Performance of Emerging Technologies for Measuring Solid and Liquid Precipitation in Cold Climate as Compared to the Traditional Manual Gauges
1. Introduction
Precipitation plays a critical role on our planet by modulating the hydrological cycle and by influencing daily human activities, including air and ground transportation. Validation of climate and numerical weather prediction models, and radar and satellite remote sensing algorithms, require accurate precipitation measurements. Precipitation amount is normally measured using a weighing gauge, which is an open container on the ground that collects precipitating hydrometeors, including raindrops, snow and hail particles, etc. However, accurately measuring the precipitation is usually more complex, particularly for snow because of many factors, including losses from wind, wetting, and evaporation (Sevruk and Klemm 1989; Goodison et al. 1998; Rasmussen et al. 2012; Yang et al. 2005) and potential enhancement due to blowing snow (Yang et al. 1999). Recent studies also indicate that snow gauge catch efficiency depends on snow type and density (Thériault et al. 2015, 2012; Colli et al. 2015, 2016a,b). This is particularly challenging in the cold northern latitudes, where the snowfall intensities are relatively low. The manual gauges that are deployed in the Canadian precipitation networks include the Type B and Canadian Nipher (Mekis and Vincent 2011; Metcalfe and Goodison 1993). These gauges are normally referred to as standard rain gauges. Automatic gauges, such as Geonor and Pluvio, are currently being considered for the networks.
One way to reduce the wind-induced loss is by using some kind of wind shield, by placing a given gauge inside some bushes, or using specially designed shields, including the double-fenced structure with a Tretyakov manual gauge suggested by the World Meteorological Organization (WMO), which is normally referred to as Double Fence Intercomparison Reference (DFIR) (Goodison et al. 1998). There is no absolute reference standard for snow measurement. The DFIR is usually considered the secondary standard with the bush-surrounded gauge being the primary standard (Goodison et al. 1998; Yang 2014). Part of the reason why the bush gauge is considered to be the primary standard is because it showed higher catch efficiency as compared to the gauge placed inside the DFIR (Goodison et al. 1998). A more recent study by Yang (2014) showed that the DFIR underestimated solid precipitation by 5% as compared to the bush gauge. During the WMO Solid Precipitation Measurement Intercomparison (SPMI), the Canadian Nipher gauge outperformed many other partially shielded or unshielded gauges when compared against a Russian Tretyakov manual gauge as the DFIR (Goodison et al. 1998). One of the outcomes of the WMO SPMI was the development of correction factors for unshielded and partially shielded gauges that are normally referred to as transfer functions (Goodison et al. 1998). These transfer functions were developed based on daily manual snow measurements, and so there are some uncertainties with their use at shorter time scales due to the strong variability normally seen during precipitation and associated weather conditions. These automatic gauges are currently being used with various wind shield configurations, including the double-fenced structure suggested by the WMO, but their accuracy under various atmospheric conditions are not well known (Rasmussen et al. 2012; Theriault et al. 2015), particularly for the light solid precipitation that normally occurs in cold climates. There are also noncatchment-type optical instruments that employ a forward light scattering method, and hot plates and distrometers that measure hydrometer size and fall velocity distributions to estimate the precipitation intensities (Rasmussen et al. 2011; Boudala et al. 2014a,b; Brandes et al. 2007). These emerging technologies usually are more sensitive and measure precipitation intensity at higher temporal resolution down to minutes. Some of these instruments do not suffer from the wind-induced and other losses mentioned earlier and hence are suitable for measuring particularly light precipitation. However, these instruments are relatively new and their response under various atmospheric conditions is not well known.
This study aims to address some of the issues associated with both catchment- and noncatchment-type gauges, including testing their performance under a cold climate, providing suitable guidance to improve the accuracy of the catchment-type gauges, and characterize snow density under various weather conditions. For this purpose, precipitation and type, and snow depth data were collected and analyzed as part of the 4Wing Cold Lake Research Project (4WCLRP). Various instruments were used, including the Vaisala present weather sensors PWD22 and FS11P, the OTT Pluvio2 gauge, the Canadian Nipher and Type B manual gauges, and a snow ruler to measure snow depth. The 4WCLRP was initiated by the Cloud Physics and Severe Weather Research Section of Environment and Climate Change Canada (ECCC) with the cooperation of the Department of National Defence (DND). The data used in this study were collected during the period between September 2014 and August 2015. The paper is organized as follows: The observation site and instrumentation are discussed in section 2, the precipitation amount and type data analysis and results are given in section 3, the snow density is discussed in section 4, and the summary and conclusions are given in section 5.
2. Observation sites and instrumentation
The study area, the Cold Lake Regional Airport (CYOD), is located in northeastern Alberta, Canada (at 541 m MSL; 54°23′59.8″N, 110°17′6.2″W). The geographical locations of CYOD (Fig. 1a) and the surrounding areas, and inside COYD where the instruments were located (Figs. 1a,b) are shown. In Fig. 1b, in addition to the ECCC observation site used in this study, there is a collocated permanent meteorological observation site used by the DND. The region is generally characterized by a humid continental climate with warm summers and cold winters as indicated in Fig. 2, which shows monthly averaged temperature (Fig. 2a), relative humidity (Fig. 2b), and the frequency distributions for temperature and humidity (Figs. 2c,d, respectively) based on hourly observations taken during the 2014–15 period. The monthly mean temperature varies from −12°C in January to 18°C in July. The minimum temperature in winter reached near −35°C, the warmest temperature is close to 27°C in summer, and the most frequent temperature is close to 0°C (Fig. 2c). The monthly averaged humidity varied from 85% in winter (December) to near 55% in summer (May). The humidity may reach as low as 15% in April and May, but within a given month the humidity may reach 100%, and the most frequent humidity is 90% (Fig. 2d). As indicated in Fig. 1a, there are a number of local conditions that can potentially affect the local weather and produce precipitation. To the west and northwest of the airport, there are four small lakes (Marie, Ethel, Crane, and Hilda) and to the northeast there are two large lakes (Cold and Primrose) known to produce precipitation and other weather phenomena when the flow is from a northeasterly direction. The west and southwest sides of the airport are surrounded by the Beaver River valley with an east–west orientation, which is also known to produce various weather conditions at the airport.
Google map showing (a) Cold Lake and surrounding areas and (b) two instrument sites. In (b) the site marked as ECC is where the ECCC instruments were located, and the site marked as DND is where the DND permanent observing site is located.
Citation: Journal of Atmospheric and Oceanic Technology 34, 1; 10.1175/JTECH-D-16-0088.1
Climatology of Cold Lake based on 2 years of hourly reported data.
Citation: Journal of Atmospheric and Oceanic Technology 34, 1; 10.1175/JTECH-D-16-0088.1
The list of the deployed instruments along with the measured microphysical and meteorological parameters, measurement principles, and associated uncertainties are given in Table 1. The instrument setup at the ECCC site is given in the top panel of Fig. 3, and the instruments installed at the DND site are given in Figs. 3a,b. The DND observation site is located approximately 948 m from the ECCC site. The descriptions and measurement principles of the instruments installed at the ECCC and DND sites are given below.
The description of the instruments installed at CYOD and related information.
(top) The ECCC instrument setup at CYOD. (a) The Type B and (b) the Canadian Nipher manual gauges at the DND site.
Citation: Journal of Atmospheric and Oceanic Technology 34, 1; 10.1175/JTECH-D-16-0088.1
The Vaisala PWD22 and FS11P present weather sensors measure precipitation, precipitation type, and visibility. The operating principles of these two Vaisala sensors are similar with some minor differences (Table 1). These probes have two arms facing each other, one equipped with a near-infrared transmitter and the other with a receiver. The two arms are arranged in a way that the infrared light can reach the receiver only if it is forward scattered at a given angle (45° for PWD22 and 42° for FS11P) by particles between the arms. Signal processing software analyzes the voltage output from the receiver, along with the current temperature, to determine the type and intensity of precipitation. The reported precipitation types include the WMO synoptic (SYNOP) and METAR codes, as well as the National Weather Service (NWS) code. Each instrument also has a heated capacitive surface that provides a liquid water equivalent measurement. The FS11P is also equipped with a background light detection sensor.
The OTT Pluvio2 precipitation gauge used here has a precipitation-collecting container with a collecting area of 200 cm2 and a capacity of up to 1500 mm. The weight of the precipitation collected in the container is measured every 6 s by an electronic weighing cell at a resolution of 0.01 mm. The measured precipitation intensity and amount are reported every minute. The OTT Pluvio2 uses a special filter algorithm to correct the 6-s precipitation weights for wind effects. Additionally, the gauge is fitted with a single Alter shield, as shown in Fig. 3a, to minimize the wind effect during snow (see Table 1 for more information). The OTT Pluvio2, version 200, is also supplied with an orifice rim heating system. This reliably keeps the orifice ring rim free of snow and ice during low-temperature operations.
The Vaisala weather transmitter (WXT) 520 measures temperature, relative humidity, and wind speed and direction. The wind sensor has an array of equally spaced ultrasonic transducers on a horizontal plane. The wind speed and direction are determined by measuring the time it takes the ultrasound to travel from one transducer to the other. The instrument uses separate sensors for measuring temperature and relative humidity. The accuracy and resolution of these measurements are given in Table 1.
The manual gauges used for measuring the total precipitation and rain amounts are the Type B and Canadian Nipher gauges shown in Fig. 3b. The Type B and Nipher gauges are the standard instruments for rain and snow observations, respectively, in Canada (Mekis and Vincent 2011; Metcalfe and Goodison 1993). The snow amount in its water equivalent form is determined using the Canadian Nipher gauge. The falling snow is collected and melted and then measured by pouring it into a graduated cylinder. The rainfall amount is measured using the Type B rain gauge. Inside the Type B gauge, there is a graduated cylinder that holds up to 25 mm of rain. Rainfall of more than 25 mm can be made by measuring the overflow of the cylinder into the surrounding container. The manually measured data were collected every 6 h and segregated as solid, mixed, and liquid phases. More details about how the precipitation phase was segregated are given in section 3. As well the depths of snow accumulated over a Weaver snow board were measured by taking the average of 10 snow ruler measurements in an undisturbed area around the meteorological compound.
3. Data analysis and results
a. Frequency distribution of precipitation type based on 1-min data
Figure 4 shows the frequency distribution of precipitation type (PT) reported by the PWD22 and FS11P present weather sensors based on 1-min data between September 2014 and August 2015. The weather symbols shown in Fig. 4 and Table 2 are C, S, SP, IP, SG, IC, R, ZR, ZL, P, RLS, L, and RL represent clear, snow, snow pellets, ice pellets, snow grains, ice crystals, rain, freezing rain, freezing drizzle, unknown, mixed, drizzle, and rain and drizzle, respectively. When no precipitation was reported (~88% of the time) or the clear case, the PWD22 agreed well with the FS11P. Based on the PWD22 and FS11P sensors, the frequency of precipitation events during the entire measurement period was quite low, only 12%. Most of the reported precipitation types were snow with a frequency of 8% and followed by rain 3%. There were only a few cases of L and IP reported (see Fig. 4).
The frequency distribution of precipitation type PT reported by the PWD22 and FS11P present weather sensors reported based on 1-min data between September 2014 and August 2015.
Citation: Journal of Atmospheric and Oceanic Technology 34, 1; 10.1175/JTECH-D-16-0088.1
The precipitation amounts measured for solid, liquid, and mixed events.
b. Frequency distribution of precipitation type compared with human observer
For this comparison hourly data collected during the same period mentioned in section 3a were used. From the present weather sensors, hourly precipitation type reports were extracted by using the nearest to human observer time via an interpolation method. It is customary to segregate precipitation phases using temperature; for example, Rasmussen et al. (2011) assumed precipitation phase as solid for (T < 0°C), liquid for (T > 4°C), and mixed for other temperatures. To test such assumptions, the hourly precipitation type frequency data were also segregated based on these temperature intervals as shown in Fig. 5. When all the temperatures were included, the observer reported no precipitation with a frequency of 86%, which was very close to the frequency of ~88% that was reported by both the PWD22 and FS11P, which is similar to the 1-min data. The human observer reported a frequency of 9.8% for snow events and 3.5% for rain events as compared to ~8% and 4% for snow and rain events, respectively, reported by both the PWD22 and FS11P sensors. This shows that the observer reported only ~1% more snow cases, indicating excellent agreement with the sensors. According to the observer, a relatively small frequency of drizzle cases (0.15%) occurred during the observation period, and the optical sensors reported more drizzle cases (~0.4%). Both optical sensors, the PWD22 and FS11P probes, reported significantly more ice pellets events with frequencies of 0.14% and 0.2%, respectively, as compared to the value reported by the human observer, which was 0.024%. According to the observer, there were a small number of solid precipitation cases associated with IC and SG that were not seen by the probes (see Fig. 5). The observer reported some freezing rain, freezing drizzle, and mixed-phase cases, but the sensors are not capable of reporting these events. When the data were segregated for T > 4°C, the occurrence of no-precipitation events increased from 86% to 94%, and the rain events also increased as expected, but based on the observer, snow and mixed-phase events were not totally eliminated under this condition. Similarly, when the temperatures were below freezing, although the proportion of snow events increased as compared to the unsegregated data (by a factor of 2.5 based on the observer), there were still a few reported cases of rain, freezing precipitation, and mixed-phase events. When the temperatures were between 0° and 4°C, based on the human observer the proportion of the mixed-phase events increased from 0.2% to 1.13%, which was quite significant, but the rain events were also increased by more than a factor of 1.6 as compared to the unsegregated data. Based on the human observation, it was only when T < −2°C that all the rain and drizzle cases were eliminated, but the liquid phase in the form of freezing rain and freezing drizzle was still reported (see Fig. 5). The sensors appeared to be reporting the ZL and ZR cases as rain. Therefore, based on this study, identification of precipitation type based on temperature alone could be misleading, particularly for near-freezing temperatures.
The frequency distributions of precipitation type PT based on the PWD22 measurements for various temperatures.
Citation: Journal of Atmospheric and Oceanic Technology 34, 1; 10.1175/JTECH-D-16-0088.1
c. Precipitation
1) Before consideration of the wind effect
The data used for these comparisons include 1-min-averaged precipitation measured using the Vaisala FS11P and PWD22, and OTT Pluvio2 gauges, and 6-hourly manually measured liquid water equivalent (LWE) and rain amounts measured using the Nipher shield and Type B gauges, respectively. The data were collected during the same September 2014–August 2015 observation period.
The time series of the total precipitation intensities (Fig. 6a), accumulations and type (Fig. 6b), temperature (Fig. 6c), and wind speed (Fig. 6d) for the entire measurement period is given. In Fig. 6d, the red line represents median filtered wind speed data using a 6-h window for clarity. The monthly amounts of total precipitation accumulation are given in Fig. 6b. In September the present weather sensor indicated all liquid phase precipitation except for one ice pellet case (Fig. 6b). For the liquid phase, all sensors are expected to perform well, particularly the catchment-type gauges (Pluvio2 and Type B), since the wind effect is minimal. Both optical sensors, the FS11P and PWD22, agreed with the Type B manual gauge, but the Pluvio2 measured a relatively lower amount (Fig. 6b). The precipitation amount measured in October 2014 was relatively low and mostly occurred in liquid form with some snow cases, since the temperature remained mainly above freezing. In this case the Pluvio2 agreed with the total precipitation measured manually using the Nipher and Type B gauges. The two sensors, PWD22 and FS11P, measured a much larger precipitation amount. Relatively large amounts of snowfall occurred in November. There were also a few cases of rain and drizzle during the early period of the month, when the temperatures were warmer. The FS11P agreed with the manual measurement remarkably well within 98% of the manual measurements (Fig. 6b, November 2014). On the other hand, the Pluvio2 gauge collected only 50% of the amount that was collected manually and the PWD22 measured about 43% higher than the manual measurements (Fig. 6b, November 2014). The underestimation of the snow amount by the Pluvio2 gauge is associated with wind speed, since the wind speed sometimes reached 10 m s−1 (Fig. 6d), and this will be discussed in more depth later. As with October, December was a relatively dry month, only 5 mm of snow was measured based on the manual measurements, but the FS11P and PWD22 measured higher amounts. The precipitation intensities measured in December and also in October were mostly light and hence the optical probes measured higher amounts of precipitation (Fig. 6b), indicating that noncatchment-type optical sensors are more sensitive than the catchment-type gauges. There were some warmer periods in December: the temperature ranged from −30°C to almost 10°C, exhibiting very large temperature variations (Fig. 6c). Similar temperature variations also occurred in January 2015 and were associated with mixed-phase precipitation (drizzle, rain, ice pellets, and snow). Under this mixed-phase condition, the total amounts of precipitation measured in January were 15, 19, 26, and 26 mm using the Pluvio2 gauge, manual method, and PWD2 and FS11P optical probes, respectively, indicating that the Pluvio2 gauge underestimated the precipitation relative to the other probes (Fig. 6b, January 2015). The temperature in February mostly remained below freezing and hence the precipitation type measured was mainly snow and ice pellets. In this case the snowfall amount measured using the PWD22 sensor (37 mm) was closer to the manual measurement (30 mm) compared to the relatively smaller amounts measured with the FS11P sensor (23 mm) and Pluvio2 gauge (19 mm). In March 2015, the temperature varied from −35° to 16°C and received mixed precipitation, although not very significant, and the measured values ranged only from approximately 8 to 10 mm, indicating that all the instruments showed similar performance. In April the temperature varied from −15° to 25°C and the precipitation phase was mixed, and relatively significant total precipitation amounts of 64, 57, 40, and 38 mm were measured by the FS11P, PWD22, Pluvio2, and manual gauges, respectively. In this case the Pluvio2 and manual gauges agreed reasonably well. In May and June, the precipitation phase was mainly rain except for a few snow and ice pellet cases that occurred during the early part of May. For both months the Pluvio2 and manual gauges agreed reasonably well, indicating that when rain dominates, the two probes measure similar amounts of precipitation.
(a) The time series of the total precipitation intensities, (b) accumulations and type, (c) temperature, and (d) ws for the entire measurement period. In (d), the red line represents the median filtered wind speed data using a 6-h window. (bottom) Monthly precipitation amount measured using all the instruments.
Citation: Journal of Atmospheric and Oceanic Technology 34, 1; 10.1175/JTECH-D-16-0088.1
In July and August 2015, the Pluvio2 and Type B both measured 34 mm of rain in August, and 72 and 76 mm, respectively, in July, indicating good agreement. The optical probes, PWD22 and FS11P, both measured relatively higher rain amounts of 86 and 41 mm for July and August, respectively. The ratios of the total precipitation relative to the manual measurements were 1.21, 1.24, and 0.87 for the FS11P, PWD22, and Pluvio2, respectively, indicating that the FS11P and PWD22 overestimate precipitation by 21% and 24%, respectively, and the Pluvio2 underestimates the amount by 13% as compared to the manually measured value. The underestimation of the Pluvio2 could be partly attributed to wind-induced loss during snowfall (Rasmussen et al. 2012 and references therein). Since the optical probes are not expected to be significantly affected by wind, the overestimation of the precipitation amount by these probes could be attributed to the fact that they are more sensitive than the manual gauges. Validity of these possibilities will be explored in the next section.
2) Determination of precipitation type for 6-hourly precipitation data
To assess the effect of wind on the collection efficiency of the gauges, it is necessary to use 6-hourly precipitation amounts measured using the manual gauges and the present weather sensors because manual measurements are available only on a 6-hourly time scale. Since the sensor-based available observed precipitation type data have a 1-min time resolution, and there is an hourly time resolution for the human observer, it is challenging to identify a 6-hourly precipitation type using these datasets. As discussed earlier, for relatively warmer near-freezing temperatures, it is particularly difficult to segregate the precipitation phase. However, the manual 6-hourly precipitation data are already segregated as solid, liquid, and mixed. This was normally done by a human observer using the hourly weather observations, 6-hourly snow depth, and total precipitation gauge (Nipher shield gauge) measurements. For all snow cases, the snowfall amount is determined by using the Canadian Nipher gauge as described in section 2. For mixed-phase cases, the amount of rain is estimated by subtracting the liquid water equivalent estimate based on the measured snow depth assuming a 10:1 snow water ratio from the total precipitation measured using the Canadian Nipher. Thus, in mixed-phase cases, there is some uncertainty associated with the 10:1 snow density assumption. Figure 7 shows the fraction of LWE precipitation amount plotted against the observed mean temperature based on 6-hourly data. In the mixed-phase cases, the solid fraction approximately linearly increased from 0.3 to 0.9 with decreasing temperature from about 4° to −4°C, but there were also all solid and all liquid cases within this temperature interval. In this study, the precipitation phase is segregated based on solid fraction as shown in Fig. 7.
The observed fraction of LWE precipitation amount plotted against the observed mean temperature.
Citation: Journal of Atmospheric and Oceanic Technology 34, 1; 10.1175/JTECH-D-16-0088.1
3) Sensitivity
Figure 8 shows the frequency distributions of 1-min-averaged precipitation intensities for solid precipitation (Fig. 8a), and the associated frequency distributions for temperature (Fig. 8b) and wind (Fig. 8c), and the distribution of liquid phase intensities (Fig. 8d), and the associated temperature and wind distributions are given in Figs. 8e and 8f, respectively. The snow and rain cases were identified based on 1-min PWD22 data. The Pluvio2 gauge reported more no-precipitation events (94%) as compared to the optical probes (~70%) in solid phase, and 78% in liquid phase as compared to 41% in rain cases for the optical probes, indicating the Pluvio2 gauge missed some precipitation events. The number of light snow and rain-rate (<1 mm h−1) cases were much lower for the Pluvio2 than measured by the PWD22 and FS11P, but the number of cases that the Pluvio2 measured increased for higher precipitation intensities (>2 mm h−1). In fact, the Pluvio2 gauge appears to report more precipitation cases for rates higher than 2 mm h−1 for snow. This could have some compensating effect in the total amount. The temperature distribution during snow varied from about 0° to −30°C with a maximum near −10°C, and during the liquid phase the temperature varied from −5° to 20°C with a maximum near 15°C. The wind speed distributions were identical during both snow and liquid phase cases, varying from near 0 to 11 m s−1. For comparison, the sensitivity of these instruments during snow and rain were also investigated using 10-min-averaged data. In this case the precipitation phase was identified based on the 10-min-averaged temperature (snow for T < −4°C and rain for T > 4°C) and the results are given in Fig. 9. The figure shows the fraction of the snowfall-rate (Fig. 9a) and rain-rate (Fig. 9b) contribution to the perspective total snow and rain amount, respectively. As depicted in the figure, for 10-min-averaged data, the light snowfall rate (<≈0.4 mm h−1) contributes a relatively small fraction to the total amount for the Pluvio2 (≈20%) as compared to almost 40%–50% for the optical FS11P and PWD22 probes (Fig. 9a). For 10-min-averaged data, the sensitivity of Pluvio2 for rain is relatively comparable to the optical probes (Fig. 9b).
(a) The frequency distributions of precipitation intensities for solid, and the associated frequency distributions for (b) temperature and (c) wind; (d) the distribution of liquid phase intensities; and the associated (e) temperature and (f) wind distributions.
Citation: Journal of Atmospheric and Oceanic Technology 34, 1; 10.1175/JTECH-D-16-0088.1
The fractional contribution of the observed precipitation rate to the total amount for (a) solid phase and (b) liquid phase.
Citation: Journal of Atmospheric and Oceanic Technology 34, 1; 10.1175/JTECH-D-16-0088.1
4) The effect of wind and corrections
Figure 10 shows the 6-hourly precipitation amounts measured using the PWD22, FS11P, and Pluvio2 plotted against the manual measurements for solid precipitation (Figs. 10a–c) and liquid precipitation (Figs. 10d–f). For the liquid phase precipitation, all instruments agreed quite well with a correlation coefficient (R) close to 0.96. For the solid phase precipitation, however, the correlation coefficient varied from 0.62 for the FS11P to 0.88 for the Pluvio2, indicating that the Pluvio2 was correlated with the manual gauge better than the optical gauges. However, as depicted in Figs. 10a–c, there is significant scatter for the solid phase case, particularly for the Pluvio2 gauge when LWE rates < 1 mm h−1. This scatter is mainly due to the wind effect and also the type of snow (Thériault et al. 2015, 2012; Colli et al. 2015, 2016a,b).
The 6-hourly precipitation amounts measured using the PWD22, FS11P, and Pluvio2 plotted against the manual measurements for (a)–(c) solid precipitation and (d)–(f) liquid precipitation.
Citation: Journal of Atmospheric and Oceanic Technology 34, 1; 10.1175/JTECH-D-16-0088.1
The precipitation amounts measured by each instrument and for each phase are given in Table 2. Generally, the measured liquid phase precipitation amounts were larger than the solid phase and mixed-phase amounts. Overall, the instruments measured similar amounts in liquid precipitation as compared to the solid phase case, particularly the Pluvio2 gauge. The collection efficiencies of the instruments as compared to the manual gauges are given in Table 3. The collection efficiency of the Pluvio2 gauge was 0.98 for the liquid phase and 0.84 for the mixed-phase cases, indicating relatively good agreement with the manual measurements. During the solid precipitation events, the Pluvio2 gauge underestimated the precipitation by 40%, which is quite significant and suggests the effect of wind. The PWD22 sensor overestimated the amount as compared to the manual measurements by 33% during solid phase and 21% during liquid precipitation events. The FS11P probe agreed with the manual measurements in the solid phase but overestimated the amount by about 30% in the liquid phase. The measurement differences that were observed between these instruments during liquid precipitation events could be partly associated with the difference in the sensitivity of the instrument, but generally such small differences are expected and hence it can be assumed that the instruments agreed well within measurement uncertainty. However, the undercatch by the Pluvio2 gauge in snow is significant and this is investigated in the following section.
The collection efficiency of the instruments as compared to the manual gauges before correction for wind speed.
Transfer function CENipher proposed by Goodison et al. (1998) is plotted as function of the observed wind speed during snow.
Citation: Journal of Atmospheric and Oceanic Technology 34, 1; 10.1175/JTECH-D-16-0088.1
The collection efficiency of the Pluvio2 data relative to the corrected data [given in Eq. (1)] plotted against (a) the observed mean wind speed and (b) the number of data points averaged for each wind speed interval.
Citation: Journal of Atmospheric and Oceanic Technology 34, 1; 10.1175/JTECH-D-16-0088.1
Comparisons of corrected values of precipitation data from the Nipher gauge and Pluvio2 and measured using the PWD22 and FS11P are plotted against the number of hours of solid precipitation during the entire measurement period.
Citation: Journal of Atmospheric and Oceanic Technology 34, 1; 10.1175/JTECH-D-16-0088.1
5) Parameterization of the frequency distribution of precipitation intensity
Based on the analysis discussed earlier in this study, it was demonstrated that the PWD22 performed quite well as compared to the corrected manual Nipher gauge measurements. Thus, for this statistical analysis, the entire 1-min-averaged dataset was used. The PWD22 data were segregated as rain, solid, and drizzle precipitation types measured using the same instrument. To test the lognormality of the observed precipitation intensity, the frequency distribution of the log of the observed precipitation intensity [
(top) The frequency distribution of the natural log of the observed precipitation rate values for (a) rain, (b) snow or solid, (c) drizzle, and (d) total precipitation. The black line represents the best-fit curve for a normal distribution. (bottom) As in (top), but for the frequency distribution of precipitation rate values. The LogN and InvG are the best-fit curves for lognormal and inverse Gaussian, respectively.
Citation: Journal of Atmospheric and Oceanic Technology 34, 1; 10.1175/JTECH-D-16-0088.1
The best-fit coefficients for normal, lognormal, and inverse-Gaussian pdfs.
4. Snow density
(a) Liquid wqater equivalent 
Citation: Journal of Atmospheric and Oceanic Technology 34, 1; 10.1175/JTECH-D-16-0088.1
5. Summary and conclusions
In this paper precipitation amount and type, and snow depth data collected during the 4Wing Cold Lake Research Project covering the period from September 2014 to August 2015 using various instruments, including the Vaisala PWD22 and FS11P sensors, the OTT Pluvio2 gauge, the Canadian Nipher and Type B manual gauges, and a snow rule measuring depth, have been analyzed. The analysis indicated that most (80% of the total) of the measured snow intensities, by LWE, were relatively low (<1.5 mm h−1; Fig. 9). During snow events the observed temperature varied from −35° to 4°C and the wind speed reached 12 m s−1 with a mean value of 4 m s−1 and a standard deviation of 2 m s−1. Based on the 1-min-averaged data collected using the optical PWD22 and FS11P sensors, the precipitation events during the entire measurement period represented only 12% of the total time. Most of the reported precipitation types were snow with a frequency of 8% followed by rain at 3% of the time. There were only a few cases of drizzle and ice pellet events. The precipitation types reported by the present weather sensors were also compared with the hourly human observation data for various temperature intervals. When all the temperatures were included, the observer reported no precipitation with a frequency of 86%, which was very close to the frequency of ~88% that was reported by both the PWD22 and FS11P. The human observer reported snow and rain 10% and 4% of the time, respectively, as compared to approximately 8% and 4% measured by both PWD22 and FS11P sensors. This shows that the observer reported only ~1% more snow cases, indicating excellent agreement with the sensors. According to the observer, drizzle cases occurred 0.15% of the time during the observation period, which was very close to the value from the sensors (~0.4%). Both the PWD22 and FS11P probes reported significantly more ice pellets events with frequencies of 0.14% and 0.2%, respectively, as compared to the value reported by the human observer, which was 0.024% of the time.
The amounts of 6-hourly precipitation measured using the Pluvio2 gauge with a single Alter shield, Canadian Nipher and Type B gauges, and the Vaisala PWD22 and FS11P sensors were also compared. The comparisons revealed that the collection efficiencies of the Pluvio2 gauge as compared to the manual measurements were 0.98 and 0.83 for liquid and mixed-phase cases, respectively, indicating relatively good agreement with the manual observations. For the solid phase, however, the Pluvio2 gauge significantly undercaught with a ratio of only 0.57, possibly due to wind effects. The Vaisala PWD22 overestimated the amount as compared to the manual measurements by 33% and 21% during solid and liquid precipitation events, respectively. The FS11P probe agreed with the manual measurements during snow, but it overestimated the amount by about 30% during rain. After correcting for the effect of wind during snow for both the Canadian Nipher and the Pluvio2 gauge, the data agreed remarkably well with the PWD22 measurements, but the FS11P measurement was still slightly lower. Thus, these findings have demonstrated the usefulness of emerging technologies, such as the PWD22 and FS11P probes, showing that they can be used for measuring snowfall in cold climates, where the snowfall intensity tends to be relatively low.
After demonstrating the good performance of the PWD22, the 1-min-averaged precipitation intensity frequency distributions were obtained using this probe, including the solid and liquid phases. Several probability density functions, including gamma, normal, lognormal, and inverse Gaussian, were used to fit the observed frequency distributions. The goodness of the fits were tested using null hypothesis chi-square statistics, and based on these tests it was found that the observed precipitation intensities can be described by lognormal and inverse-Gaussian distributions.
Using the snow depth and corrected Nipher gauge data, snow densities were derived. The snow density or snow-to-liquid ratio varied from 4.2 to 35 with a mean value of 12.2 or 0.082 g cm−3, which suggests that the mean value derived in this study is higher than the 10:1 ratio usually assumed for converting snow depth to snow water equivalent in Canada. On average, the snow density depends on temperature, increasing with increasing temperature, and the 10:1 ratio appears to be more appropriate for relatively warmer temperatures.
Acknowledgments
This work was partially funded by the Department of National Defence (DND) and the Canadian National Search and Rescue New Initiative Fund (SAR-NIF) under the SAR Project (SN201532). We like to thank Mike Harwood and Robert Reed for helping with the installation of the instruments at Cold Lake. The authors also would like to extend their thanks to Ramond Dooley, Amy Slade-Campbell, and Gordon Lee of DND for providing the hourly precipitation data, and also Randy Blackwell for helping during the instrument installation process.
REFERENCES
Baxter, M. A., C. E. Graves, and J. T. Moore, 2005: A climatology of snow-to-liquid ratio for the contiguous United States. Wea. Forecasting, 20, 729–744, doi:10.1175/WAF856.1.
Boudala, F. S., G. A. Isaac, R. Rasmussen, S. G. Cober, and B. Scott, 2014a: Comparisons of snowfall measurements in complex terrain made during the 2010 Winter Olympics in Vancouver. Pure Appl. Geophys., 171, 113, doi:10.1007/s00024-012-0610-5.
Boudala, F. S., R. Rasmussen, G. A. Isaac, and B. Scott, 2014b: Performance of hot plate for measuring solid precipitation in complex terrain during the 2010 Vancouver Winter Olympics. J. Atmos. Oceanic Technol., 31, 437–446, doi:10.1175/JTECH-D-12-00247.1.
Brandes, E. A., K. Ikeda, G. Zhang, M. Schönhuber, and R. M. Rasmussen, 2007: A statistical and physical description of hydrometeor distributions in Colorado snowstorms using a video disdrometer. J. Appl. Meteor. Climatol., 46, 634–650, doi:10.1175/JAM2489.1.
Cho, H.-K., K. P. Bowan, and G. R. North, 2004: A comparison of gamma and lognormal distributions for characterizing satellite rain rates from the Tropical Rainfall Measuring Mission. J. Appl. Meteor., 43, 1586–1597, doi:10.1175/JAM2165.1.
Colli, M., R. Rasmussen, J. M. Thériault, L. Lanza, C. Baker, and J. Kochendorfer, 2015: An improved trajectory model to evaluate the collection performance of snow gauges. J. Appl. Meteor. Climatol., 54, 1826–1836, doi:10.1175/JAMC-D-15-0035.1.
Colli, M., L. Lanza, R. Rasmussen, and J. M. Thériault, 2016a: The collection efficiency of shielded and unshielded precipitation gauges. Part I: CFD airflow modeling. J. Hydrometeor., 17, 231–243, doi:10.1175/JHM-D-15-0010.1.
Colli, M., L. Lanza, R. Rasmussen, and J. M. Thériault, 2016b: The collection efficiency of shielded and unshielded precipitation gauges. Part II: Modeling particle trajectories. J. Hydrometeor., 17, 245–254, doi:10.1175/JHM-D-15-0011.1.
Dan'azumi, S., S. Shamsudin, and A. Aris, 2010: Modeling the distribution of rainfall intensity using hourly data. Amer. J. Environ. Sci., 6, 238–243, doi:10.3844/ajessp.2010.238.243.
Goodison, B. E., P. Y. T. Louie, and D. Yang, 1998: WMO Solid Precipitation Measurement Intercomparison. Final Rep., World Meteorological Organization Instruments and Observing Methods Rep. 67, WMO/TD-872, 211 pp.
Jonas, J., C. Marty, and J. Magnusson, 2009: Estimating the snow water equivalent from snow depth measurements in the Swiss Alps. J. Hydrol., 378, 161–167, doi:10.1016/j.jhydrol.2009.09.021.
Kedem, B., and L. S. Chiu, 1987: On the lognormality of rain rate. Proc. Natl. Acad. Sci. USA, 84, 901–905, doi:10.1073/pnas.84.4.901.
Kedem, B., H. Pavlopoulos, X. Guan, and D. A. Short, 1994: Probability distribution model for rain rate. J. Appl. Meteor., 33, 1486–1493, doi:10.1175/1520-0450(1994)033<1486:APDMFR>2.0.CO;2.
Mekis, É., and L. A. Vincent, 2011: An overview of the second generation adjusted daily precipitation dataset for trend analysis in Canada. Atmos.–Ocean, 49, 163–177, doi:10.1080/07055900.2011.583910.
Metcalfe, J. R., and B. E. Goodison, 1993: Correction of Canadian winter precipitation data. Preprints, Eighth Symp. on Meteorological Observations and Instrumentation, Anaheim, CA, Amer. Meteor. Soc., 338–343.
Potter, J. G., 1965: Water content of freshly fallen snow. CIR-4232, TEC-569, Meteorology Branch, Dept. of Transport, Toronto, ON, Canada, 12 pp. [Available from National Snow and Ice Data Center User Services, University of Colorado, Campus Box 449, Boulder, CO 80309-0449.]
Rasmussen, R. M., J. Hallett, R. Purcell, S. D. Landolt, and J. Cole, 2011: The hotplate precipitation gauge. J. Atmos. Oceanic Technol., 28, 148–164, doi:10.1175/2010JTECHA1375.1.
Rasmussen, R. M., and Coauthors, 2012: How well are we measuring snow: The NOAA/FAA/NCAR winter precipitation test bed. Bull. Amer. Meteor. Soc., 93, 811–829, doi:10.1175/BAMS-D-11-00052.1.
Roebber, P. J., S. Bruening, D. M. Schultz, and J. V. Cortinas Jr., 2003: Improving snowfall forecasting by diagnosing snow density. Wea. Forecasting, 18, 264–287, doi:10.1175/1520-0434(2003)018<0264:ISFBDS>2.0.CO;2.
Sevruk, B., and S. Klemm, 1989: Catalogue of national standard precipitation gauges. Tech. Rep. WMO, Instruments and Observing Methods, Rep. 39, WMO, Geneva, Switzerland, 50 pp. [Available online at https://www.wmo.int/pages/prog/www/IMOP/publications/IOM-39.pdf.]
Sturn, M., B. Taras, G. E. Liston, C. Derksen, T. Jonas, and J. Lea, 2010: Estimating snow water equivalent using snow depth data and climate classes. J. Hydrometeor., 11, 1380–1394, doi:10.1175/2010JHM1202.1.
Thériault, J., R. Rasmussen, K. Ikeda, and S. Landolt, 2012: Dependence of snow gauge collection efficiency on snowflake characteristics. J. Appl. Meteor. Climatol., 51, 745–762, doi:10.1175/JAMC-D-11-0116.1.
Thériault, J., R. Rasmussen, E. Petro, J.-Y. Trépannier, M. Colli, and L. G. Lanza, 2015: Impact of wind direction, wind speed, and particle characteristics on the collection efficiency of the Double Fence Intercomparison Reference. J. Appl. Meteor. Climatol., 54, 1918–1930, doi:10.1175/JAMC-D-15-0034.1.
Wolff, M., K. Isaksen, A. Petersen-Øverleir, K. Ødemark, T. Reitan, and R. Bækkan, 2015: Derivation of a new continuous adjustment function for correcting wind-induced loss of solid precipitation: Results of a Norwegian field study. Hydrol. Earth Syst. Sci., 19, 951–967, doi:10.5194/hess-19-951-2015.
Yang, D., 2014: Double Fence Intercomparison Reference (DFIR) vs. Bush Gauge for “true” snowfall measurement. J. Hydrol., 509, 94–100, doi:10.1016/j.jhydrol.2013.08.052.
Yang, D., S. Ishida, B. E. Goodison, and T. Gunther, 1999: Bias correction of daily precipitation measurements for Greenland. J. Geophys. Res., 104, 6171–6181, doi:10.1029/1998JD200110.
Yang, D., D. Kane, Z. Zhang, D. Legates, and B. Goodison, 2005: Bias corrections of long-term (1973–2004) daily precipitation data over the northern regions. Geophys. Res. Lett., 32, L19501, doi:10.1029/2005GL024057.
