identify areas with large discrepancies between model and and/or to uncertainties in the statistical gridding techniques

Unauthenticated | Downloaded 10/04/21 04:28 AM UTC 2556 JOURNAL OF HYDROMETEOROLOGY VOLUME 18

FIG. 11. (a) Standard deviation of winter precipitation bias between gridded datasets and WRF as a function of elevation. (b) Correlation coefficient between observational datasets and WRF as a function of gauge density, which is defined as the number of gauges in a range of 30 km for each gauge. (c) Winter precipitation difference between observational datasets and WRF as a function of wind speed, for grid points 3 km MSL.

over complex terrain. In the IWUS domain, we find par- and isolated. A previous study shows the summer pre- ticularly large discrepancies in the Wind River, the Gros cipitation over Colorado mountains is significantly over- Ventre, and Bighorn Ranges in Wyoming, where most estimated in simulations with relatively coarse resolution gauges are in the foothills, not in the high country. This may (12 km); the results are significantly improved after in- motivate the reevaluation of some gauge records over these creasing the resolution to 4 km (Rasmussen et al. 2014). mountains, and the installation of additional SNOTEL The model representativeness of the summertime pre- gauges, especially in regions marked by large discrepancies cipitation is affected also by its cloud microphysics scheme, between modeled and gauge-driven precipitation esti- which controls the precipitation efficiency and cloud life mates. In Colorado and Utah, more gauges exist closer to cycle (e.g., Khain et al. 2015). Further improvements in the mountain tops, and the discrepancies between WRF and resolution of convection-permitting models and in micro- observed winter precipitation generally are smaller. The physics schemes may improve the simulation of summer- differences between WRF and the four gauge-driven time precipitation and its diurnal cycle in the IWUS. gridded datasets are subtly related to the wind speed, sug- gesting measurements of ambient conditions may be 5. Conclusions helpful to improve the gauge-driven gridded datasets. The 4-km convection-permitting WRF simulation A 10-yr, 4-km-resolution, convection-permitting WRF overestimates summertime precipitation in the high plains simulation is used to study seasonal precipitation in the of Wyoming and Colorado. Changing land surface, PBL, IWUS, in particular over the mountains, and to compare and radiation parameterizations have no large impacts on orographic precipitation with SNOTEL gauge data, four the results (Rasmussen et al. 2014; Liu et al. 2017). gauge-driven gridded datasets (PRISM, Daymet, NLDAS Increasing the model resolution may improve the results II, and CUSEG), and a radar-based dataset (NCEP IV). because orographic convection generally is rather small The main conclusions are as follows:

Unauthenticated | Downloaded 10/04/21 04:28 AM UTC SEPTEMBER 2017 J I N G E T A L . 2557 d The IWUS WRF simulation captures wintertime Clark, M. P., and A. G. Slater, 2006: Probabilistic quantitative orographic precipitation well, and biases over specific precipitation estimation in complex terrain. J. Hydrometeor., mountain ranges are identical to those in another 7, 3–22, doi:10.1175/JHM474.1. Daly, C., R. P. Neilson, and D. L. Phillips, 1994: A statistical– WRF simulation (Liu et al. 2017), suggesting that topographic model for mapping climatological precipitation these biases are at least partly due to uncertainties in over mountainous terrain. J. Appl. Meteor., 33, 140–158, the measurement of snowfall and/or uncertainties in doi:10.1175/1520-0450(1994)033,0140:ASTMFM.2.0.CO;2. the gridding of these measurements over complex ——, W. P. Gibson, G. H. Taylor, G. L. Johnson, and P. Pasteris, terrain. 2002: A knowledge-based approach to the statistical mapping of climate. Climate Res., 22, 99–113, doi:10.3354/cr022099. d Cross validation of the WRF simulation against the ——, ——, ——, M. K. Doggett, and J. I. Smith, 2007: Observer gauge-driven gridded dataset suggests that uncer- bias in daily precipitation measurements at United States co- tainties of the gridded datasets are larger over moun- operative network stations. Bull. Amer. Meteor. Soc., 88, 899– tains in winter. Among the four gauge-driven gridded 912, doi:10.1175/BAMS-88-6-899. datasets, NLDAS II estimates the least winter pre- ——, M. Halbleib, J. I. Smith, W. P. Gibson, M. K. Doggett, G. H. Taylor, J. Curtis, and P. P. Pasteris, 2008: Physio- cipitation and the most summer precipitation in the graphically sensitive mapping of climatological temperature IWUS, while CUSEG estimates the most winter pre- and precipitation across the conterminous United States. Int. cipitation and the least summer precipitation. PRISM J. Climatol., 28, 2031–2064, doi:10.1002/joc.1688. and Daymet are more consistent with WRF than Doesken, N. J., and G. L. Schaefer, 1987: The contribution of NLDAS II and CUSEG. NCEP IV compares well SNOTEL precipitation measurements to climate analysis, mon- against other datasets in the vicinity of radars, but itoring and research in Colorado. Proc. 55th Annual Western Snow Conf., Vancouver, BC, Canada, Western Snow Confer- grossly underestimates winter precipitation over the ence, 20–30, https://westernsnowconference.org/node/675. mountains. Ebert, E. E., J. E. Janowiak, and C. Kidd, 2007: Comparison of d The discrepancy between WRF and gauge-driven gridded near-real-time precipitation estimates from satellite observa- datasets is larger at higher elevations and larger in areas tions and numerical models. Bull. Amer. Meteor. Soc., 88, 47– with fewer gauges. Compared to WRF, gauge-driven 64, doi:10.1175/BAMS-88-1-47. Fulton, R. A., J. P. Breidenbach, D. J. Seo, D. A. Miller, and datasets tend to overestimate the precipitation in areas T. O’Bannon, 1998: The WSR-88D rainfall algorithm. Wea. with prevailing weak surface winds and tend to underes- Forecasting, 13, 377–395, doi:10.1175/1520-0434(1998)013,0377: timate the precipitation in areas with strong winds. TWRA.2.0.CO;2. d In general, regional climate simulations can be used to Gourley, J. J., Y. Hong, Z. L. Flamig, L. Li, and J. Wang, 2010: improve the algorithms of gauge-driven gridded data- Intercomparison of rainfall estimates from radar, satellite, gauge, and combinations for a season of record sets in complex terrain, to inform where gauges are too rainfall. J. Appl. Meteor. Climatol., 49, 437–452, sparsely distributed, and to indicate which SNOTEL doi:10.1175/2009JAMC2302.1. sites may be questionable. Gutmann, E. D., R. M. Rasmussen, C. Liu, K. Ikeda, D. J. Gochis, M. P. Clark, J. Dudhia, and G. Thompson, 2012: A comparison In a separate study in progress, we are using the IWUS of statistical and dynamical downscaling of winter pre- WRF simulation and gridded datasets to evaluate sum- cipitation over complex terrain. J. Climate, 25, 262–281, mertime precipitation over the IWUS, in particular its doi:10.1175/2011JCLI4109.1. diurnal cycle. Henn, B., A. J. Newman, B. Livneh, C. Daly, and J. D. Lundquist, 2017: An assessment of differences in gridded precipita- tion datasets in complex terrain. J. Hydrol., doi:10.1016/ Acknowledgments. This work was funded by the j.jhydrol.2017.03.008, in press. Wyoming Water Development Commission and the Hong, S. Y., and H. L. Pan, 1996: Nonlocal boundary layer United States Geological Survey, under the auspices of vertical diffusion in a medium-range forecast model. Mon. Wea. the University of Wyoming Water Research Program. Rev., 124, 2322–2339, doi:10.1175/1520-0493(1996)124,2322: . The lead author was funded by a Wyoming Engineering NBLVDI 2.0.CO;2. Hou, A. Y., and Coauthors, 2014: The Global Precipitation Mea- Initiative Doctoral Fellowship. The research benefitted surement Mission. Bull. Amer. Meteor. Soc., 95, 701–722, from comments by Roy Rasmussen. doi:10.1175/BAMS-D-13-00164.1. Iacono, M. J., J. S. Delamere, E. J. Mlawer, M. W. Shephard, REFERENCES S. A. Clough, and W. D. Collins, 2008: Radiative forcing by long-lived greenhouse gases: Calculations with the AER ra- Bales, R. C., N. P. Molotch, T. H. Painter, M. D. Dettinger, R. Rice, diative transfer models. J. Geophys. Res., 113, D13103, and J. Dozier, 2006: Mountain hydrology of the western doi:10.1029/2008JD009944. United States. Water Resour. Res., 42, W08432, doi:10.1029/ Ikeda, K., and Coauthors, 2010: Simulation of seasonal snowfall 2005WR004387. over Colorado. Atmos. Res., 97, 462–477, doi:10.1016/ Barnett, T. P., and Coauthors, 2008: Human-induced changes in the j.atmosres.2010.04.010. hydrology of the western United States. Science, 319, 1080– Jimenez, P. A., J. J. Dudhia, F. Gonzalez-Rouco, J. Navarro, 1083, doi:10.1126/science.1152538. J. P. Montavez, and E. Garcia-Bustamante, 2012: A revised

Unauthenticated | Downloaded 10/04/21 04:28 AM UTC 2558 JOURNAL OF HYDROMETEOROLOGY VOLUME 18

scheme for the WRF surface layer formulation. Mon. Wea. PRISM Climate Group, 2016: Precipitation-elevation Regressions Rev., 140, 898–918, doi:10.1175/MWR-D-11-00056.1. on Independent Slopes Model. Northwest Alliance for Com- Khain, A. P., and Coauthors, 2015: Representation of microphys- putational Science and Engineering, Oregon State University, ical processes in cloud-resolving models: Spectral (bin) accessed 16 February 2017, http://www.prism.oregonstate.edu/. microphysics versus bulk parameterization. Rev. Geophys., 53, Rasmussen, R., and Coauthors, 2011: High-resolution coupled 247–322, doi:10.1002/2014RG000468. climate runoff simulations of seasonal snowfall over Colorado: Lin, X., and A. Y. Hou, 2012: Estimation of rain intensity spectra A process study of current and warmer climate. J. Climate, 24, over the continental United States using ground radar–gauge 3015–3048, doi:10.1175/2010JCLI3985.1. measurements. J. Climate, 25, 1901–1915, doi:10.1175/ ——, and Coauthors, 2012: How well are we measuring snow: The JCLI-D-11-00151.1. NOAA/FAA/NCAR winter precipitation test bed. Bull. Amer. Lin, Y., and K. E. Mitchell, 2005: The NCEP stage II/IV hourly Meteor. Soc., 93,811–829,doi:10.1175/BAMS-D-11-00052.1. precipitation analyses: Development and applications. 19th ——, and Coauthors, 2014: Climate change impacts on the water Conf. on Hydrology, San Diego, CA, Amer. Meteor. Soc., 1.2, balance of the Colorado headwaters: High-resolution regional https://ams.confex.com/ams/pdfpapers/83847.pdf. climate model simulations. J. Hydrometeor., 15, 1091–1116, Liu, C., K. Ikeda, G. Thompson, R. Rasmussen, and J. Dudhia, doi:10.1175/JHM-D-13-0118.1. 2011: High-resolution simulations of wintertime precipitation Saha, S., and Coauthors, 2010: The NCEP Climate Forecast System in the Colorado Headwaters region: Sensitivity to physics Reanalysis. Bull. Amer. Meteor. Soc., 91, 1015–1057, parameterizations. Mon. Wea. Rev., 139, 3533–3553, doi:10.1175/2010BAMS3001.1. doi:10.1175/MWR-D-11-00009.1. Seo, D. J., 1998: Real-time estimation of rainfall fields using radar ——, and Coauthors, 2017: Continental-scale convection-permitting rainfall and rain gage data. J. Hydrol., 208, 37–52, doi:10.1016/ modeling of the current and future climate of North America. S0022-1694(98)00141-3. Climate Dyn., 49, 71–95, doi:10.1007/s00382-016-3327-9. Serreze, M. C., M. P. Clark, R. L. Armstrong, D. A. McGinnis, and Lundquist, J. D., M. Hughes, B. Henn, E. D. Gutmann, B. Livneh, R. S. Pulwarty, 1999: Characteristics of the western United J. Dozier, and P. Neiman, 2015: High-elevation precipitation States snowpack from Snowpack Telemetry (SNOTEL) data. patterns: Using snow measurements to assess daily gridded Water Resour. Res., 35, 2145–2160, doi:10.1029/1999WR900090. J. Hydrometeor. datasets across the Sierra Nevada, California. , Smalley, M., T. L’Ecuyer, M. Lebsock, and J. Haynes, 2014: A 16, 1773–1792, doi:10.1175/JHM-D-15-0019.1. comparison of precipitation occurrence from the NCEP Stage Maddox, R. A., J. Zhang, J. J. Gourley, and K. W. Howard, IV QPE product and the CloudSat Cloud Profiling Radar. 2002: Weather radar coverage over the contiguous J. Hydrometeor., 15, 444–458, doi:10.1175/JHM-D-13-048.1. United States. Wea. Forecasting, 17, 927–934, doi:10.1175/ Tesfagiorgis, K., S. E. Mahani, N. Y. Krakauer, and 1520-0434(2002)017,0927:WRCOTC.2.0.CO;2. R. Khanbilvardi, 2011: Bias correction of satellite rainfall Mote, P. W., A. F. Hamlet, M. P. Clark, and D. P. Lettenmaier, estimates using a radar-gauge product – A case study in 2005: Declining mountain snowpack in western North Amer- Oklahoma (USA). Hydrol. Earth Syst. Sci., 15, 2631–2647, ica. Bull. Amer. Meteor. Soc., 86, 39–49, doi:10.1175/ doi:10.5194/hess-15-2631-2011. BAMS-86-1-39. Thompson, G., R. M. Rasmussen, and K. Manning, 2004: Explicit NCEP, 2001: National Centers for Environmental Prediction forecasts of winter precipitation using an improved bulk Hourly Multisensor Precipitation Analysis Stage IV. Subset: March 2003–February 2012, NOAA NCEP Environmental microphysics scheme. Part I: Description and sensitivity Modeling Center, accessed 5 August 2016, http://www.emc. analysis. Mon. Wea. Rev., 132, 519–542, doi:10.1175/ , . ncep.noaa.gov/mmb/ylin/pcpanl/stage4/. 1520-0493(2004)132 0519:EFOWPU 2.0.CO;2. ——, 2014: North American Land Data Assimilation System Thornton, P. E., S. W. Running, and M. A. White, 1997: Generating project phase 2. NOAA NCEP Environmental Modeling surfaces of daily meteorological variables over large regions of Center, accessed 11 September 2016, http://www.emc.ncep. complex terrain. J. Hydrol., 190, 214–251, doi:10.1016/ noaa.gov/mmb/nldas/. S0022-1694(96)03128-9. Newman, A. J., and Coauthors, 2015a: Gridded ensemble pre- ——, M. M. Thornton, B. W. Mayer, N. Wilhelmi, Y. Wei, cipitation and temperature estimates for the contiguous R. Devarakonda, and R. B. Cook, 2014: Daymet: Daily Sur- United States. J. Hydrometeor., 16, 2481–2500, doi:10.1175/ face Weather Data on a 1-km Grid for North America, Ver- JHM-D-15-0026.1. sion 2. ORNL DAAC, accessed 11 September 2016, https:// ——, and Coauthors, 2015b: Gridded ensemble precipitation doi.org/10.3334/ORNLDAAC/1328. and temperature estimates over the contiguous United Woodhouse, C. A., 2004: A paleo perspective on hydroclimatic States. Earth System Grid at NCAR, accessed 7 variability in the western United States. Aquat. Sci., 66, November 2016, https://www.earthsystemgrid.org/dataset/ 346–356, doi:10.1007/s00027-004-0723-8. gridded_precip_and_temp.html. Xia, Y., and Coauthors, 2012: Continental-scale water and energy Niu, G.-Y., and Coauthors, 2011: The community Noah land sur- flux analysis and validation for the North American Land face model with multiparameterization options (Noah-MP): Data Assimilation System project phase 2 (NLDAS-2): 1. Model description and evaluation with local-scale mea- 1. Intercomparison and application of model products. surements. J. Geophys. Res., 116, D12109, doi:10.1029/ J. Geophys. Res., 117, D03109, doi:10.1029/2011JD016048. 2010JD015139. Yang, Z.-L., and Coauthors, 2011: The community Noah land sur- NRCS, 1977: Snow Telemetry gauge network. Subset: March 2003– face model with multiparameterization options (Noah-MP): February 2012, NRCS, Department of Agriculture, accessed 2. Evaluation over global river basins. J. Geophys. Res., 116, 31 March 2016, http://www.wcc.nrcs.usda.gov/. D12110, doi:10.1029/2010JD015140.

Unauthenticated | Downloaded 10/04/21 04:28 AM UTC