The Tibetan Plateau (TP) has the most abundant water resources (such as glaciers, rivers, and lakes) and the highest altitude in the world. The TP not only plays a significant role in the formation of Asian monsoon circulations but also has a profound impact on the global water cycles, climate, and environment (Xu et al., 2014, 2015; Wan et al., 2017). Clouds and precipitation over the TP are important components of global hydrological cycles and energy budgets (Xu et al., 2008; Kang et al., 2010; Li, 2018). However, the shortage of in situ observations and the low spatiotemporal resolution and uncertainties in satellite measurements have restricted the understanding of the physical properties of clouds and precipitation over the TP (Zhao et al., 2019).
To strengthen the observations of clouds and precipitation over the TP, three Tibetan Plateau Atmospheric Scientific Experiments (TIPEX) were carried out in the summers of 1979, 1998, and 2013 (Liu et al., 2002; Chen et al., 2017; Zhao et al., 2018, 2019). In particular, during the third TIPEX, advanced measurements such as Ka-band cloud radar, X-band dual polarization radar, disdrometers, and microwave radiometers were deployed in Nagqu on the TP to comprehensively analyze the physical properties and climatic characteristics of clouds and precipitation (Liu et al., 2015; Chang and Guo, 2016; Chen et al., 2017).
However, due to the complex topography of the TP, the representativeness of single station observations is very poor. Therefore, the Second Tibetan Plateau Scientific Expedition and Research (STEP) project and the “Earth-Atmosphere Interaction in the TP and its Influence on the Weather and Climate in the Lower Reaches” project were launched to establish several field observation campaign sites at which to examine clouds and precipitation in 2019. In particular, Mêdog, located in front of the main water vapor channel over the TP, is a very important campaign site where an X-band polarization phased array radar, a Ka-band cloud radar, a microwave radiometer, a disdrometer, and other instruments were deployed intermittently by the Chinese Academy of Meteorological Sciences (CAMS). One of the precise scientific objectives at this site is to obtain the development and precipitation characteristics of convective clouds in the valley of the Yarlung Zangbo Grand Canyon (YZGC).
The raindrop size distribution (DSD) has received much attention over the past few decades due to its great importance in reflecting the fundamental microphysics of rainfall (Rosenfeld and Ulbrich, 2003). A better understanding of the DSD and its variation is not only critical for microphysical parameterizations in numerical weather prediction models (Milbrandt and Yau, 2005; Morrison and Milbrandt, 2015) but is also important for the remote sensing of precipitation (Cifelli et al., 2011; Chen et al., 2017). Microphysical parameterization is a key element in numerical models that affects the prediction accuracy of convective systems (Gilmore et al., 2004; Krishna et al., 2016). Quantitative precipitation estimations (QPEs) from ground-based weather radar or spaceborne satellite observations depend on the characteristics of the DSD to develop rainfall retrieval algorithms (Zhang et al., 2001; Chandrasekar et al., 2005; Lam et al., 2015; Ji et al., 2019). To this end, numerous DSD observations have been conducted around the world to elucidate the variability in DSDs among different climate regions and rainfall types (Tokay and Short, 1996; Yuter and Houze, 1997; Maki et al., 2001; Testud et al., 2001; Bringi et al., 2003; Zhang et al., 2003; Thurai, et al., 2010; Lam et al., 2015; Chen et al., 2016, 2017; Wu et al., 2019; Ji et al., 2019).
In the last several decades, many DSD studies have also been conducted over various regions in China using optical disdrometers. Most of these studies were carried out in eastern and southern China (Niu et al., 2010; Chen et al., 2013, 2016; Tang et al., 2014; Wang et al., 2015; Wen et al., 2016; Wu and Liu, 2017; Huo et al., 2019). Recently, the DSD characteristics over the TP were studied using disdrometer data collected at Lhasa [3600 m above sea level (ASL)] and Nyingchi (3300 m ASL), and it was revealed that collisional breakup occurred at a lower rainfall intensity and with a smaller maximum raindrop size than that in low-altitude regions (Porcù et al., 2014). Based on DSD measurements taken at Nagqu (31.29°N, 92.04°E; 4508 m ASL) during the third TIPEX, Chen et al. (2017) showed that convective precipitation was characterized by smaller generalized intercepts (Nw) and larger mass-weighted mean diameters (Dm) in the daytime than at nighttime. However, complex topography and underlying surface characteristics over the TP limit the representativeness of observations from any specific station.
In June 2019, a particle size and velocity (PARSIVEL) disdrometer was deployed at the Mêdog National Climate Observatory (MNCO; 29.31°N, 95.32°E; 1275 m ASL) to perform continuous raindrop spectra measurements. Mêdog and Nagqu are two typical regions of the TP with different geographical locations and climate regimes (Fig. 1). Mêdog is located on the southern slope of the Himalayas at the entrance of the water vapor transport channel of the YZGC, which is the most important water vapor channel through which the Indian Ocean monsoon affects precipitation over the TP (Yang et al., 1987; Zhang et al., 2016). Mêdog has a mean altitude of 1200 m ASL and subtropical climatic characteristics. The warm and wet water vapor from the Indian Ocean leads to a large amount of total precipitation in Mêdog with an annual average rainfall of more than 2000 mm (Chen and Li, 2018). Precipitation is mainly concentrated from June to September, accounting for 64% of the total annual precipitation. Nagqu is located in the center of the TP, with a mean altitude exceeding 4500 m ASL and a plateau mountain climate. Its mean annual precipitation is approximately 400 mm, and over 80% of the total precipitation occurs in summer (Chen et al., 2017). In summer, the Indian Ocean monsoon brings abundant water vapor to Mêdog, while the Nagqu region experiences interactions between the westerly wind and Indian Ocean monsoon (Yang et al., 1987; Zhang et al., 2016; Zeng et al., 2020).
Figure 1. Locations of the Mêdog and Nagqu observation fields (black dots), the topography (m, shaded) of the Tibetan Plateau (TP) superimposed with the mean vertical integral of the water vapor flux obtained in summer of (a) 2014 and (b) 2019 (kg m−1 s−1, thin black arrows), and the trajectories of the water vapor (thick gray arrows).
Therefore, the objectives of this study are to (i) explore if there were any distinct discrepancies of DSD characteristics and precipitation microphysical processes between Mêdog and Nagqu, considering the different sources of water vapor and topography in the two regions; (ii) investigate the possible causative meteorological factors if a notable DSD discrepancy does exist between the two regions of the TP; and (iii) better understand the DSD characteristics in Mêdog and Nagqu over the TP, providing a basis for improving the microphysical parameterization scheme of the numerical model over the TP. These objectives will be achieved through a comparative study on the DSD variations in precipitation between Mêdog and Nagqu based on DSD measurements taken in Mêdog during the STEP field campaign and in Nagqu during the third TIPEX. In addition to the PARSIVEL disdrometer, the automated weather station (AWS) data of the China Meteorological Administration, Moderate Resolution Imaging Spectroradiometer (MODIS) data products, and the European Centre for Medium-Range Weather Forecasts (ECMWF) Reanalysis version 5 (ERA5) are combined to illustrate the microphysical characteristics of precipitation in two regions of the TP.
The instruments and methodology adopted in this study are described in section 2. The observational results in terms of the DSD characteristics of different rainfall rates and precipitation types in Mêdog and Nagqu are presented in section 3. The possible reasons for the observed variations in the DSDs of the two regions of the TP are discussed in section 4. A summary and conclusion are given in the final section.
To examine the precipitation characteristics in the two regions of the TP, the DSD observations from Mêdog and Nagqu were divided into five categories on the basis of rainfall rate (R): R ≤ 0.1 mm h−1, 0.1 < R ≤ 1 mm h−1, 1 < R ≤ 5 mm h−1, 5 < R ≤ 10 mm h−1, and R > 10 mm h−1. The accumulated rain amounts (averaged rain rates) of the five categories in Mêdog are 9.04 mm (0.034 mm h−1), 143.12 mm (0.42 mm h−1), 346.14 mm (1.98 mm h−1), 74.42 mm (6.73 mm h−1), and 63.43 mm (17.62 mm h−1), respectively. Nagqu has accumulated rain amounts (averaged rain rates) of 2.90 mm (0.043 mm h−1), 52.20 mm (0.42 mm h−1), 234.13 mm (2.31 mm h−1), 93.16 mm (6.75 mm h−1), and 155.57 mm (22.66 mm h−1), respectively. Figure 4 gives the relative contributions of the five rainfall rate categories to the cumulative rainfall durations and rainfall totals in Mêdog and Nagqu. In Mêdog, weak precipitation with R < 1 mm h−1 often occurred, contributing more than 70% of the rainfall occurrences recorded during the observation periods in this study. In Nagqu, the second and third rainfall rate categories (0.1 < R ≤ 5 mm h−1) were the two largest contributors and were responsible for 72% of the cumulative rainfall durations. The largest contributor to the total rainfall amount was the third category (1 < R ≤ 5 mm h−1) in both regions; this category was responsible for 54% and 40% of the cumulative rainfall totals in Mêdog and Nagqu, respectively.
Figure 4. Relative contributions of each rainfall rate category to the (a) cumulative rainfall durations (min) and (b) cumulative rainfall totals (mm) in Mêdog and Nagqu on the TP.
The mean DSDs of the five rainfall rate categories and the total datasets in Mêdog and Nagqu are depicted in Fig. 5. In general, in the DSDs of the two regions of the TP, both the spectral widths and the large-drop concentrations increase with rainfall rate. It is also evident from the figure that the number concentrations of small drops are higher in Mêdog than in Nagqu for all rainfall rate categories. However, the number concentration of large drops is higher in Nagqu than in Mêdog. This discrepancy in the number concentration of large drops between Nagqu and Mêdog increases with increased rainfall rates. Distinct differences in the DSDs between Mêdog and Nagqu are noticeable in the rainfall rate categories above 5 mm h−1. In Mêdog, lower rainfall rate categories (≤ 5 mm h−1) show one peak distribution, and higher rainfall rate categories (> 5 mm h−1) show two peak distributions (e.g., peaks at 0.5 mm and 1.1 mm). However, all rainfall rate categories in Nagqu show one apparent peak distribution. The multipeak character of DSD has been studied based on different disdrometer measurements at different locations in Switzerland from 1982 to 1986 (Steiner and Waldvogel, 1987). Multiple peaks of DSD were also observed by ground-based Doppler radar in Denver (Gossard et al., 1990). Srivastava (1971) pointed out that size distribution of raindrops may not have established equilibrium in the observed falling distance. Therefore, convective rainfall with a melting level at a higher altitude increases the probability for the multipeak behavior because of the long fall distances of raindrops (List et al., 1987).
Figure 5. Comparison of the mean DSDs obtained for different rainfall rate categories in Mêdog (solid lines) to those obtained for Nagqu (dashed lines) on the TP.
The average raindrop concentrations with drop diameters in the total datasets collected in Mêdog and Nagqu are depicted in Fig. 5f. It is apparent that the number concentrations of midsize and large drops are higher in Nagqu than in Mêdog, whereas the number concentration of drops with diameters smaller than 0.6 mm is lower in Nagqu than in Mêdog.
The largest uncertainty in model predictions of convective precipitation originates from microphysical parameterizations (Krishna et al., 2016). Therefore, one of the most important aspects of DSD research is to improve the parameterization scheme of the cloud–precipitation microphysical processes in numerical models. For this purpose, the Mêdog and Nagqu precipitation DSDs observed from disdrometers are fitted to gamma distributions (Eq. 6) by using MOM. The Dm, Nw, μ, and Λ variations based on the rainfall rate categories in Mêdog and Nagqu are given in Fig. 6.
Figure 6. Variations in the average Dm, lgNw, μ and Λ values for each rainfall rate category in Mêdog and Nagqu.
The average Dm values have similar trends at the two sites, continuously increasing with the rainfall rate in both regions. This feature is in line with the results of previous studies and is a result of the enhancement of large raindrops with increasing rainfall rate (Testud et al., 2001; Rosenfeld and Ulbrich, 2003). Nagqu has higher (lower) average Dm (lgNw) values than those of Mêdog for all rainfall rate categories. The discrepancy in the average Dm values between Nagqu and Mêdog increases with rainfall rate and changes from 0.115 mm to 0.568 mm.
The average lgNw values increase up to rainfall rate category four (below 10 mm h−1) and then decrease in both regions. In particular, the differences in both Dm and lgNw between Nagqu and Mêdog were significant when the rainfall rate was higher than 10 mm h−1. A rainfall rate higher than 10 mm h−1 is usually determined by convective clouds (Tokay and Short, 1996; Wu et al., 2019). That is, for convective precipitation with the same rainfall rate (corresponding to the same rainwater content), smaller drops with larger number concentrations were dominant in Mêdog, whereas Nagqu had a higher number concentration of large drops than Mêdog.
The average μ values are higher in Mêdog than in Nagqu, except for those of the first rainfall rate category (R ≤ 0.1 mm h−1). The average μ values in Nagqu show a monotonic decrease with an increasing rainfall rate and range from 8.676 to 0.598. The μ values in Mêdog decrease from 4.572 to 2.720 with an increasing rainfall rate, increase to 3.869 at R ranging from 5 to 10 mm h−1, and then decrease again to 2.963 when R > 10 mm h−1. The variation trends in the Λ values are found to be similar to those in the μ values, which may be due to the μ-Λ relation of ΛDm = 4 + μ. The Λ values range from 4.308–15.328 mm−1 (2.105–18.803 mm−1) in Mêdog (Nagqu).
Studies have shown that the microphysical dynamics of raindrop spectra are significantly different in different precipitation types (Tokay and Short, 1996; Bringi et al., 2003; Ulbrich and Atlas, 2007). Therefore, we investigated the DSD characteristics of stratiform and convective precipitation types in Mêdog and Nagqu. Due to the scarcity of observation instruments on the TP, a simple stratification method proposed by Bringi et al. (2003) based on the standard deviation (STD) of the rainfall rate over ten consecutive 1-min DSD samples was used in this study. If the STD ≤ 1.5 mm h−1, stratiform precipitation is identified; if the STD > 1.5 mm h−1 and R > 5 mm h−1, convective precipitation is identified. As a result, the data from Mêdog and Nagqu consist of 95.1%/1.5% (45 426/707) and 88.2%/5.4% (16 393/995) stratiform/convective rain samples, respectively. For stratiform precipitation in Mêdog/Nagqu, the accumulated rain amount and mean rain rate were 448 mm/266 mm and 0.6 mm h−1/1.0 mm h−1, respectively. For convective precipitation, the accumulated rain amount and mean rain rate were 110 mm/222 mm and 9.3 mm/13.4 mm h−1, respectively.
The relative-frequency histograms of Dm and lgNw values derived from 1-min DSD samples for stratiform and convective precipitation events in Mêdog and Nagqu are given in Fig. 7. Regarding the stratiform precipitation type (Figs. 7a and c), the patterns of the Dm and lgNw distributions in Mêdog are generally close to those in Nagqu, as are the statistical values [mean value (MEAN), standard deviation (STD), and skewness (SKEW)]. However, a discrepancy also shows that Mêdog has a smaller mean Dm (0.84 mm) than that of Nagqu (0.93 mm); additionally, Mêdog has a larger mean lgNw (3.65) than that of Nagqu (3.58). When the raindrop diameter is larger than 1.0 mm, the occurrence frequency of Dm in Nagqu is higher than that in Mêdog, suggesting larger raindrops in stratiform rain in Nagqu. This discrepancy may reflect microphysical differences in stratiform precipitation DSDs from Nagqu and Mêdog. Stratiform precipitation results from the melting of snowflakes and/or tiny, rimed ice particles. The low-density, large snow particles result in DSDs characterized by relatively larger Dm and lower Nw, compared to the tiny, rimed ice particles (Fabry and Zawadzki, 1995; Bringi et al., 2003).
Figure 7. Comparisons of occurrence frequencies between Mêdog (red) and Nagqu (blue): (a) Dm values for stratiform precipitation, (b) Dm values for convective precipitation, (c) lgNw values for stratiform precipitation, and (d) lgNw values for convective precipitation. The units of the Dm and Nw values are mm and mm−1 m−3, respectively. Mean values (MEAN), standard deviations (STD), and skewness (SKEW) are given in the respective panels.
On the other hand, the distributions of Dm and lgNw for convective precipitation show significant differences between Mêdog and Nagqu (Figs. 7b and d). For instance, the Dm histogram representing convective precipitation in Nagqu is much broader than that of Mêdog, ranging from 1.0 mm to 3.5 mm with a mean value of 1.82 mm, which is similar to the range found for Colorado convective cases with a broader spectral width due to microphysical precipitation processes involving the melting of frozen particles (e.g., tiny hailstones and graupel) in the high plains (Bringi et al., 2003). In contrast, the Mêdog convective precipitation has a significantly narrower Dm distribution than that of Nagqu, ranging from 0.9 mm to 2.0 mm with a significantly smaller mean value of 1.33 mm, which is very close to the value of 1.41 mm measured in East China during the summer monsoon season (Wen et al., 2016). This similarity may be related to the warm and humid air currents in the two regions. The lgNw histogram representing Nagqu is quite skewed, with a lower mean value of 3.61 (Nw ~ 4000 mm−1 m−3), whereas the Mêdog histogram is nearly symmetric with a higher mean value of 4.08 (Nw ~ 12 000 mm−1 m−3). On the whole, Mêdog convective rainfall is distinguished by relatively small mass-weighted mean diameter Dm but high normalized intercept parameter Nw, which is similar to the characteristics of tropical convective regimes due to sufficient water vapor supplies producing abundant small particles, whereas Nagqu had relatively larger Dm at lower Nw, reflecting the DSD characteristics of continental convective regimes.
Figures 7c and d show that the bimodality distribution of lgNw is also obvious in Mêdog; this distribution has been found in tropical precipitation cases in previous studies (Ulbrich and Atlas, 2007; Thompson et al., 2015; Dolan et al., 2018). The occurrence frequency of lgNw in Mêdog peaks at 3.6 and at 4.2, corresponding to stratiform and convective precipitation, respectively. This bimodal distribution is lacking in the convective and stratiform rainfall components in Nagqu.
To further understand the characteristics of the Dm and lgNw values of stratiform and convective precipitation types in the two typical regions of the TP, we compared the results between Nagqu and Mêdog, as well as comparing the results with statistical results obtained for other climate regimes (Fig. 8). Two black rectangles in Fig. 8 correspond to the maritime- and continental-like convective clusters defined by Bringi et al. (2003), respectively. Stratiform (convective) precipitation cases are marked with blue (red) color symbols. The results indicate that the summer convective precipitation in Mêdog is maritime-like, exhibiting smaller Dm and higher lgNw values, whereas the summer convective events in Nagqu could be identified as continental-like, characterized by relatively larger Dm and lower lgNw values. Figure 8 also shows that the mean lgNw values versus the mean Dm values for stratiform precipitation cases in Mêdog are close to those for Nagqu, which is in line with the report by Thompson et al. (2015) showing that stratiform rainfall in the tropics is similar to that in other climate regimes.
Figure 8. Distribution of the mean values of lgNw and Dm from the present study and from the literature, denoted with different symbols as shown in the legend. The blue/red symbols represent stratiform/convective precipitation. The two black rectangles represent the maritime and continental convective populations, respectively, from Bringi et al. (2003). The dotted and solid lines indicate the C–S separation lines from Bringi et al. (2003) for continental regions and from Thompson et al. (2015) for the tropics, respectively.
The results were also compared with other climate regimes in China (i.e., Nanjing in East China, Beijing in North China, and Foshan in South China, as reported by Wen et al., 2016, Ji et al., 2019, and Wang, 2019, respectively). For convective rain, Mêdog exhibited characteristics close to the two-dimensional video disdrometer observations of East and South China during the summer monsoon period, where the observed summer convective clusters are also maritime-like in nature. This may be due to the abundant warm and humid moisture in summer in the three regions, which produces large quantities of small particles. Nagqu consists of a lower concentration of relatively larger-sized drops, similar to that seen in North China, where the mean values of Dm and lgNw are 2.03 mm and 3.61, respectively (Ji et al., 2019). The DSD characteristics in Nagqu and Beijing appear to be evidence of the microphysics of precipitation in the midlatitudes, where ice processes likely play an important role in precipitation processes (Dolan et al., 2018; Ji et al., 2019).
Previous studies have revealed that the μ-Λ relation has the ability to represent variability in DSDs of natural precipitation well, and can be approximately described by a second-degree polynomial (Zhang et al., 2003; Chen et al., 2017; Wu et al., 2019). The relation varies with climatological regimes, geographical locations, and precipitation types (Zhang et al., 2003; Cao et al., 2008; Chen et al., 2013, 2016).
Following the method of Zhang et al. (2003), to minimize the scatter, samples with drop counts > 1000 and rainfall rates R > 5 mm h−1 were used to compute the μ and Λ values in Mêdog and Nagqu. The μ-Λ relation can be used for the range of Λ between 0–20 mm−1, and larger values of Λ indicate smaller raindrops (Zhang et al., 2003; Cao et al., 2008). Then, a second-degree polynomial μ-Λ relation was further fitted by the least squares method based on these data (Fig. 9). The relation for Mêdog is given below.
Figure 9. Scatterplots of μ versus Λ and the empirical fitting relations for cases with rainfall rates >5 mm h−1 and raindrop counts > 1000 in Mêdog and Nagqu on the TP. The black dots represent Mêdog rainfall cases, and the gray dots represent Nagqu precipitation clusters. The red solid line and blue solid line indicate the fitted empirical μ-Λ relations in Mêdog and Nagqu, respectively. The green dashed line represents the empirical μ-Λ relation in Florida, obtained from Zhang et al. (2003).
The relation for Nagqu is as follows.
Figure 9 shows that the shape factor μ of Mêdog is close to that of Nagqu when Λ < 10 mm−1, while it is obviously lower than that of Nagqu when Λ is increasing. This could be related to higher numbers of small drops in Mêdog than in Nagqu.
Comparing the μ-Λ relations of Mêdog and Nagqu with that determined in Florida, USA, as derived by Zhang et al. (2003), the shape factor μ of Florida is distinctly smaller than those of Mêdog and Nagqu with increasing Λ (i.e., Λ ≥ 10 mm−1). On the one hand, these differences could be attributed partly to the different types of instruments used. The 1D PARSIVEL disdrometer used in our study tends to underestimate the numbers of small and midsize drops compared to the 2D video disdrometers used in the study in Florida, leading to larger μ values found in our study (Zhang et al., 2003; Wen et al., 2016). On the other hand, although a 1D PARSIVEL disdrometer was used in Mêdog and Nagqu, the μ-Λ relation obtained in Mêdog is slightly different from that determined in Nagqu, which implies that the microphysics of precipitation vary with geographical locations and climate regimes.
The major uncertainty in radar-based QPEs is caused by DSD variability, which can be affected by climate regimes, rainfall types, and geographical locations (Tokay and Short, 1996; Uijlenhoet, 2001; Rosenfeld and Ulbrich, 2003; Steiner et al., 2004; Lee and Zawadzki, 2005; Tokay et al., 2008). These DSD variabilities fundamentally affect the radar reﬂectivity factor (Z) and rainfall rate (R) relation, which is widely used in radar QPE algorithms. For example, Tokay and Short (1996) recommended the use of the relations Z = 367R1.30 and Z = 139R1.43 for stratiform and convective rainfall types in tropical regions, respectively. The Next-Generation Weather Radar (NEXRAD) system recommends the empirical relationships of Z = 300R1.4 and Z = 200R1.6 for convective and stratiform precipitation in the midlatitudes, respectively (Fulton et al., 1998). To improve radar rainfall estimates over the TP, the Z-R relations in Mêdog and Nagqu are discussed in this section based on the DSD characteristics observed during the rainy season.
Scatterplots displaying the relation between the Z and R are given in Fig. 10, superimposed with the Z-R fittings based on the least squares method for stratiform and convective precipitation cases in the two studied regions. Details of the fitted coefficients and exponents of the power-law relations for different precipitation types in the two regions are given in Table 1. For comparison with previous studies, the Z-R relations suggested in the midlatitudes and tropics are also superimposed upon Fig. 10 with differently colored solid lines. The corresponding power-law relationships are also given with the same colors as those of the solid lines. Following Wu et al., (2019), statistical parameters such as the normalized mean bias (NB) and normalized standard error (NSE) were used in this study to evaluate the performances of different Z-R relations.
Figure 10. Scatterplots of radar reflectivity (Z) versus rainfall rate (R), superimposed with the fitting curves for (a) stratiform rain and (b) convective rain. The black dots represent Mêdog rainfall cases, and the gray dots represent Nagqu precipitation clusters. The red and blue solid lines represent the fitting Z-R relations in Mêdog and Nagqu, respectively. The green solid lines denote the empirical relations used in NEXRAD. The black solid line represents the fitting relations from Tokay and Short (1996).
Region Stratiform rain Convective rain A b A b Mêdog 114.79 1.34 53.69 1.71 Nagqu 124.95 1.32 89.55 1.79
Table 1. Fitted radar reflectivity and rain rate (Z-R) relations for stratiform and convective rain types in Mêdog and Nagqu on the Tibetan Plateau (TP).
The evaluation results are given in Table 2. In terms of stratiform rain, the fitted Z-R relation in Mêdog is close to that in Nagqu (Fig. 10a), which is fundamentally determined by the similar DSDs between the two regions (e.g., Figs. 7a, c, and 8). Table 2 shows that the minimum NB and NSE values in the two regions both come from the fitted Z-R relation, with NB values of 11.8% and 13.6% for Mêdog and Nagqu, respectively and NSE values of 24.8% and 28.3% for Mêdog and Nagqu, respectively. The empirical relationship at midlatitudes of Z = 200R1.6 underestimated stratiform rain by 25.5% and 26.6% on average in Mêdog and Nagqu, respectively. In particular, stratiform precipitation was overestimated (underestimated) by approximately 16% and 22% (34% and 36%) in Mêdog and Nagqu, respectively, when the rain rate was below (above) 0.1 mm h−1. Furthermore, the relation Z = 367R1.30, suggested for use in the tropics (Tokay and Short, 1996), seriously underestimated stratiform precipitation, exceeding 50% in the two regions.
Parameter Region Convective rain Stratiform rain Z=300R1.40 Z=139R1.43 Fitted Z-R Z=200R1.60 Z=367R1.30 Fitted Z-R NB Mêdog −45.7 −10.9 3.8 −25.5 −53.7 11.8 Nagqu 5.8 67.9 4.8 −26.6 −49.8 13.6 NSE Mêdog 48.6 27.6 20.6 31.9 54.6 24.8 Nagqu 40.3 77.1 31.3 36.0 52.8 28.3
Table 2. NB and NSE (%) values of empirical relations for convective/stratiform rain in the midlatitudes (Z=300R1.40/Z=200R1.60) and tropics (Z=139R1.43/Z=367R1.30) and of the fitted Z-R relation in this work for convective and stratiform precipitation types in Mêdog and Nagqu on the TP.
Comparing convective precipitation between the two regions, the fitted power-law relationships are Z = 53.69R1.71 and Z = 89.55R1.79 in Mêdog and Nagqu, respectively. Mêdog convective rain has smaller values for both the coefficient A and exponent b, likely associated with the relatively high concentration of small-sized raindrops in Mêdog. In other words, the same Z would derive a higher R in Mêdog, compared to Nagqu. In contrast with the Z-R empirical relations of Z = 300R1.4 and Z =139R1.43, the fitted Z-R relation of Mêdog is close to Z = 139R1.43, which was suggested for the tropics, and the Z-R relation of Nagqu approaches Z = 300R1.4, which was recommended for the midlatitudes. The minimum NB and NSE values for convective rain in Mêdog and Nagqu are also obtained from the fitted Z-R relation, with NB values of 3.8% and 4.8% and NSE values of 20.6% and 31.3%, respectively. In addition, the use of the term Z = 300R1.4 underestimates (overestimates) Nagqu convective precipitation by approximately 15% (20%) when the rainfall rate is below (above) 20 mm h−1. The use of the term Z = 139R1.43 also underestimates (overestimates) Mêdog convective precipitation by approximately 17% (5%) Mêdogwhen the rainfall rate is below (above) 30 mm h−1.