Precipitation and precipitable water: Their temporal-spatial behaviors and use in determining monsoon onset/retreat and monsoon regions

Authors


Abstract

[1] Precipitation (P) is conventionally used for determining monsoon onset/retreat. It can roughly separate monsoon regions from nonmonsoon regions. Zeng and Lu (2004) found that precipitable water (W) can also determine the monsoon onset/retreat but cannot determine the monsoon regions. Temporal-spatial behaviors of P and W are compared in this article with observed data and results of previous theoretical analyses, and the comparison is used to understand the performances of P and W in determining the monsoon onset/retreat and monsoon regions. It is shown that W increases everywhere from winter to summer, dominated by the large seasonal change of temperature. P increases from winter to summer mainly in monsoon regions; it decreases or does not change much in most of the nonmonsoon regions. Whether P increases or not from winter to summer depends on whether the increase of W is greater than the increase of temperature. Synoptic variations of P and W have positive correlations everywhere. The increases of P and W from winter to summer in monsoon regions make both able to determine the climatic monsoon onset and retreat. The positive correlations of daily P and W in monsoon onset and retreat seasons make the interannual variations of the monsoon onset and retreat determined from P able to be determined from W. The decrease or small change of P from winter to summer in most of the nonmonsoon regions and the increase of W in nonmonsoon regions make P able to roughly determine the monsoon regions while W fails to.

1. Introduction

[2] The Globally Unified Monsoon Index (GUMI), proposed by Zeng and Lu [2004] for comparing and linking the different monsoons over the globe, uses precipitable water (W) as the single parameter. The rationale of the GUMI is using the annual cycle of W to represent the annual monsoon process, which includes summer and winter monsoons as well as the two transitions between them, and regarding the monsoon onset and retreat as phase-locked phenomena with each corresponding to a relatively fixed stage of the annual cycle, which is indicated with the golden ratio. It was hoped that this single parameter W could also determine the monsoon regions, with the hypothesis that for each grid point in the globe, one could first identify if it is a monsoonal grid point or not, and if it is a monsoonal one then provide the monsoon onset and retreat dates. The results showed that the climatic monsoon onset and retreat determined from W are in general consistent with those obtained from the different methods used for the different monsoon regions, which are mostly based on precipitation. However, W fails to determine the monsoon regions; when the generally recognized monsoon regions in the globe were all included through choosing a suitable critical value for the annual range of W, some nonmonsoon areas were also included. These areas were shaded gray in Figures 2 and 3 of Zeng and Lu [2004], and had to be removed from the monsoon regions subjectively.

[3] Precipitation (P) is conventionally used for determining the monsoon onset and retreat [e.g., Das, 1987; Tao and Chen, 1987; Hendon and Liebmann, 1990; Douglas et al., 1993; Murakami and Matsumoto, 1994; Higgins et al., 1997]. The fundamental reason is that P can significantly affect agriculture and other human's activities, can be observed from the surface, and, as a by-product of the monsoon circulation, can well reflect the physical essence of the monsoon. W has been applied to determine the monsoon onset and retreat in both research [e.g., Cadet, 1986] and operational work (e.g., by the National Weather Service at Arizona). The consideration of using W in the GUMI is that it can be available over oceans and lands, and can be less affected by the surface condition than P. It is relatively stable during the monsoon seasons compared with P; the latter may have breaks within monsoon season and may appear before and after the monsoon season in some monsoon areas, forming the premonsoon P and the fall P, respectively. Because of the characteristics that it increases rapidly from winter to summer in monsoon regions, P has been used to determine the monsoon onset and retreat. Since the increase of P from winter to summer is mainly in monsoon regions and in most of the nonmonsoon regions it does not increase, P can roughly separate the monsoon regions from the nonmonsoon regions.

[4] How can we understand that on the one hand, both P and W can determine the monsoon onset and retreat, but on the other hand, P can roughly determine the monsoon regions but W fails to? It was later realized from observed data [Lu and Zeng, 2005] that these can be attributed to the different geographical characteristics of the seasonal patterns of P and W. Theoretical analyses were subsequently made to interpret the temporal-spatial behaviors of these two quantities. Lu [2007] examined how the seasonal change of W is dominated by the change of temperature. Lu and Zeng [2005] investigated how the seasonal change of P is regulated by the change of W and the change of temperature. Although P and W are different quantities, with the former being the water vapor that has been consumed for condensation and left the atmosphere and the latter being the water vapor that is still contained in the atmosphere, and in different units, they are both two-dimensional observational hydrological quantities. Their linkage is of an interest in the hydro-climate community. The first purpose of this article is to explore from observed data how P and W relate and differ at different timescales and locations, and understand the results briefly with the previous theoretical studies.

[5] The GUMI has received extensive attention in the monsoon community and has been applied or cited by many researchers [e.g., Li and Chen, 2005; Liu et al., 2005; Kitoh and Uchiyama, 2006; Guan and Chan, 2006; Taniguchi and Koike, 2006; J. Xu et al., 2006; M. Xu et al., 2006; Smith et al., 2008; Zhang, 2009]. For example, based on the principle of the GUMI, Kitoh and Uchiyama [2006] constructed a monsoon onset and retreat index using P for the East Asian summer monsoon. Smith et al. [2008] compared the climatic onset dates of the Northern Australian monsoon that were determined with the eight indexes or methods from the published references, and showed that the result of the GUMI is the best; the date is the same as what he determined from the precipitation data. The second purpose of this article is, to ensure the reliability of the GUMI in monsoon studies, to clarify the issues on the rationality of the use of W in indicating the monsoon onset/retreat and the reason why W cannot determine the monsoon regions with the understanding of the temporal-spatial behaviors of P and W obtained from the data and theoretical studies.

[6] The observed data of P and W are introduced in section 2. In section 3, the temporal-spatial behaviors of P and W and their relations are first revealed from the observed data and then interpreted briefly with results of theoretical studies. In section 4, the issues of the GUMI on the monsoon onset/retreat and monsoon regions are clarified. Summary and discussions are given in section 5.

2. Data and Processing

[7] The general characteristics of the temporal-spatial patterns of P and W are illustrated in this study with regional data sets, which include the 10 year (1988–1997) daily analyses of P and W over the continental United States and Mexico (U.S.-Mexico) at a 1° × 1° resolution, distributed from the Climate Prediction Center of NOAA [e.g., Higgins et al., 1999] and the NASA Water Vapor Project [Randel et al., 1996], respectively.

[8] Denote the original data of P and W at a grid point as P0(d, y) and W0(d, y), respectively, where y is the year changing from 1 to 10 and d is the day of the year from 1 to 365. The mean annual cycles P(d) and W(d), obtained by averaging the original time series over y, primarily reflect the seasonal changes of P and W. The daily deviations p(d, y) and w(d, y), obtained by removing the mean annual cycles from the original time series, primarily reflect the day-to-day or synoptic variations of P and W.

3. Temporal-Spatial Behaviors of P and W

3.1. Behaviors Revealed From Observed Data

[9] Figure 1 is the distribution of the correlation between P(d) and W(d), the 10 year averaged daily P and W. It shows that at seasonal-scale P and W have different relations in different areas. Strong positive correlations, which are much greater than 0.10 (the correlation coefficient at the 95% confidence level), exist in the North American monsoon region (i.e., Mexico and the southwestern United States) and some nonmonsoon areas (e.g., the central United States and Florida). Correlations are greater than 0.80 in southern Mexico (SM). Strong negative correlations appear in the west coast of the United States (WC), and the correlation can be stronger than −0.60. Correlations are much weaker in the southeastern United States (SE).

Figure 1.

Distribution of the correlation coefficient between P(d) and W(d), the daily P and W averaged over the 10 years. Three grid points are marked as a, b, and c for use in Figure 2.

[10] To understand why seasonal P and W possess different relations in different areas, their seasonal patterns are examined. Figure 2 presents the seasonal cycles of P and W at three locations, which are in WC (123°W, 46°N), SM (105°W, 21°N), and SE (90°W, 33°N). The W at the three locations is similar in pattern, all increasing from winter to summer. By contrast, P exhibits different seasonal patterns in the three locations, which leads to its different relations with W. For the out-of-phase relation in WC (Figure 2a), when W is the least of the year in winter, P is the largest, and when W is the largest in summer, P is the least. For the in-phase relation in SM (Figure 2b), P and W both increase from winter to summer. In SE (Figure 2c) where the correlation is weak, W increases from winter to summer but P does not change much throughout the year. Finkelstein and Truppi [1991] analyzed the spatial distribution of precipitation seasonality in the United States. The seasonal cycles of P in Figure 2 represent the three typical precipitation seasonality patterns.

Figure 2.

The daily P and W averaged over the 10 years at grid points in (a) United States west coast (WC; 123°W, 46°N), (b) southern Mexico (SM; 105°W, 21°N), and (c) southeastern United States (SE; 90°W, 33°N) as marked in Figure 1 [from Lu and Zeng, 2005].

[11] Figure 3 is the distribution of the correlation between the W(d) at (100°W, 40°N), a reference point arbitrarily chosen in the central United States, and the entire field of W(d). The strong positive correlations over the domain show that the seasonal W has a very high spatial coherence, especially over the land. It can thus be inferred that the seasonal increase of W from winter to summer is true everywhere, not just for the three locations in Figure 2. With this, it can also be inferred that the three different seasonal patterns of P (with P increasing, decreasing, or not changing much from winter to summer) are not just for the three locations, but for the areas in Figure 1 where the correlations of seasonal P and W are strongly positive, strongly negative, and weak, respectively.

Figure 3.

Distribution of the correlation between W(d), the daily W averaged over the 10 years at 100°W, 40°N (marked) and that over the entire field.

[12] Different from the relations of seasonal P and W that can be positive and negative in different places, the correlations between synoptic P and W are positive everywhere, and the correlations can be fairly strong when the period is sufficiently long. Figure 4 presents the distribution of the correlation between p(d, y) and w(d, y) for the period from 1 May to 31 July, which can be regarded as the summer monsoon onset season. The 10 year averaged correlations are positive over the entire domain, and the coefficients are greater than 0.20, the correlation coefficient at the 95% confidence level, in most of the U.S.-Mexico. Correlations in monsoon retreat season, as well as any other periods that are sufficiently long, are also positive everywhere in the domain (figures not shown). It was pointed out that in synoptic processes P and W generally have positive correlations [Glickman et al., 2000].

Figure 4.

Distribution of the 10 year averaged correlation between p(d, y) and w(d, y), the daily deviations of P and W, during the monsoon onset season (1 May to 31 July).

3.2. Understanding the Behaviors From Theoretical Analyses

[13] Precipitation, water vapor, and temperature are fundamental climate quantities. They interact with each other through dynamic and thermodynamic processes, and finally display different relationships in different timescales and geographic regions. Theoretical analyses have been made to understand whether these relationships are dominated by the thermodynamic processes, and how they can be affected by the atmospheric circulation.

[14] The observed seasonal pattern of W, which increases everywhere from winter to summer, is generally attributed to the Clausius-Clapeyron relation [e.g., Peixoto and Oort, 1992]. It is worthwhile to note that the water vapor holding capacity of the atmosphere, which is determined by temperature, is one thing, while the actual water vapor contained in the atmosphere is another thing. The assumption behind the statement that water vapor (or W when column-integrated) increases with temperature is that relative humidity is constant or does not change much. However, because of the effect of the atmospheric circulation and other physical processes, relative humidity can vary, and whether its variation is small or not needs to be evaluated. Lu [2007] made a scale analysis to examine this, in which the effect of the atmospheric circulation is reflected with the change of relative humidity. Derived from thermodynamic laws, the change of W from winter (state 1) to summer (state 2) was finally expressed as ΔWimage − 1], where ΔT is the annual range of surface temperature, and r1 and r2 are surface relative humidity in winter and summer, respectively. This relation suggests that whether W increases or not from winter to summer depends on whether the seasonal temperature increase ΔT is greater or not than Dr ≡ 13ln(r1/r2), a scale of the decrease of relative humidity from winter to summer. As estimation, taking the annual range of surface temperature ΔT as 35°C and the winter surface relative humidity r1 as a moderate value of 0.50, the summer surface relative humidity r2 will only need to be greater than 0.03 to gain an increase of W from winter to summer. The extremely small relative humidity of being less than 0.03 is hard to reach in the real atmosphere, which is an open system and has interactions with ocean and land. This illustrates that the increase of W from winter to summer is dominated by temperature, because of its large annual range. The atmospheric circulation, even if its effect is adverse, cannot prevent W from increasing from winter to summer, although it can influence the value of the increase.

[15] The observed different seasonal patterns of P and its different relations with W can generally be attributed to the dynamic effect of the atmospheric circulation, which may play the roles of transporting water vapor, converging air, and making atmosphere instable and ascending. These roles can be different for precipitation in different places and seasons. Lu [2005] showed the difference in circulation between winter precipitation in WC and summer precipitation in SM. It is not the intent of this study to examine the difference of all these roles for precipitation in all the different places and seasons. While the difference in atmospheric circulation, its ultimate effect is the same; the circulation contributes to change the local water vapor and temperature to make air saturated for precipitation. With this consideration, a simple unified method was used by Lu and Zeng [2005] to understand the different seasonal patterns of P and its different relations with W in the context of the changes of water vapor and temperature. It was found from observed data that the seasonal changes of P can be well described by relative humidity, and this is consistent with the finding of Bretherton et al. [2004], in which the observed seasonal precipitation and column-relative humidity have a very strong positive relation. This implies that whether relative humidity, thus precipitation, increases or not from winter to summer depends on the comparison between the seasonal change of water vapor and the seasonal change of temperature, which were defined as Cvape2/e1 and Ctemes(T2)/es(T1), respectively, where e is vapor pressure and es the saturation vapor pressure at temperature T. The above analysis of the seasonal variation of W shows that W increases everywhere from winter to summer. However, this increase can be large in some areas and small in other areas, when compared with the increase of temperature. In the places where the increase of water vapor from winter to summer is much greater than the increase of temperature (e.g., in SM), seasonal P and W can be positively correlated. In the places where the increase of temperature is much more than the increase of water vapor (e.g., in WC), seasonal P may conversely change with temperature, and thus be negatively correlated with seasonal W. In places where the seasonal changes of water vapor and temperature are equivalent (e.g., in SE), P and W are weakly correlated.

[16] Bretherton et al. [2004] pointed out that precipitation and column-relative humidity also hold a strong positive relation at daily timescales. Using the North American Regional Reanalysis, we calculated the correlations of daily P with W and column-relative humidity at different locations in period 1 May to 31 July of different years, and found that the P-relative humidity correlations are generally stronger than the P-W correlations. From the physics of precipitation, the necessary condition of forming precipitation is the saturation of the air, and what links directly with precipitation is relative humidity, rather than water vapor amount. For synoptic processes that are sufficiently long and contain both rainy and dry days, especially when the numbers of the rainy and dry days are equivalent, the variations can majorly be reflected as the contrast between the rainy days and the dry days. Normally, in rainy days (there is precipitation) relative humidity is higher, and in dry days (there is no precipitation) relative humidity is lower. The synoptic variations of precipitation can therefore be well described by relative humidity. The relations between synoptic P and W can thus be regulated by the variation of temperature. The correlations of synoptic P and W that are positive everywhere revealed from the observed data suggest that water vapor changes more in synoptic processes than temperature.

4. Use of P and W in Determining Monsoon Onset/Retreat and Monsoon Regions

4.1. Monsoon Onset/Retreat

[17] The use of P for determining the monsoon onset and retreat is based on its characteristics that it is related to the monsoon circulation and increases (decreases) rapidly during the monsoon onset (retreat) season. Any other quantity, when considered for indicating the climatic monsoon onset and retreat, should at least possess these characteristics either. Dominated by local temperature, the seasonal variation of W is formed majorly through the change in the thermally induced atmospheric circulation [Lu, 2007]. In monsoon regions, W is also related to the monsoon circulation. According to section 3, W also increases from winter to summer everywhere in monsoon regions. Thus fundamentally W is a qualified candidate for determining the climatic means of the monsoon onset and retreat.

[18] It can be inferred from the significant positive correlations between synoptic P and W during the monsoon onset and retreat seasons in monsoon regions that a larger (smaller) daily W statistically corresponds to a larger (smaller) daily P. For each of the P and W, if its values around the climatic monsoon onset (retreat) date in a specific year are larger, then an earlier onset (later retreat) date can statistically be expected in this year. These suggest that when an earlier (later) onset or retreat date is determined from P, an earlier (later) onset or retreat date can also be statistically determined from W. In other words, the interannual variations of the monsoon onset and retreat indicated from P can also be indicated from W.

[19] What analyzed here are the preliminary requirements for becoming a monsoon index. Compared with W, temperature also relates to the monsoon circulation and increases from winter to summer. However, its synoptic relations with P are primarily negative or weak. So, maybe it can indicate the climatic monsoon onset and retreat, but it cannot determine the interannual variations of the monsoon onset and retreat.

4.2. Monsoon Regions

[20] As revealed in section 3, W increases from winter to summer in both monsoon and nonmonsoon regions. The annual range of W in some of the nonmonsoon regions can be as large as or larger than that in monsoon regions, which makes it infeasible to separate the monsoon regions from the nonmonsoon regions with the annual range of W. By contrast, P increases from winter to summer mainly in monsoon regions. It is noticed in section 3 that P may also increase from winter to summer in a small portion of the nonmonsoon regions, e.g., in part of the central United States and Florida. In most of the nonmonsoon regions, P decreases or does not change much from winter to summer. So, overall, P can roughly (though cannot fully) distinguish the monsoon regions from the nonmonsoon regions.

5. Summary and Discussions

[21] P and W are both two-dimensional observational hydrological quantities. Their linkage is of an interest in the hydro-climate community. In this study, the temporal-spatial behaviors of P and W and their relationships are explored with observed data and theoretical analyses. The results are used to clarify the issues in the GUMI, which shows that W can determine the monsoon onset and retreat but cannot determine the monsoon regions. The schematic in Figure 5 presents an overall picture of the analyses in sections 3 and 4.

Figure 5.

A schematic for comparing the temporal-spatial behaviors of P and W and their use in determining the monsoon onset/retreat and monsoon regions.

[22] Observed data show that W has a similar seasonal pattern; it increases everywhere from winter to summer. However, the seasonal pattern of P can vary with location. In addition to the increases from winter to summer in some regions, which are mainly monsoon regions, P may have decreases or small changes from winter to summer in other regions, which include most of the nonmonsoon regions. Although seasonal P and W have different relations in different places, synoptic P and W possess positive correlations everywhere.

[23] The variations of W can be influenced by the dynamic and thermodynamic processes in the atmosphere. Lu [2007] studied the seasonal pattern of W with scale analysis, in which the dynamic effect is reflected with the change of relative humidity, and demonstrated that the seasonal pattern of W is dominated by temperature because of its large seasonal change, and the effect of the atmospheric circulation cannot prevent W from increasing from winter to summer.

[24] The variations of P can also be attributed to the dynamic and thermodynamic effects of the atmosphere. The roles of the atmospheric circulation in the formation of P may include the transport of water vapor, convergence of air, atmospheric instability, and ascending motion. Although the difference of the roles in different places and seasons, their ultimate effects are all to change the local water vapor and temperature to make the air saturated to form precipitation. Lu and Zeng [2005] provided a unified understanding of the different seasonal patterns of P in the context of the changes of water vapor and temperature.

[25] The understanding is based on the observed fact that relative humidity can well describe the seasonal variations of P. This suggests that, although W increases with temperature from winter to summer, whether P increases or not from winter to summer depends on the comparison between the increase of W and the increase of temperature. In regions where water vapor increases much more than temperature, which are mainly monsoon regions, seasonal P and W have positive correlations. In regions where temperature has equivalent or large increase compared with water vapor, which cover most of the nonmonsoon regions, seasonal P and W have negative or weak correlations.

[26] At daily timescales, relative humidity can be linked with P in physics, and can well describe the variations of P. This can help to understand the observed correlations of synoptic P and W that are positive everywhere, suggesting that temperature does not change much in synoptic variations compared with water vapor. On the basis of our recent work, the P-relative humidity relationship is also strong and positive at interannual timescale, and this general relationship can be used to better understand the interannual P-water vapor and P-temperature relationships.

[27] P increases from winter to summer in monsoon regions, and is conventionally used for determining the monsoon onset and retreat. It decreases or does not change much in most of the nonmonsoon regions, so can roughly determine the monsoon regions. Differently, W increases from winter to summer in both monsoon and nonmonsoon regions, so it can determine the climatic means of the monsoon onset and retreat but cannot separate the monsoon regions from the nonmonsoon regions. Since the daily P and W in monsoon onset and retreat seasons have significant positive correlation, the interannual variations of the monsoon onset and retreat determined from P can statistically be determined from W.

[28] Another issue of the GUMI is the rationality of the criterion, in which the golden ratio (0.618) is used semiempirically as the threshold. The monsoon onset (retreat) date is defined as the day when the GUMI, the daily W normalized in the year, reaches (decreases to) the golden ratio. Fasullo and Webster [2003] used 0.50 as the threshold for simplicity. The use of the golden ratio allows to have transitions between the summer and winter monsoons. As estimation, take the annual cycle of the index as a sine wave. Then the annual monsoon process can be divided into a summer monsoon with the index greater than 0.618, a winter monsoon with the index less than 0.382, and two transitions with the index between 0.382 and 0.618. This means that each of the summer and winter monsoons lasts for 3.5 months, and each of the transitions lasts for 2.5 months. For a typical monsoon, e.g., not in the edge of a monsoon region, the onset (retreat) may usually be a rapid process, and it takes only a few days for the index to increase (decrease) from a very lower (higher) value to a very higher (lower) value, thus the determination of the onset and retreat dates is not very sensitive to the selection of the threshold.

Acknowledgments

[29] The analysis data of precipitation and precipitable water are distributed by the NCEP/CPC and the NASA/DAAC, respectively. The three anonymous reviewers and the editor are thanked for their helpful comments and suggestions. This work was partially supported by grant KLME0803 from the Key Laboratory of Meteorological Disaster of Ministry of Education at Nanjing University of Information Science and Technology.

Advertisement