The diurnal cycle of precipitation is investigated using ground based hourly observations for more than 10 years both in Japan and Malaysia. The diurnal cycle of precipitation in Japan is classified into three clusters. The first one has a peak in the morning, and the stations categorized into this cluster are located in coastal regions. The second cluster has two peaks in the morning and in the evening. These stations are located in inland region. The morning peak in above two cluster is dominant in June, when it is `Baiu' in Japan. Baiu is the rainy season related to the south-west Asian monsoon. The third cluster is a exceptional case. No morning peak is observed in the stations of the third cluster and they have comparatively strong evening peak.In the case of Malay Peninsula, the inland region has a pronounced peak of rainfall amount at 16 LST and the magnitude excesses by 200% of the mean amount of each month. This evening peak is too sharp to represent by a sine wave of 24 hour cycle decomposed by Fourier transform. The intensity also becomes higher at the peak time(15 LST). The magnitude of the diurnal cycle of mean intensity is larger than the annual cycle of monthly mean intensity. The morning peak of precipitation is observed during the south-west monsoon season in the west coast, and during the north-east monsoon season in the east coast. The intensity of precipitation is not significantly high during the period, namely the increase of the probability or the duration of precipitation forms this morning peak. These evidences indicate the mechanism of the convective rainfall by the thermodynamic forcing in the evening, and the low level convergence between the local land-sea breeze and the predominant monsoon wind in the morning.
TRMM(Tropical Rainfall Measuring Mission) satellite is planned to be launched in 1997, and the rainfall measuring instruments, such as the first space-borne rain radar, microwave imager and visible and infrared scanner etc., will be on board (Simpson et al., 1988). TRMM satellite is designed to have the non-sun-synchronous orbit in order to prevent the estimation error caused by the large diurnal cycle of precipitation, observed at some fixed times in a day. Besides the observational error of the instruments, the error may also occur due to the averaging procedure of the intermittent observation both spatially and temporally. In these cases, it is important to know the temporal and spatial characteristics of precipitation in order to evaluate the sampling error of the observation (McConnell and North, 1987; Shin and North, 1988). On the other hand, TRMM is also expected to observe the diurnal cycle of precipitation itself in global tropics (Bell and Reid, 1993).
The knowledge of the diurnal cycle of precipitation is also important for the evaluation of the evaporation especially on daily values, because the precipitation time of a day and successive sunshine duration effects the amount of daily evaporation.
There are many reports about the diurnal cycle of precipitation in various areas in the world. Some studies used only a few station data( Ray, 1928; Prasad, 1974; Haldar et al., 1991), because the analysis of the diurnal cycle of precipitation needs relatively higher temporal resolution of precipitation data. Therefore the geographical distribution of the diurnal cycle pattern has been presented in the region with systematic observational network( Dexter, 1944; Bleeker and Andre, 1951; Wallace, 1975 ) or about the region of special observation program (Gray and Jacobson, 1977; McGarry and Reed, 1978). The diurnal variation of wind field was also analyzed in such project ( Nitta and Esbensen, 1974; Pedder, 1978; Johnson and Priegnits, 1981 ) or using general sounding data(Hastenrath, 1967), because the relationship between the diurnal solar atmospheric tide and the precipitation had been investigated( Brier, 1965; Brier and Simpson, 1969).
Weather ship records were the only data source to estimate the diurnal cycle of precipitation over the ocean (Kraus, 1963), but the various forms of satellite data are available now for the evaluation of the diurnal cycle in the atmosphere (Murakami, 1983; Augustine, 1984; Hartmann et al., 1991; Meisner and Arkin, 1987).
In this study, the diurnal cycle of precipitation is investigated using the hourly observation data of more than 10 years from the stations in both Japan and Malaysia. The seasonal change and the geographical distribution of the diurnal cycle are presented, and the mechanism of the diurnal cycle of precipitation is discussed with the diurnal cycle of the mean intensity.
In the case of Malaysian data, the hourly data of 17 stations were provided from Malaysian Meteorological Service. Data records are from 1981 to 1990 except for Tanah Rata which has only seven years from 1984 to 1990. The resolution of the observation is 0.1 mm h^{-1}.
Both data from Japan and Malaysia are uncorrected for the wind effect, and all the times used here are given in official local standard time(LST). The LST is nine hours ahead to UTC in Japan and it is eight hours in Malaysia, and these LST correspond to the latitude of 135deg. E and 120deg. E respectively. The 135deg. E line is located at the center of Japan, but the 120deg. E line runs in the east offshore of Kalimantan Island, the east Malaysia. Therefore official LST used below is more than 1 hour ahead of the local time of the sun in the Malay Peninsula.
Mean precipitation at hour h in the m^{th} month P(m,h) (mm h^{-1} month^{-1}) were calculated from data sets mentioned above.
The diurnal cycle of precipitation can be defined by
It is the normalized diurnal cycle of each month, which can be
interpreted as
Anomaly(%) of the h LST precipitation in the m^{th} month.
Estimation error(%) of monthly precipitation by the 1 hour observation only at h LST in the m^{th} month.
Annual precipitation in each LST is calculated by
Then annual mean diurnal pattern is obtained by
Mean intensity of the precipitation is estimated by
Here, N(m,h) is the mean number of precipitation occurrence
at h LST in m^{th} month.
I(m, h) is the mean rainfall only over the periods when rainfall is sensed.
The diurnal cycle of precipitation intensity can be defined similar to
the equation (1),
In the calculation of the climatological mean of the intensity for
several years, zero values were omitted from the averaging.
This algorithm was adopted considering the
mixed probability distribution(see Kedem(1990))
to describe the frequency of precipitation intensity.
Finally, the cluster analysis by maximum length method was applied to the annual mean diurnal pattern DP(y, h) of Japan. Annual mean diurnal patterns are combined in the order of nearest distance in orthogonal 24 dimensional domain. The results of this objective procedure shows the relative distance among each 24 dimensional vector and/or cluster. The distance between clusters is represented by the maximum length between the vectors belonging to each cluster. Even though the classification is objective, the decision of threshold to determine the proper number of clusters is subjective. Seven clusters were differentiated in the preliminary report(Oki and Musiake, 1992), but some of them are very similar together. The small differences among these similar clusters seem to be caused by the variation of the magnitude of the diurnal behavior, and this variation is related to the small-scale climatological conditions. The detailed description of the regional climate within Japan from the view point of the diurnal cycle of precipitation is not the objective of this JAM94, therefore only three clusters are discriminated in order to extract the basic characteristics of the diurnal cycle over Japan.
In the case of Malaysia, the diurnal cycle patterns were categorized subjectively.
The mean precipitation P(m,h), at h LST in m^{th} month, are shown in Fig. 2 by the ensemble mean of cluster I. Contour and tone in the center figure indicate the mean precipitation (mm h^{-1} month^{-1}) at that time of a day. Annual mean diurnal cycle D(y, h)(mm h^{-1} year^{-1}) and mean monthly precipitation(mm month^{-1}) are shown in the upper and right-hand side figures respectively.
The diurnal cycle patterns of precipitation DP(m, h)(%) are shown in Fig. 3 for the cluster I. Contour and tone in the center figure indicate the anomaly (%) of the diurnal cycle of precipitation, normalized in each month. Anomaly (%) for the annual mean diurnal cycle of precipitation DP(y, h) and monthly anomaly values to the annual mean are shown in the upper and right-hand side figures respectively. The diurnal cycle of precipitation in each month becomes clear in these normalized figures.
Similar figures are shown for clusters II and III in Figs. 4 to 7.
These results are similar to the earlier reports about the diurnal cycle of severe rainfall in Japan (Fujibe, 1988; Tatehira and Hoshina, 1993). Clearly the stations in cluster I have a morning peak, but there are no pronounced peaks throughout the year. It probably represents its maritime climate. The stations of cluster II have an evening peak. The precipitation amounts in the evening are greater by over 40% than the daily mean in summer. This must be caused by severe convective showers in the evening with the local circulation. However, the morning peak can be found still in cluster II. As Tagami(1990) has pointed out, this should be a composite of two different phenomena of the diurnal cycles. In addition, clusters I and II have the morning peaks in June and March. The mechanism producing the morning peak in this month should be an interesting topic for further study. The frontal system stays near Japan during June and March, and it brings precipitation. The frontal rainfall in June is especially called as `Baiu', and it is related to the south-west Asian monsoon. The evening peak seems to cause the low DP(m,h) at 20-24 LST in summer. There are no peaks near midnight except for November.
From the comparison of these two clusters, it is noted that there is a morning peak from March to September almost everywhere in Japan, and that there is another system producing an evening peak in summer(and sometimes in May), in some inland regions. If the evening peak system works stronger than the morning peak system, that station belongs to cluster II.
Cluster III has comparatively severe precipitation in the evening in summer, and the annual mean diurnal cycle is much different from others. They are stations where precipitation is brought mainly by thunder storms in the summer evenings, and they have less precipitation at other times of the day during the other seasons.
A similar classification was also done for the seasonal change pattern of the diurnal cycle ratio DP(m,h), the vectors of (24 12) elements at each station. Although the stations were divided into localized clusters corresponding to the seasonal change of monthly precipitation, the result has basically the same geographical distribution as Fig. 1.
In the case of the west coast regime, the mean precipitation P(m,h) and DP(m, h) are shown in Figs. 9 and 10 respectively by the ensemble mean of the 4 stations in the Peninsula. Notations are the same as the figures for Japan, but the scale is adjusted adequately in each figure. The precipitation occurs in the morning from May to October. This period is the south-west monsoon season and the monthly precipitation is also large during these months. On the other hand, similar amounts of precipitation are observed in the evening in the other season.
The inland regime includes Kuala Lumpur and Senai, and they seem to be located near the coast on the map in Fig. 8, but they are actually more than 30 km from the nearest coast. This indicates that the land-sea effect for the diurnal cycle of precipitation penetrates less than 30 km in the Malay Peninsula. The ensemble mean of the 4 stations of this regime in the Peninsula is shown in Fig. 11. There is a distinct peak of the diurnal cycle of precipitation. Precipitation is observed in the evening and very little precipitation is observed at other times of the day.
The monthly precipitation is higher in May and October. These months are several weeks after the equinox, when the solar incident angle is highest. May and October are also the period of monsoon exchange, the beginning and the ending of the south-west and the north-east monsoons. The amplitude of the diurnal cycle is approximately 200% for annual mean precipitation, and the anomaly of 15-16 LST for monthly mean precipitation is nearly 300% in March and April(Fig. 12). It indicates that precipitation observation at a fixed time, such as by the remote sensing using a sun-synchronous satellite, may produce an estimation error of a few hundred percent.
In the case of the east coast regime, the evening peak is as strong as for the inland regime, but little precipitation is observed in other LST(Fig. 13). It is clear from the figure that the precipitation is observed from midnight to evening in November, December and January. This is the north-east monsoon season, and the monthly precipitation is much higher than in other months. However, the normalized diurnal cycle of precipitation(Fig. 14) has the clear peak in the evening from May to October. The weak peak can be found in the morning from December to February, but the morning peak is weak in comparison with the evening one. Kuching has similar characteristics of precipitation, too. Although the station is located on Kalimantan Island, it faces to the ocean in the north direction and it may be affected by the north-east monsoon. Kuala Terengganu and Kota Bharu in the east coast of the Peninsula have a moderate evening peak, and the maximum LST is later at night. Sandakan in Kalimantan Island has similar pattern as these two stations, and it also faces to the ocean in the direction of north-east.
Although the Fourier transformation is often applied and the component of one-day(24 hour) cycle is often used for the analysis of the phase and the amplitude of the diurnal cycle (McGarry and Reed, 1978), the diurnal cycle pattern is not always similar to the trigonometrical function. The annual mean diurnal cycle DP(y, h) has a sudden front and a long tail, and it is difficult to say that DP(y, h) has a simple periodic shape. Figs. 15 and 16 shows the Fourier transform and integral of DP(y, h) for the inland regime of Malay Peninsula. The shape of the annual mean diurnal cycle cannot be expressed only by the sinusoid of 24 hour cycle, and 12 hour(semi-diurnal) cycle and 8 hour cycle components are required(Fig. 16). However, it can be seen clearly from Fig. 15, the semi-diurnal and 8 hour cycles are computational mode caused by the insufficiency to represent the annual mean diurnal cycle by the sin function of 24 hour cycle. The phases of these three waves correspond well, but the diurnal component delays by 1 hour. Also one should be careful to see the amplitude of the diurnal cycle by Fourier transform. The amplitude of 24 hour cycle reproduces only to the half of the evening peak in the case of Fig. 15. Therefore the maximum values of rainfall and the anomaly percent are used as they are in this study to express the diurnal cycle.
From the above considerations, the morning peak of the precipitation in Japan prevails almost all over Japan, except for the region of cluster III, inland in the Kanto area, and this peak is brought by the increase of the duration or the events of precipitation. This morning peak can be found almost throughout the year, but the evening peak can be found only in the months from May to September, when Japan is governed by the Pacific High and affected by the south-west monsoon. In this season the atmospheric conditions should be similar to that of tropical regions, and the convective rainfall will occur in the evening at inland stations.
The mean intensity of precipitation in the west coast regime of the Malay Peninsula is shown in Fig. 23. IP_m(m, h) is also shown in Fig. 24. Although the peak of the annual mean intensity at 4 LST can be found, both figures of the intensity for the west coast regime are not clear, and there may be some uncertainty. Gray and Jacobson(1977) described the morning peak of heavy rainfall at tropical land stations, but the morning peak does not consist of severe precipitation in this case.
In the case of the central Malay Peninsula, Fig. 25 shows the I_m(m, h) of the inland regime. The mean intensity at each LST in each month corresponds well to the precipitation volume(Fig. 11). The anomaly of the annual mean intensity in each LST(Fig. 26) is approximately 75%, whereas the precipitation volume has an anomaly of 200%. It indicates that the diurnal cycle of precipitation is formed by the diurnal cycle of both the intensity and the probability of precipitation. The situation is the same in the case of the evening peak of precipitation in inland Japan(Fig. 5 and Fig. 18), although the amplitude is ten times smaller than that of Malaysia.
When the figures are examined carefully, the peak of the diurnal cycle of precipitation is seen to be at 16 LST, whereas the peak of the mean intensity is at 15 LST in the west coast and the inland regimes. The significance of this gap is not clear and there are no gaps in the clusters of Japan. It is also noticeable that the diurnal cycle of the annual mean intensity is higher than the seasonal change of the monthly mean intensity in Malaysia, but the situation is reversed in Japan.
The precise mechanism of the morning peak of precipitation is still uncertain. The morning peak over the ocean is well known (Murakami, 1983), and some research (Meisner and Arkin, 1987; Kito, 1993; Nitta and Sekine, 1993) shows the diurnal cycle and morning peak is weak over the ocean far from the continents. It suggests the interaction between land and sea breeze with the synoptic wind (Houze et al., 1981; Johnson and Priegnits, 1981). From the results of this study, morning peaks can be found especially during the affecting monsoon season both in Malaysia and Japan. In the case of Malaysia, the coast lines are perpendicular to the predominant monsoon wind. It will cause the low level convergence between the monsoon wind and the offshore breeze of the local land-sea circulation. Some evidence in this study supports this mechanism. Fig. 27 and Fig. 28 show P(m, h) for Kuching and Bintulu in Kalimantan Island. Both stations are located on the coast but only Kuching faces to the north-east. Clearly the diurnal cycle of precipitation P(m, h) of Kuching is similar to the east coast regime but Bintulu is not. It indicates the importance of the convergence between the predominant monsoon wind and the offshore breeze. The negligible morning peak in Kuala Lumpur and Senai, around 30 km from the nearest coast also suggests the importance of the local land-sea circulation for the morning peak.
The morning peak in Japan is very weak and the the radiative cooling of the cloud top may be enough to form the morning peak. However there is a close relationship between the morning peak and the 'Baiu' monsoon. The hypothesis proposed by Gray and Jacobson (1977) would explain this, that the difference between the weather system and its surrounding cloud free region produces the morning maximum.
Numerical models including radiation processes coupled with cloud physics and land surface processes will be useful for a better understanding of the diurnal cycle of precipitation. Recently Randall et. al(1991) showed that an oceanic diurnal cycle of precipitation occurs even in the absence of neighboring continents and tends to have a morning maximum, but their numerical experiment was carried out by 4deg. latitude and 5deg. longitude grid. A numerical experiment with a higher spatial resolution model would be necessary for the interpretation of the diurnal cycle of Japan and Malaysia.
The seasonal change and the intensity were separated in this study, but the further detailed analysis in relation to synoptic environment or classification by intensity will help towards a better understanding of the mechanism of the diurnal cycle of precipitation. The improvement of the statistical model to describe the spatial and temporal distribution of precipitation including the diurnal cycle will be necessary for the feasibility study on the satellite observation of precipitation.
The diurnal cycle of precipitation in Japan is classified into three clusters. The first one has a peak in the morning, and the stations in this cluster are located in coastal region. The second cluster has two peaks in the morning and in the evening and these stations are located in inland region. The morning peak in the above two clusters is dominant in June, during the `Baiu' season in Japan. Baiu is the rainy season related to the south-west Asian monsoon. The third cluster is an exceptional case. No morning peak is observed in the stations of the third cluster and they have a comparatively strong evening peak.
There are three regimes of the diurnal cycle of precipitation in Malaysia. The stations in the central Malay Peninsula have a pronounced diurnal cycle of precipitation and the peak time is 16 LST for rain volume, and it is 15 LST for mean intensity. The normalized anomaly percent of the diurnal cycle in a month exceeds 300% but the seasonal change is small.
In the region affected by the north-east monsoon, namely the east coast of the Malay Peninsula, the diurnal cycle is weak in the northern hemisphere winter, and the peak is observed in the early morning. Evening peak is observed in the northern hemisphere summer.
In the west coast of the Malay Peninsula, where the south-west monsoon effects, the diurnal cycle has a peak in the morning from May to October. There is still a pronounced peak in the evening during other seasons.
In both Japan and Malaysia, an evening peak is observed in inland regions with the high intensity of precipitation. It should be caused by thermodynamic convection. The diurnal cycle of the intensity is responsible for around half of the diurnal cycle of precipitation, and in the case of Malaysia, it is higher than the seasonal change of the monthly mean intensity.
The morning peak of precipitation is observed in coastal regions in the monsoon season and the intensity of precipitation is not significantly high at these times, so the increase of the probability or the duration of precipitation produces this morning peak. The importance of the local land-sea circulation is pointed out, and the morning peak in Malaysia is probably caused by the low level convergence between the predominant monsoon wind and the offshore breeze of the local circulation. The radiative cooling of cloud tops and the contrast with surrounding cloud-free region may also be important, especially for the morning peak in Japan. The authors thank the Japan Meteorological Agency and Malaysian Meteorological Service for providing hourly precipitation data. Part of this research is supported by the Grant-in-Aid for Scientific Research from the Japanese Ministry of Education, Science and Culture. The authors also would like to thank Prof. T. Nitta in Center for Climate System Research, University of Tokyo and Dr. J. Matsumoto in Department of Geography in University of Tokyo for their helpful suggestions. Figures were drawn by Dennou Club Libraries, and maps were produced using GMT system.
Kraus, E. B., 1963: The diurnal precipitation change over the sea. J. Atmos. Sci., 20, 551-556.
Nieuwolt, S., 1968: Diurnal rainfall variation in Malaya. A. A. A. G., 58, 313-326.
Ray, C. L., 1928: Diurnal variation of rainfall at San Juan, P. R. Mon. Wea. Rev., 56, 140-141.
Total precipitation (mm h^{-1} month^{-1}) at each time of a day in each month. Mean 13 years for 90 stations of the cluster I, in coastal region of Japan.
The anomaly(%) of the diurnal cycle of precipitation, normalized in each month for the cluster I, in coastal region of Japan.
Total precipitation (mm h^{-1} month^{-1}) at each time of a day in each month. Mean 13 years for 49 stations of the cluster II in inland region of Japan.
The anomaly(%) of the diurnal cycle of precipitation, normalized in each month for the cluster II, in inland region of Japan.
Total precipitation (mm h^{-1} month^{-1}) at each time of a day in each month. Mean 13 years for 3 stations of the cluster III of Japan.
The anomaly(%) of the diurnal cycle of precipitation, normalized in each month for the cluster III of Japan
Malaysian hourly rainfall data stations used in this study
Total precipitation (mm h^{-1} month^{-1}) at each time of a day in each month. Mean 4 stations for 10 years of the west coast regime of Malay Peninsula
The anomaly(%) of the diurnal cycle of precipitation, normalized in each month. The mean of the west coast regime.
Total precipitation (mm h^{-1} month^{-1}) at each time of a day in each month. Mean 4 stations for 10 years of the inland regime of Malay Peninsula
The anomaly(%) of the diurnal cycle of precipitation, normalized in each month. The mean of the inland regime.
Total precipitation (mm h^{-1} month^{-1}) at each time of a day in each month. Mean 4 stations for 10 years of the east coast regime of Malay Peninsula
The anomaly(%) of the diurnal cycle of precipitation, normalized in each month. The mean of the east coast regime.
Fourier transform of annual mean diurnal cycle of precipitation for the inland regime of Malay Peninsula. Average, 24 hour(diurnal), 12 hour(semi-diurnal) and 8 hour cycle components are compared with the original diurnal cycle of precipitation.
Fourier integrals of Fig. 15. Solid curves indicate the composites of cycles, and original observations are plotted by circles.
Mean intensity of precipitation (mm h^{-1}) at each time of a day in each month, mean 49 stations for 13 years in inland region of Japan(Cluster II).
The anomaly(%) of the diurnal cycle of mean intensity of precipitation, normalized in each month, in inland region of Japan (Cluster II).
Mean intensity of precipitation (mm h^{-1}) at each time of a day in each month, mean 90 stations for 13 years in coastal region of Japan(Cluster I).
Mean intensity of precipitation (mm h^{-1}) at each time of a day in each month, mean 3 stations for 13 years(Cluster III).
Mean intensity of precipitation (mm h^{-1}) at each time of a day in each month by the mean of east coast regime in Malay Peninsula
The anomaly(%) of the diurnal cycle of mean intensity of precipitation, normalized in each month, by the mean of east coast regime in Malay Peninsula
Mean intensity of precipitation (mm h^{-1}) at each time of a day in each month by the mean of the west coast regime in Malay Peninsula
The anomaly(%) of the diurnal cycle of mean intensity of precipitation, normalized in each month, by the mean of the west coast regime in Malay Peninsula
Mean intensity of precipitation (mm h^{-1}) at each time of a day in each month by the mean of inland regime in Malay Peninsula
The anomaly(%) of the diurnal cycle of mean intensity of precipitation, normalized in each month, by the mean of inland regime in Malay Peninsula
Total precipitation (mm h^{-1} month^{-1}) at each time of a day in each month. At Kuching mean for 10 years.
Total precipitation (mm h^{-1} month^{-1}) at each time of a day in each month. At Bintulu mean for 10 years.
(Last updated at Thursday, 14-Feb-2002 17:43:22 JST, by Taikan OKI)
