Atmospheric Research 119 (2013) 131–139
Contents lists available at SciVerse ScienceDirect
Atmospheric Research journal homepage: www.elsevier.com/locate/atmos
Evaluation of the warm season diurnal cycle of precipitation over Sweden simulated by the Rossby Centre regional climate model RCA3 Alexander Walther a,⁎, Jee-Hoon Jeong a, Grigory Nikulin b, Colin Jones b, Deliang Chen a, 1 a b
Department of Earth Sciences, University of Gothenburg, Gothenburg, Sweden Rossby Centre, SMHI, Norrköping, Sweden
a r t i c l e
i n f o
Article history: Received 20 January 2011 Received in revised form 19 October 2011 Accepted 21 October 2011 Keywords: Precipitation diurnal cycle Regional climate model Resolution dependency Convective precipitation
a b s t r a c t This study examines the diurnal cycle of precipitation over Sweden for the warm season (April to September) both in hourly observational data and in simulations from the Rossby Centre regional climate model (RCA3). A series of parallel long-term simulations of RCA3 with different horizontal resolutions – 50, 25, 12, and 6 km – were analyzed to investigate the sensitivity of the model's horizontal resolution to the simulated diurnal cycle of precipitation. Overall, a clear distinction between an afternoon peak for inland stations and an early morning peak for stations along the Eastern coast is commonly found both in observation and model results. However, the diurnal cycle estimated from the model simulations show too early afternoon peaks with too large amplitude compared to the observation. Increasing horizontal model resolution tends to reduce this bias both in peak timing and amplitude, but this resolution effect seems not to be monotonic; this is clearly seen only when comparing coarser resolution results with the 6 km resolution result. As the resolution increases, the peak timing and amplitude of the diurnal cycle of resolved large-scale precipitation become more similar to the observed cycle of total precipitation while the contribution of subgrid scale convective precipitation to the total precipitation decreases. An increase in resolution also tends to reduce too much precipitation of relatively light intensity over inland compared to the observation, which may also contribute to the more realistic simulation of the afternoon peak in convective precipitation. © 2011 Elsevier B.V. All rights reserved.
1. Introduction The diurnal cycle of precipitation is one of the fundamental natural variations found in the earth's climate system. The punctual diurnal variation of incoming solar radiation associated with diurnal changes in other weather and climate variables like temperature and vertical motion yield the diurnal cycle of precipitation. Extensive studies based on in-situ and satellite observation have addressed various dynamical mechanisms to cause and modulate diurnal variation of precipitation (e.g. Brier and Simpson, 1969; Dai and Deser, 1999; Dai and Wang, 1999; Wallace, 1975), but still much uncertainty remains.
⁎ Corresponding author at: Department of Earth Sciences, University of Gothenburg, Box 460, 405 30 Gothenburg, Sweden. Tel.: + 46 31 7862867. E-mail address:
[email protected] (A. Walther). 1 Currently on leave. 0169-8095/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.atmosres.2011.10.012
Overall, the nighttime peak in maritime climate vs. the late afternoon peak in continental climate are the most pronounced features of the diurnal cycle of precipitation found in the climate system (Dai, 2001a, 2001b), although detailed regional representations vary greatly depending on the season and topographical conditions (e.g. Lee et al., 2007; Yin et al., 2009). Despite of its fundamentality in weather and climate, global and regional climate models still have large difficulties in reproducing the observed diurnal cycle of precipitation, especially those found on local and regional scale (Guichard et al., 2004; Lenderink and van Meijgaard, 2008; Michaelides et al., 2009; Rio et al., 2009; Shin et al., 2007; Zhang et al., 2008). One of the most common biases in climate models is too early initiation of the afternoon peak in the continental area, which is known to attribute considerably to difficulties in simulating realistic convective activities occurring on sub-
A. Walther et al. / Atmospheric Research 119 (2013) 131–139
grid scales. For instance, too much sensitivity in the convective formulation in the current models initiate too frequent convective activities of relatively light intensity in the early afternoon which prevent the accumulation of available energy to invoke stronger convection in the later afternoon (e.g. Lee et al., 2008; Rio et al., 2009; Shin et al., 2007). Previous studies with climate models suggested that this bias is sensitive to the choice of the models' convective or cumulus parameterization (e.g. Liang, 2004; Rio et al., 2009) as different schemes show different diurnal cycle characteristics depending on the geographical locations. Another crucial factor to determine the characteristics of a realistic diurnal cycle is the model's spatial resolution, which needs to be sufficiently fine to resolve sub-grid scale processes associated with convective activities, and it has been generally suggested that a finer horizontal resolution simulates the diurnal cycle and other precipitation parameters more realistically (Lee et al., 2007; Rauscher et al., 2010; Shin et al., 2007; Wehner et al., 2010). The purpose of this study is to investigate the general performance of the Rossby Centre regional climate model (RCA3) to simulate the observed warm season diurnal cycle of precipitation over Sweden. In our recent work (Jeong et al., 2011), the overall features of the diurnal precipitation cycle in Sweden were estimated both in hourly observational data, and an initial comparison with the simulated diurnal cycle by the Rossby Centre regional climate model (RCA3) with 50 km horizontal resolution was made. Here we utilized a parallel series of RCA3 simulations for the same period but with finer horizontal resolution: 25, 12, and 6 km, in order to investigate the sensitivity of the horizontal resolution on the simulated characteristics of the diurnal cycle. The contribution of convective precipitation vs. precipitation from large-scale condensation to the total precipitation amount and its diurnal cycle was investigated. The following Section 2 describes the hourly observational data of precipitation in Sweden and the RCM simulations, and the method used to estimate the peak and amplitude of the diurnal cycle. In Section 3 the general characteristics of the diurnal precipitation cycle in the observations and simulations are presented, followed by discussions in Section 4. 2. Data and methods 2.1. Observed hourly precipitation data The observed hourly precipitation records from 93 stations in Sweden for 1996–2008 provided by the Swedish Meteorological and Hydrological Institute (SMHI) are utilized. Measurements at each station are conducted by wind-shielded GEONOR automatic precipitation gauges (Bakkehoi et al., 1985) which yield hourly precipitation rates with a resolution of 0.1 mm/h. The precipitation designated to a certain hour is the amount accumulated during the past 60 min. The stations are relatively well distributed over the entire Sweden, covering most of the plains in southern and middle Sweden, the northern mountainous region, as well as the coastal region along the Baltic Sea (Fig. 1a). Following the guidelines of the SMHI data distribution center, initial quality control was conducted to remove erroneous negative values. Some additional quality control were conducted by following Jeong et al. (2011) in order to remove unrealistically high values. All these erroneous values were set to missing records, but occupy only a very small fraction (0.0004%) of
a)
b) MALEXANDER, AMJJAS 0.22
Obs 50km 25km 12km 6km
0.2
0.18
amount [mm] (solid)
132
0.16
0.14
0.12
0.1
0.08
0.06
0
2
4
6
8
10
12
14
16
18
20
22
time [hour] Fig. 1. a) Geographical locations of observation stations of the hourly precipitation network over Sweden. The fraction of missing values in the original station series is indicated by symbols. b) An example of the estimated diurnal cycle of precipitation amount from observation and RCA3 simulations with 4 different resolutions for the ‘Malexander’ station (marked with a black circle in the map).
total records. The research area, 55–70 N/10–25E, is located in the zone of boreal and sub-arctic climates with maritime character along the west coast and the coast to the Baltic Sea and more continental influence in the Swedish inland
A. Walther et al. / Atmospheric Research 119 (2013) 131–139
on the lee-side of the Scandinavian mountains. There is yearround precipitation peaking summer to autumn with a big amount originating from thunderstorms. The highest official daily total rainfall recorded is 198 mm (1997, Northern Sweden). Daily amounts above 20 mm are considered as heavy rain. According to Dahlström (2006), the maximum 10 min (1hour) rainfall recorded is around 30 mm (110 mm). 2.2. RCA3 experiments We evaluated the characteristics of the diurnal precipitation cycle simulated with a regional climate model by comparing those with observational records. Long-term simulations of the Rossby Centre Atmospheric Model version 3 (RCA3) were analyzed. The RCA3 is known to be one of the best models in simulating weather and climate over Sweden and Scandinavia (Kjellström et al., 2005) and has been extensively utilized for numerous climate impact studies (Ghatak et al., 2010). We utilized RCA3 experiments conducted with 4 different spatial resolutions: approximately 50, 25, 12, and 6 km with 24 unequally spaced hybrid terrain-following vertical levels. Originally, all experiments were run from 1961-present. Only the 6 km experiment covers a shorter period — 1987present. For the comparison only the results for the overlapping period with the hourly observations, 1996–2008, was used. From the model experiments, total precipitation (1-hourly), large-scale and convective precipitation (3-hourly) was used. All simulations were driven with data from the ERA40 re-analysis (Uppala et al., 2005) until 8/2002 and ERAInterim (Dee et al., 2011) thereafter. Identical model physics and dynamics were utilized. The basic topography and land use type datasets used are at 1 km resolution and were converted to each of the designated resolutions. The only difference between the simulations is the computation time-step interval (30 min, 20 min, 10 min, and 7.5 min for 50 km, 25 km, 12 km, and 6 km simulation respectively) in order to meet computational stability for the chosen horizontal resolution. The model domain covers most of Europe using a rotated latitudelongitude coordinate system; the latitude range is about 26–72°N, and the corresponding longitude range is about 4°W–34°E and 33°W–57°E at the southern and northern boundary, respectively. The land-surface scheme used is a so-called tiled scheme (Samuelsson et al., 2006) meaning that each grid cell consists of several tiles with individual surface properties and for which a separate energy balance is calculated. The Kain–Fritsch-scheme was used for the parameterization of convective clouds at subgrid scale. Resolved large-scale clouds are described using the scheme of Rasch and Kristjánsson. For further details on the dynamical and physical treatments of RCA3 the reader is referred to Samuelsson et al. (2006). When comparing the simulation results with observations, the simulated precipitation data were taken from grid points closest to the 93 observational stations without any spatial interpolation process in order to preserve the rather localized characteristics of precipitation events. 2.3. Determination of the diurnal cycle of precipitation The diurnal cycle of precipitation was objectively estimated by using long-term observational records and corresponding
133
results from RCA3 simulations. Here we followed the methodology described in Jeong et al. (2011). Firstly, a long-term average of the hourly precipitation amount P– at each station (or corresponding grid points for model simulations) for the entire analysis period was calculated. Secondly, applying the harmonic analysis technique (Angelis and McGregor, 2004; Wilks, 2006) the diurnal variation of a chosen station (Pe ) at hour h was determined by the summation of sinusoidal harmonics as − 2πkh PeðhÞ ¼ P þ ∑ Ck cos −θk þ residual ð1Þ 24 k −
−
where P is the 24-hour mean of P , k is the harmonic number (i.e., 1 for the 24-hour cycle), and Ck and θk are the amplitude and the phase of the given kth harmonic. In this study, the summation of the 1st (24-hour frequency; k = 1) and 2nd (12-hour frequency; k = 2) part of the harmonics was defined as mean (smoothed) diurnal cycle. The amplitude of this smoothed diurnal cycle was determined as half the difference between maximum and minimum value within the average 24-hour cycle, and the peak timing as the time when the maximum value is found. An example of the diurnal precipitation cycle observed and modeled is shown in Fig. 1b. The diurnal cycle modeled by the 1st and 2nd harmonics well represent both, amplitude and peak phase of the diurnal cycle obtained from observed raw data (Jeong et al., 2011). 3. Results The observed diurnal cycle for the summer season over Sweden shows clear temporal and spatial characteristics (Fig. 2a). A pronounced rainfall peak in the afternoon is found for most inland stations all over Sweden, both in the Southern plains and in the Northern mountains. The rainfall amounts peak largely between 15 and 18 LST. In contrast, stations along the Eastern coast show a night or early morning peak around 02–06 LST. Jeong et al. (2011) suggested that convective precipitation activity is the main cause for the inland afternoon peak while the night or early morning peak for stations along the eastern coast is linked to non-convective rainfall events emerging from frontal systems within the land-sea breeze circulation. It appears that all model simulations relatively well capture the principal spatial pattern of the diurnal cycle in terms of the amplitude and peak timing, but there is a more or less pronounced deficiency. The amplitude is much overestimated in the 50 km model especially for the mid and southern plains, and gets closer to the observed as the model resolution increases to 25, 12 and 6 km. For the peak timing, the 50 km simulation exhibits the largest bias; most of the inland stations produce a too early afternoon peak showing quite uniform peak timing around noon, which is earlier than observed by up to 5 h. It appears that this bias becomes smaller as the horizontal resolution increases. However, this effect is the most distinct when the resolution comes to 6 km while there is only little improvement from the 50 via the 25 to the 12 km simulation. This is more clearly seen in Fig. 3 where the bias between observed and modeled peak timing is presented. For the vast majority of stations the peak timing is simulated too early (downward triangle). Light red to orange colors, i.e. bias of up to 4 h, are dominating the maps showing the bias of the 50, 25 and 12 km simulations. There is a slight trend toward smaller
134
A. Walther et al. / Atmospheric Research 119 (2013) 131–139
69°N
69°N
Obs AMJJAS
69°N
50km AMJJAS
69°N
25km AMJJAS
69°N
12km AMJJAS
0.12
6km AMJJAS
0.1 66°N
66°N
66°N
66°N
66°N
63°N
63°N
63°N
63°N
63°N
0.08
0.06 60°N
60°N
60°N
60°N
60°N
57°N
57°N
57°N
57°N
57°N
54°N
12°E 15°E 18°E 12°E 24°E
54°N
12°E 15°E 18°E 12°E 24°E
54°N
12°E 15°E 18°E 12°E 24°E
54°N
12°E 15°E 18°E 12°E 24°E
54°N
0.04
0.02
12°E 15°E 18°E 12°E 24°E
Fig. 2. Observed and modeled peak timing of the diurnal cycle of precipitation amount (from left to right). The peak timing at each station is indicated by the filling color of the circle. The strength of the diurnal cycle is indicated by gray shading [mm].
bias in the 12 km simulation. However, as mentioned before, a significant improvement in performance can be seen in the 6 km simulation. Table 1 shows the bias between observed and modeled peak timing as fraction of stations in [%] for different intervals. For the 50, 25 and 12 km simulations, the largest fraction is found in the bias range −4 and −2 h, but in the range between −1 h and +1 h for 6 km resolution (49.46%). For 15% of the locations the peak timing modeled matches the peak observed with no bias in the 6 km run. The numbers are
69°N
AMJJAS
AMJJAS
AMJJAS
12
69°N
69°N
69°N
AMJJAS
5.4%/9.7%/9.7% for the 50/25/12 km runs, respectively which clearly underlines the better performance of the higher resolutions, and of the 6 km resolution in particular. Jeong et al. (2011) found that the RCA3 model with 50 km resolution tends to overestimate the occurrence frequency of relatively light precipitation over inland, which might contribute the bias of the simulated diurnal cycle. This can also be seen in Fig. 4 where the observed versus simulated precipitation frequency is shown for the three regional domains: Northern
10 8
66°N 66°N
66°N
6
66°N
4 2
63°N 63°N
63°N
63°N
0 -2
60°N 60°N
60°N
-4
60°N
-6 57°N
-8
57°N
57°N
57°N
-10 54°N
12°E
15°E
18°E
21°E
12km−Obs
25km−Obs
50km−Obs 24°E
54°N
12°E
15°E
18°E
21°E
24°E
54°N
12°E
15°E
18°E
21°E
6km−Obs 24°E
54°N
12°E
15°E
18°E
21°E
-12 24°E
Fig. 3. Difference between simulated and observed peak timing of precipitation amount. Deviation in hours indicated with a color scale. Black circle (downward triangle, square) indicate the model simulating the peak at the same hour (earlier, later) compared to the observation.
A. Walther et al. / Atmospheric Research 119 (2013) 131–139
are shown in Fig. 5. Again, a clear distinction between an afternoon peak for inland vs. an early morning peak in Eastern coast, as well as the model bias is seen. For the inland domain, the convective precipitation tends to decrease almost monotonically with increasing resolution, whereas the large-scale part increases: among the simulations the 6 km run has the highest large-scale and the least convective precipitation. For the inland stations in the Northern and Southern domain the convective precipitation cycle shows a distinct afternoon peak with the peak timing slightly shifting later, that is closer to the observed one, with increasing model resolution, of which amplitude and timing is getting to be more similar to those from total precipitation observed. The amplitude of the large-scale precipitation cycle is much weaker than that of convective precipitation and the peak is found in the early morning for the 50, 25 and 12 km simulations. In contrast, the 6 km simulation exhibits an afternoon peak, even though with small amplitude, which occurs slightly after the peak timing of the observed diurnal cycle of total precipitation. Summarizing these results, we may conclude that the least bias of peak timing and amplitude of the diurnal cycle in the 6 km simulation is contributed by a combination of the more realistic simulation of the convective precipitation cycle, and partly by the late afternoon peak in large-scale condensational precipitation. For the diurnal cycle in the East coast, all simulations roughly capture the early morning peak, which seems to be mostly contributed by large-scale precipitation. However, the effect of the increasing resolution is not obviously found in the East coast. The amplitude and peak timing in the 50 km and 25 km simulations are more similar to observations than the 6 km experiment. The peak timing in the early morning peak along the Eastern coast is more contributed by the non-convective precipitation processes linked to the coastal environment along the Baltic Sea, thus increased resolution or improved simulation of convective precipitation does not have effective significant effect on the performance of the simulation. All 4 simulations largely overestimate the precipitation amount observed. Although the peak time and amplitude of the simulated diurnal cycle is much improved in the 6 km resolution, the model still overestimates the total amount of precipitation by about 30% compared to the observation
Table 1 Fraction of stations [%] where the bias of the simulated peak timing with respect to the peak timing observed falls into the designated range. Example: in the 25 km simulation for 11.83% of the stations the diurnal peak is modeled 5 to 7 h earlier than observation.
50 km 25 km 12 km 6 km
− 7 to − 5 h
− 4 to − 2 h
− 1–1 h
5.38 11.83 5.38 2.15
58.06 59.14 50.54 32.26
25.81 18.28 25.81 49.46
mountains (north of 60°N), Southern plains (south of 60°N), and Eastern coast (within 20 km from the shore, east of 15°E) to examine the regional difference. First, it is found that there is a clear difference between the two inland domains and the Eastern coast. At the Eastern coast stations the model bias is relatively smaller and there is no clear difference in performance for different periods during the day, again confirming that the diurnal cycle along the Baltic Sea coast is present, but rather weak. All simulations largely overestimate the precipitation frequency compared to observation with a factor of 3–5. Model resolution increase is coupled to a decrease in precipitation frequency coming closer to the observed values. This is most pronounced in the Northern mountains having the most complicated terrain, where the mean frequency is around 60% in the 50 km simulation decreasing to 46% in the 6 km run. When examining the observed vs modeled precipitation amount (not shown here) it becomes obvious that too high precipitation amounts originate from the run with the coarsest resolution (50 km) around noon (12 LST). The inter-model difference of the mean amount is not as large as for the modeled mean precipitation frequency as shown in Fig. 4. Based on the identified bias of the simulated diurnal cycle depending on the horizontal resolution, we further investigated the contribution from different precipitation processes – largescale condensational and convective precipitation – to the bias in the diurnal cycle separately. The diurnal cycle estimated from the total, convective and large-scale precipitation from the 4 different RCA3 simulations averaged for the tree regions 80
80
80
EastCoast
75
70
70
70
65
65
65
60 55 50 45 40
5
10
15
freq obs [%]
60 55 50 45 40
50km 25km 12km 6km
35
freq modeled [%]
75
freq modeled [%]
75
30
24
Southern plains
Northern mountains
freq modeled [%]
135
20
30
5
10
15
freq obs [%]
20 18 16
60
14
55
12
50
10 8
45 40
50km 25km 12km 6km
35
22
35 20
30
6
50km 25km 12km 6km
5
10
15
4 2 20
freq obs [%]
Fig. 4. Scatter plot of observed versus simulated precipitation amount from the estimated diurnal cycle of precipitation for three representative regions. Each dot represents the observed rainfall amount at one station at an hour of the diurnal cycle (x-axis) versus those for the respective RCA3 resolutions (y-axis). The corresponding time of the day is indicated by colors and the respective model resolution by symbol. The mean for each model simulation is shown as black symbol.
136
A. Walther et al. / Atmospheric Research 119 (2013) 131–139 0.18
0.18
South, AMJJAS
0.16
0.16
0.14
0.14
0.12
0.12
amount [mm]
amount [mm]
North, AMJJAS
0.1 0.08 0.06
0.08 0.06
0.04
0.04
0.02
0.02
0
0
2
4
6
8
10 12 14 16 18 20 22
time [hour] 0.18
EastCoast, AMJJAS 0.16 0.14
amount [mm]
0.1
0.12 0.1 0.08 0.06 0.04
0
0
2
4
6
10 12 14 16 18 20 22
8
time [hour] Obs 50km 25km 12km 6km Conv 50km Conv 25km Conv 12km Conv 6km LS 50km LS 25km LS 12km LS 6km
0.02 0
0
2
4
6
8
10 12 14 16 18 20 22
time [hour] Fig. 5. Diurnal cycle of precipitation amount for 3 representative regions: observed total precipitation (bold blue line), simulated total precipitation (bold solid lines), simulated convective precipitation (thin solid lines) and simulated large-scale precipitation (dotted lines).
(see lines with total precipitation in Fig. 3). This could be contributed by the too sensitive and frequent initiation of convective precipitation events as suggested by previous model studies (Lee et al., 2007; Sato et al., 2009; Shin et al., 2007). Focusing on the afternoon peak, we compare the observed and simulated total, convective and large-scale precipitation frequency for 12–18 LST in Fig. 6. All simulations overestimate the frequency of total precipitation by about 30 to 60% compared to observations with a frequency range between 6 and 16%. The minimum frequency in the simulations is around 35% with a maximum of about 75% (50 km), 65% (25 km), 55% (12 km) and 50% (6 km). For the RCA3 simulations, the most pronounced pattern is too frequent precipitation events in the Scandinavian mountain range having the most complicated topography within the study area, while the observations provide a spatially diverse pattern with more regions of frequent precipitation events peaking in southern and central Sweden. As the horizontal resolution increases, this bias continuously decreases. Recalling the decreasing amount of convective precipitation with increasing horizontal resolution shown in Fig. 5, the reduced amount of precipitation in finer resolution seems to be mostly contributed by the reduced convective precipitation, while showing more
similar amplitude and peak timing of the diurnal cycle to those from observed total precipitation. Too high precipitation frequency over the mountains is also evident in the convective and large-scale simulations (Fig. 6). Resolution increase also here leads to frequency decrease over this area. Outside the core mountain area large-scale precipitation frequency increases whereas convective precipitation frequency decreases with increasing model resolution. The observed spatial frequency pattern (bottom left) in Southern and Central Sweden is rather diverse and the spatial variability is best reproduced in the simulated 12 and 6 km convective precipitation. 4. Summary and discussion The amplitude and peak timing of the diurnal cycle in Sweden were objectively determined from hourly observations and RCA3 simulations with 4 different horizontal resolutions. The simulations capture the principal characteristics of the observed diurnal cycle quite well but, in general, show a too early afternoon peak in large parts of the inland regions being mostly contributed by convective activity which is found in late afternoon (15–18 LST) in the observational data. The finer the model's horizontal resolution, the more
A. Walther et al. / Atmospheric Research 119 (2013) 131–139
50km p−tot
12km p−tot
25km p−tot
69°N
137
6km p−tot
69°N
69°N
75
69°N
[%] 65 66°N
66°N
66°N
66°N
60 63°N
63°N
63°N
63°N
60°N
60°N
60°N
60°N
57°N
57°N
57°N
57°N
54°N
54°N
54°N
54°N
55 50 45 40 35
12°E
18°E
24°E
12°E
50km p−l s 69°N
18°E
24°E
12°E
18°E
24°E
18°E
24°E
6km p−l s 75
69°N
69°N
69°N
12°E
12km p−l s
25km p−l s
[%] 65 66°N
66°N
66°N
66°N
60 63°N
63°N
63°N
63°N
60°N
60°N
60°N
60°N
57°N
57°N
57°N
57°N
55 50 45 40 35
54°N
OBS prec
12°E
18°E
24°E
54°N
50km p−conv
69°N
22
12°E
18°E
24°E
54°N
25km p−conv
12°E
18°E
24°E
54°N
12km p−conv
12°E
18°E
24°E
6km p−conv
69°N
69°N
69°N
69°N
66°N
66°N
66°N
66°N
22
[%] 18
66°N
[%] 18
16
16
14 63°N
12
14 63°N
63°N
63°N
63°N
60°N
60°N
60°N
60°N
57°N
57°N
57°N
57°N
12
10 60°N
8
10 8
6 57°N
4
6
2 54°N
12°E
18°E
24°E
54°N
12°E
18°E
24°E
54°N
12°E
18°E
24°E
54°N
12°E
18°E
24°E
54°N
4 2 12°E
18°E
24°E
Fig. 6. Frequency of precipitation events (prec > 0 mm) for LST 12–18 obtained from the observations (bottom left) and from the RCA3 simulated total precipitation (p-tot, top panel), large-scale precipitation (p-ls, middle panel) and convective precipitation (p-conv, bottom panel) for all model resolutions (50-6 km from left to right) in [%]. Note the different color scales for obs/p-conv and p-tot/p-ls.
realistically simulated the peak timing and amplitude of the diurnal cycle compared to the observations, and the less of the simulated precipitation coming from the parameterized convective precipitation on the sub-grid scale. The fraction of convective/large-scale precipitation gradually decreases/ increases with finer grid resolution and the afternoon peak in the convective rainfall curve is shifting later towards the rainfall peak observed. However, in the 6 km simulation this peak is still 1–2 h earlier than the afternoon peak observed. There is an abrupt change in the shape of the diurnal cycle of convective/large-scale precipitation from 12 to 6 km. The
more realistic simulation of the afternoon peak in the 6 km model is mostly contributed by the diurnal cycle of resolved large-scale precipitation showing an inland afternoon peak one hour ahead the observed one, which is dominant over the diurnal cycle of subgrid scale convective rainfall at the finest resolution. An improved convective parameterization scheme could significantly improve the performance of the diurnal cycle simulations. The scheme used for convective parameterization in RCA3 – the Kain–Fritsch convection scheme – was enhanced compared to the previous model version (RCA2) in the way
138
A. Walther et al. / Atmospheric Research 119 (2013) 131–139
that shallow convection is assumed to be non-precipitating. Nevertheless, the scheme seems to be still too sensitive when it comes to the initiation of deep convection, which does not change significantly with an increase in resolution regarding the peak timing coinciding with findings obtained by Lee et al. (2007). The relative influence of convective and large-scale precipitation on the simulated diurnal cycle was suggested, but it remains uncertain and not easy to validate because the distinction between the convective and large-scale precipitation in the observations is almost impossible. Depending on the model resolution, convective and large-scale precipitation contributes differently to the diurnal cycle of total precipitation. The rather abrupt improvement from the 12 km to the 6 km simulation may indicate a direct coupling between these two precipitation types. Convective processes in the 6 km simulation may act as a moisture supply triggering increased largescale precipitation with some hours delay instead of producing precipitation from the convective scheme. This question cannot be answered satisfactorily with the data available at this stage and within the scope of this study. However, results from an initial analysis suggest that convective and large-scale precipitation are not coincident and in this way not related directly. The results of this study are based on precipitation records only. Further examination of the local circulation, and energy and heating distribution associated with convective instability is required to reveal underlying physical processes in detail. For instance, our results suggest that the sensitivity of the simulated diurnal cycle change is found to be not monotonic with a considerable jump between 12 km and 6 km resolution. This may imply that a certain dynamical/physical process being resolvable with a resolution finer then 12 km may greatly contribute the realistic simulation of the diurnal cycle. The summer season 24-h cycle is the main focus of this study. However, also a semidiurnal (12-h) peak is evident during nighttime over the inland regions and afternoon along the East coast. The 24-h cycle is well dominant over the 12-h cycle and is rather weak when looking at the regional means. Dai (2001b) distinguishes between different precipitation categories ranging from drizzle to showery precipitation, where the former contributes to nighttime peaks and the latter mostly to afternoon peaks over land. The cause for the semidiurnal peak can be twofold. First, it could be a real secondary peak as suggested by Dai (2001b). Another explanation is the superposition of precipitation peaks originating from different days being mixed when calculating the average cycle. Results from an initial analysis based on the data available suggest that both causes contribute to similar extent in the study area. The time-mean total rainfall amount and the strength of the diurnal cycle seem to be related to some degree. For example, precipitation amounts along the East coast are lower and the diurnal cycle weaker compared to higher amounts and a stronger cycle for the inland regions. The performance to resolve a realistic diurnal cycle of precipitation is, of course, very sensitive to the model's physical parameterizations especially associated with parameterizations in planetary boundary layer processes and cumulus convection. Given the aforementioned role of convective and large-scale precipitation among the simulations, an explicit sensitivity test of the convective parameterization scheme could help answering the question whether an improved scheme would also lead
to a significantly better simulated total precipitation cycle. We plan to test the changes in characteristics of the precipitation diurnal cycle from the improvement of model physics in the updated version of RCA – RCA4. The comparison of the sensitivity of the horizontal resolution and that of updated physics on the precipitation diurnal cycle from two successive versions of the model will possibly be made, and could provide a good and simple metric to test the change in model performance. Also, efforts will be put on the investigation of possible future changes of the diurnal cycle under climate change conditions using regional climate model simulations forced with different GCM scenarios. Acknowledgment We thank the Swedish Meteorological and Hydrological Institute (SMHI) for providing the hourly rainfall observations and the Rossby Centre for the access to the climate simulations. The Swedish Research Council (VR), the Gothenburg Atmospheric Science Centre (GAC) and FORMAS (grant #20071048-8700∗51) are thanked for support to Deliang Chen and Alexander Walther. The Centre of Earth System Science (TELLUS) at the University of Gothenburg supported Jee-Hoon Jeong. Part of this work was conducted under the Swedish Mistra-SWECIA program. Last but not least we thank the two reviewers for very constructive and critical comments, which helped to improve this manuscript significantly. References Angelis, C.F., McGregor, G.R., 2004. Diurnal cycle of rainfall over the Brazilian Amazon. Climate Res. 26, 139–149. Bakkehoi, S., Oien, K., Forland, E.J., 1985. An automatic precipitation gauge based on vibrating-wire strain gauges. Nord. Hydrol. 16, 193–202. Brier, G.W., Simpson, J., 1969. Tropical cloudiness and rainfall related to pressure and tidal variations. Q. J. R. Meteorol. Soc. 95, 120–147. Dai, A., Deser, C., 1999. Diurnal and semidiurnal variations in global surface wind and divergence fields. J. Geophys. Res. 104, 31109–31125. Dai, A., Wang, J., 1999. Diurnal and semidiurnal tides in global surface pressure fields. J. Atmos. Sci. 56, 3874–3891. Dai, A., 2001a. Global precipitation and thunderstorm frequencies. Part I: Seasonal and Interannual Variability. J. Climate 14, 1092–1111. Dai, A., 2001b. Global precipitation and thunderstorm frequencies. Part II: Diurnal Variations. J. Climate 14, 1112–1128. Dahlström, B., 2006. Rain intensity in Sweden - a climatological analysis. VAForsk. Svensk Vatten AB, Stockholm. p. 69. Dee, D.P., et al., 2011. The ERA-Interim reanalysis: configuration and performance of the data assimilation system. Q. J. R. Meteor. Soc. 137, 553–597. Ghatak, D., Frei, A., Gong, G., Stroeve, J., Robinson, D., 2010. On the emergence of an Arctic amplification signal in terrestrial Arctic snow extent. J. Geophys. Res. 115, D24105. Guichard, F., et al., 2004. Modelling the diurnal cycle of deep precipitating convection over land with cloud-resolving models and single-column models. Q. J. R. Meteorol. Soc. 130 C, 3139–3172. Jeong, J.H., Walter, A., Nikulin, G., Chen, D., Jones, C., 2011. Diurnal Cycle of Precipitation Amount and Frequency in Sweden: Observation Versus Model Simulation, Tellus. (accepted) . Kjellström, E., et al., 2005. A 140-year simulation of European climate with the new version of the Rossby Centre regional atmospheric climate model (RCA3). SMHI Reports Meteorology and Climatology, No. 108. p. 54. Lee, M.I., et al., 2007. Sensitivity of horizontal resolution in the AGCM simulations of Warm season diurnal cycle of precipitation over the United States and Northern Mexico. J. Climate 20, 1862–1881. Lee, M.I., et al., 2008. Role of convection triggers in the simulation of the diurnal cycle of precipitation over the United States Great Plains in a general circulation model. J. Geophys. Res. D: Atmospheres 113. Lenderink, G., van Meijgaard, E., 2008. Increase in hourly precipitation extremes beyond expectations from temperature changes. Nat. Geosci. 1, 511–514.
A. Walther et al. / Atmospheric Research 119 (2013) 131–139 Liang, X.-Z., 2004. Regional climate model simulation of summer precipitation diurnal cycle over the United States. Geophys. Res. Lett. 31. Michaelides, S., et al., 2009. Precipitation: measurement, remote sensing, climatology and modeling. Atmos. Res. 94, 512–533. Rauscher, S.A., Coppola, E., Piani, C., Giorgi, F., 2010. Resolution effects on regional climate model simulations of seasonal precipitation over Europe. Clim. Dyn. 35, 685–711. Rio, C., Hourdin, F., Grandpeix, J.Y., Lafore, J.P., 2009. Shifting the diurnal cycle of parameterized deep convection over land. Geophys. Res. Lett. 36. Samuelsson, P., Gollvik, S., Ullerstig, A., 2006. The Land-surface Scheme of the Rossby Centre Regional Atmospheric Climate Model (RCA3). SMHI Meteorologi, SMHI, Norrköping. p. 25. Sato, T., Miura, H., Satoh, M., Takayabu, Y.N., Wang, Y., 2009. Diurnal cycle of precipitation in the tropics simulated in a global cloud-resolving model. J. Clim. 22, 4809–4826.
139
Shin, D.W., Cocke, S., LaRow, T.E., 2007. Diurnal cycle of precipitation in a climate model. J. Geophys. Res. D: Atmospheres 112. Uppala, S.M., et al., 2005. The ERA-40 re-analysis. Q. J. R. Meteorol. Soc. 131, 2961–3211. Wallace, J.M., 1975. Diurnal-variations in precipitation and thunderstorm frequency over conterminous United-States. Mon. Weather Rev. 103, 406–419. Wehner, M.F., Smith, R.L., Bala, G., Duffy, P., 2010. The effect of horizontal resolution on simulation of very extreme US precipitation events in a global atmosphere model. Clim. Dyn. 34, 241–247. Wilks, D.S., 2006. Statistical Methods in the Atmospheric Sciences, 2nd ed. Academic Press, Burlington, MA. xvii, 627 p. pp. Yin, S., Chen, D., Xie, Y., 2009. Diurnal variations of precipitation during the warm season over China. Int. J. Climatol. 29, 1154–1170. Zhang, X., Sorteberg, A., Zhang, J., Gerdes, R., Comiso, J.C., 2008. Recent radical shifts of atmospheric circulations and rapid changes in Arctic climate system. Geophys. Res. Lett. 35, L22701.