Changes in global land monsoon area and total rainfall accumulation over the last half century

Authors


Abstract

[1] This study reports the changes of global land monsoon coverage and monsoon rainfall accumulation. We showed that the combination of monsoon area and rainfall intensity change has led to an overall weakening trend of global land monsoon rainfall accumulation during the last 54 years. This decreasing tendency is mainly caused by the North African monsoon and South Asian monsoon, due to the significant decreasing tendencies of both rainfall intensity and monsoon coverage. The long-term changes of the other monsoon subsystems are not significant in the context of regional average. The monsoon area and rainfall accumulation indices are consistent with the rainfall intensity index in revealing both the inter-decadal and interannual variability. The former appears as an inter-decadal oscillation with a peak phase through 1970-1980; the latter appears as the North African, North American and Australian monsoons having stronger spectrum power in the LF (3-7yr) band, while the South African, South and East Asian monsoons having greater powers in the QB (2-3yr) band.

1. Introduction

[2] Changes in monsoon rainfall are of great scientific and societal importance owing to their impacts on more than two-thirds of the world's population. Monsoons occur in various regions around the world. Each monsoon system generally has its own specific characteristics in terms of variability [Kripalani and Kulkarni, 2001; Jones et al., 2004; Schreck and Semazzi, 2004; Yu et al., 2004]. At the same time, the connections in the global divergent circulation necessitated by mass conservation bring about coordination among regional monsoons [Trenberth et al., 2006]. This kind of coordination has been manifested in the coherent variations between East Asian summer monsoon and North American summer rainfall [Lau and Weng, 2002; Zhang et al., 2005], South American monsoon and African monsoon [Biasutti et al., 2003; Hoerling et al., 2006], South Asian summer monsoon, Australian and eastern African monsoon [Meehl, 1997]. Hence, it is desirable to examine the change of global monsoon as a whole. Wang and Ding [2006] (hereafter WD2006) proposed an objective criterion to define global monsoon domain and the variation in global monsoon intensity. They showed an overall downward trend in monsoon rainfall intensity over global land monsoon regions during 1950-2004. Recently, this weakening tendency has been reproduced using Atmospheric General Circulation Model driven by historical sea surface temperatures covering the same period [Zhou et al., 2008].

[3] Changes in global monsoon appear not only in rainfall intensity but also in area covered by monsoon rain. The monsoon domains can migrate spatially over time with changes in internal and external forcing agents of the climate system. There has been a substantial change in land-sea thermal contrast over the past 50 years; the monsoon domain may have been affected by the change of thermal contrast. The criteria used in WD2006 focuses on rainfall intensity. Since the monsoon index was calculated as an average of precipitation intensity falling within the climatological annual range-defined monsoon region, it was unable to describe the changes of area covered by monsoon rain. The present study aims to address the following questions: 1) Are there any changes in global land monsoon area over the last half century? 2) What are the major changes of accumulated precipitation falling within the annual range-defined monsoon region? 3) Which portion of global monsoon system dominates the decreasing tendency of WD2006?

2. Data and Analysis Method

[4] Three sets of monthly rain-gauge precipitation data are used: (1) the data set compiled by Delaware University (Delaware) for the period of 1950-1999 (C. J. Willmott and K. Matsuura (2001), Terrestrial Air Temperature and Precipitation: Monthly and Annual Time Series (1950-1999), 2001; available at http://climate.geog.udel.edu/∼climate/html_pages/README.ghcn_ts2.html), (2) the data set constructed by the Climatic Research Unit (CRU) for the period of 1901-2002 [Mitchell and Jones, 2005], and (3) the Precipitation REConstruction data over Land (PREC/L) compiled for the period of 1948-2006 by the Climate Prediction Center at the National Center for Environmental Prediction [Chen et al., 2002]. To reduce uncertainties arising from difference in data sources and interpolation algorithms, an ensemble mean of the three data sets is calculated over the global land areas on a 0.5° by 0.5° grid for the period of 1949-2002. Previous studies indicate that the magnitude of the trend due to the changes in gauge networks is much less than that of the trend itself over most of the land areas [e.g., New et al., 2001; Dai et al., 2004]. A recent analysis found that the modern satellite-gauge combination data are highly consistent with the rain-gauge data used here in revealing the interannual variation of land monsoon rainfall intensity [Zhou et al., 2008], adding confidence to the reliability of the analyses based on gauge data.

[5] The local summer-minus-winter precipitation, defined as the annual range (AR), is used in the analyses. Here, summer means June-July-August (JJA) in the Northern Hemisphere (NH) and December-January-February (DJF) in the Southern Hemisphere (SH). As WD2006, the monsoon rainfall domain is defined by the region in which the AR exceeds 180 mm and the local summer rainfall exceeds 35% of annual rainfall (see auxiliary material for further information on the monsoon definition and the land-ocean mask used in the calculation of land monsoon rainfall). The NH-averaged JJA monsoon rainfall and SH-averaged DJF monsoon rainfall are used to measure the strength of NH and SH summer monsoon rainfalls. The sum of NH index and SH index in the following season is used to quantify the strength of global monsoon. Since this index is based on rainfall intensity, we term it as Monsoon Intensity Index (MII).

[6] Two new indices are employed: 1) the land monsoon coverage index, which is the total area of grid boxes covered by monsoon rainfall; 2) The accumulated monsoon rainfall index, which is the product of monsoon area and rainfall intensity. The first index is termed as Monsoon Area Index (MAI), while the second index is termed as Accumulated monsoon Rainfall Amount Index (ARAI).

3. Results

[7] The normalized time series of the NH, SH, and global MAI are shown in Figures 1a1c. The climatological, NH (SH) monsoon areas derived from three data sets are presented in Table 1. The ensemble mean is 22.7 × 106 km2 (18.2 × 106 km2), with a standard deviation of 1.1 × 106 km2 (1.1 × 106 km2) for the Northern (Southern) Hemisphere. The results derived from CRU and PRECL data are comparable with each other, while the climatology derived from Delaware data is slightly smaller. Despite of the uncertainties in climatology, the discrepancies of the anomalies among three data sets are small, and their time evolutions are consistent. The ensemble mean time series indicates an insignificant weak decreasing trend in the NH index. The SH and global indices also show no significant trends (Figures 1b and 1c). Multi-decadal oscillations are evident among the NH, SH, and global indices, although the observed period might be too short to reliably detect either multiplicative decadal variability or underlying trend [Van Oldenborgh and Burgers, 2005]. There are two broad maxima around 1960 and 1975, and a rising trend from the middle 1980s to the present. The results derived from three individual data sets are consistent. WD2006 revealed a decreasing trend in the NH MII time series across the entire 56 years. This decreasing tendency is however not significant in the MAI time series (Table 1). Of particular interest is the peak around 1975 and the rising trend from the middle 1980s (Figures 1a1c). Similar variations are evident in the MII time series of WD2006. Using the same data sets, we re-calculate the MII as WD2006 and found the correlations between the MII and MAI time series is 0.57 (0.57) for NH (SH) and 0.62 for global average, which are statistically significant at the 5% level.

Figure 1.

Normalized time series of anomalies from (a–c) climatology of land monsoon area index and (d–f) land monsoon rainfall accumulation for June-July-August Northern Hemisphere average (Figures 1a and 1d), December-January-February Southern Hemisphere average (Figures 1b and 1e), and the global land monsoon index or the sum of Figures 1a, 1b, 1d, and 1e (Figures 1c and 1f). The trends for the ensemble are also shown.

Table 1. Climatology of Local Summer Land Monsoon Area Index and Its Standard Deviation and Trend
 Northern Hemisphere (106 km2)Southern Hemisphere (106 km2)
JJA MeanStdTrend (/50yr)DJF MeanStdTrend (/50yr)
ENS22.71.1−0.38618.21.10.285
CRU22.51.1−0.14318.01.30.311
Delaware21.11.1−0.60816.61.10.128
PRECL22.81.1−0.61018.01.10.155

[8] To reveal the change of land monsoon rainfall accumulation, the normalized time series of ARAI are shown in Figures 1d1f. The climatological, NH (SH) monsoon land monsoon rainfall accumulations derived from three data sets are presented in Table 2. The ensemble mean is 157.3 × 109 m3/day (125.3 × 109 m3/day), with a standard deviation of 10.5 × 106 km2 (9.2 × 106 km2) for the Northern (Southern) Hemisphere. Most of the points made in connection with MAI (Figures 1a1c) apply here. The evolution of ARAI matches that of MII shown in WD2006: a significant decreasing tendency is evident in the NH and global ARAI time series. The correlation between the MII and ARAI time series is 0.89 (0.80) for NH (SH) and 0.88 for global average. Both the NH and global ARAI trends are statistically significant at the 5% level (Figure 2h). The SH ARAI shows no significant trend.

Figure 2.

Normalized time series of anomalies from climatology of summer monsoon indices derived from the precipitation intensity (red), monsoon area (green), and precipitation amount (blue) for different regional monsoon domains, (a) North African monsoon; (b) South African monsoon; (c) South Asian monsoon; (d) East Asian monsoon; (e) Australian monsoon; (f) North American monsoon; and (g) South American monsoon, and (h) the trends across 1949–2002 for each index. The letter “Y” in Figure 2h delineates that the trend is statistically significant at the 5% level.

Table 2. Climatology of Local Summer Land Monsoon Rainfall Accumulation Index and Its Standard Deviation and Trenda
 Northern Hemisphere (109 m3/day)Southern Hemisphere (109 m3/day)
JJA MeanStdTrend (/50yr)DJF MeanStdTrend (/50yr)
  • a

    Bold numbers indicate the trends are statistically significant at the 5% level.

ENS157.310.5−1.132125.39.2−0.032
CRU158.611.9−0.814125.711.10.322
Delaware144.110.0−1.214113.88.9−0.040
PRECL157.410.5−1.316125.28.8−0.318

[9] The global monsoon is divided into seven portions, i.e., North African monsoon (NAfr), South African monsoon (SAfr), South Asian monsoon (SA), East Asian monsoon (EA), Australian monsoon (Aus), Northern American (NAme), and South American monsoon (SAme) domains. The time series of MII, MAI and ARAI for each individual monsoon subsystems are shown in Figures 2a2g. The trends across 1949-2002 for all indices are given in Figure 2h. Significant decreasing trends are evident in all three indices for NAfr and SA. The change of land monsoon rainfall accumulation is determined firstly by how strong it rains and secondarily by how large of the area covered by monsoon rain, as evidenced in Figure 2. The long-term trends of South Africa, Australian, South American, and North American monsoons are not statistically significant (Figure 2h). The MAI and MII trends of East Asian monsoon are not significant, mainly due to the out of phase change of rainfall in southern and northern China [e.g., Hu et al., 2003; Yu and Zhou, 2007]. Hence the weakening tendency of global monsoon rainfall reported in WD2006 is mainly caused by NAfr and SA monsoon.

[10] Besides the weakening tendency, another prominent feature of hemispheric (global) monsoon variation is the peak through 1970-1980 (Figure 1). As seen from Figure 2, this peak stands out mainly (marginally) in the SAme (Aus) monsoon index. The global climate has experienced a shift in the mid-1970s; the mid-1970s peak of monsoon index might reflect this inter-decadal shift [Trenberth and Hurrel, 1994].

[11] Besides the trends and inter-decadal variations, the time series of Figures 1 and 2 also show apparent interannual variability. To have a clear picture of the dominant time scales, power spectra of monsoon indices for each individual monsoon subsystems are shown in Figure 3. The NAfr has a single peak at 3.5-yr; all three indices are consistent in this peak. The SAfr has double spectral peaks at 2.5 and 4-yr. The 2.5-yr peak is more significant than the 4-yr peak in the area and amount indices, but the reverse is true for the intensity. All three indices of Australian monsoon have spectral peaks at 3.5-yr. The three indices of NAme (SAme) have a common peak at 3.5-yr (3-yr). Both SA and EA have 2.5-yr peak in all three indices; however, the 3.5-yr peak only exists in the intensity and amount indices. In summary, the 3.5-yr peak is significant mainly in the North African, South and East Asian, Australian and North American monsoon indices, while the 2.5-yr peak is significant mainly in the South African, South and East Asian monsoon indices.

Figure 3.

Power spectra of summer monsoon indices derived from (left) the precipitation intensity, (middle) monsoon area, and (right) precipitation amount for different monsoon components. The horizontal thin dashed line denotes white-noise power density. A spectrum with a peak above the thin dashed line implies that it is distinguished from a white-noise spectrum with a confidence level over 95%.

[12] The El Nino-Southern Oscillation (ENSO) comprises both quasi-biennial (QB, 2-3 yr) and Low frequency (LF, 3-7 yr) components [Rasmusson et al., 1990]. The Asian-Australian monsoon also exhibits remarkable QB and LF variability [Webster et al., 1998; Meehl, 1997; Wang et al., 2003; Li et al., 2006]. The NAfr, NAme and Australian monsoons all have a stronger spectrum power in the LF band, while the SAfr, SA, and EA monsoon have a greater power in the QB band and a secondary power in LF band. Since the ENSO has a stronger spectrum power in LF band, the interannual variability of NAfr, NAme and Australian monsoon is closely related to ENSO [Castro et al., 2001; Wang et al., 2003; Schreck and Semazzi, 2004]. The dominance of QB band in the power spectrum of SAfr, SA, and EA indices suggest that the origin of interannual variability does not directly relate to ENSO. In addition, associated with interdecadal fluctuations in ENSO, the monsoon-ENSO relationship may also experience interdecadal changes [e.g., Wang et al., 2008] (also see auxiliary material for further information on the interdecadal change of monsoon-ENSO relationship). Since one might question the validity of monsoon indices derived from monsoon coverage and rainfall accumulation, the consistency of three indices in describing above well-known typical time periods suggest that these indices are useful in describing the monsoon variation.

4. Summary

[13] To measure the change of global land monsoon rainfall, we have defined two new indices according to the monsoon area and rainfall accumulation, respectively. The monsoon area index quantifies the change of area covered by monsoon rainfall, while the monsoon rainfall accumulation index measures the product of monsoon rainfall intensity and monsoon area changes. We have examined the changes of global land monsoon rainfall over the last 54 years (1949-2002) by employing the new indices. The main findings are listed below:

[14] 1) The global land monsoon rainfall accumulation exhibits an overall weakening trend in the last 54 years. This decreasing tendency is mainly caused by the North African and South Asian monsoon. Both the rainfall intensity and monsoon area changes contribute to this decreasing tendency.

[15] 2) The long-term trends of South Africa, Australian, South American, and North American monsoons are not statistically significant. The East Asian summer monsoon has an insignificant decreasing tendency in intensity but an increasing tendency in coverage. The increasing tendency of monsoon coverage offsets the decreasing tendencies of the other monsoon subsystems, resulting in an insignificant decreasing tendency of global (Northern Hemisphere) land monsoon area index.

[16] 3) The indices derived from monsoon coverage and rainfall accumulation are consistent with the intensity index of WD2006 in revealing typical variations of global land monsoon, e.g. the global monsoon exhibits an inter-decadal oscillation with a peak phase through 1970-1980; at interannual time scale, the North African, North American and Australian monsoons have stronger spectrum power in the LF band, while the South African, South and East Asian monsoons have greater powers in the QB band.

Acknowledgments

[17] This work was jointly supported by the National Basic Research Program of China (2006CB403603), the National Natural Science Foundation of China under grants 40628006, 40625014, and 40675050.

Ancillary