1. Introduction
Changes to the amount, distribution, timing, and frequency of precipitation are among the most societally relevant consequences of climate change. A vast body of literature (e.g., Durack et al. 2012; Hegerl et al. 2015; Min et al. 2008; Polson et al. 2013b; Santer et al. 2007; Zhang et al. 2007; Marvel and Bonfils 2013) has begun to identify projected and observed changes to various aspects of the hydrological cycle, but observational uncertainty, model error, large internal variability, and gaps in theoretical understanding hinder progress in detecting precipitation changes in observations and/or attributing them to anthropogenic forcing (Hegerl et al. 2015; Sarojini et al. 2016).
Expected precipitation changes are often partitioned into a thermodynamic component resulting from increased atmospheric water vapor (Allen and Ingram 2002; Held and Soden 2006) and a dynamic component associated with changes to the atmospheric circulation (Chou et al. 2009; Seager et al. 2010; Bony et al. 2013). A fundamental response to a simple, uniform warming can be described as a combination of 1) the convergence of increased moisture by the original circulation, so that wet regions become wetter and dry regions drier, and 2) changes in the zonal mean circulation that shift the locations of these wet and dry regions by expanding the tropical belt and moving the storm tracks poleward. In many places precipitation is tied to the annual cycle, with most rain falling in a distinct wet season. Forced changes in precipitation can therefore also be described as changes in the amplitude (the range between wet and dry season precipitation) and phase (the onset time of the wet season) of the precipitation annual cycle and may be attributed to thermodynamic and dynamic effects. Previous work has indicated that the amplitude of the annual cycle of tropical precipitation will increase under increased greenhouse gas forcing (e.g., Chou and Lan 2012; Dwyer et al. 2014; Tan et al. 2008) and that this change can be explained in terms of the wet-get-wetter response to uniform warming (Dwyer et al. 2014). Models from both phases 3 and 5 of the Coupled Model Intercomparison Project (CMIP3 and CMIP5) also project a delay of the onset of the rainy season in the tropics, particularly in monsoon regions (Biasutti and Sobel 2009; Biasutti 2013; Seth et al. 2011, 2013), but the mechanisms responsible for this signal are less understood, as both local land surface effect and more general changes in circulation have been implied (Dwyer et al. 2014; Seth et al. 2013).
Because the thermodynamic response is relatively robust across models (Chou et al. 2009), signals associated with this mechanism might be expected to emerge sooner in the observational record. Chou et al. (2013) have indicated that, globally, the range between the wet and dry season increased in the 1979–2010 GPCP dataset. However, the signal they identify is especially large in the higher latitudes, where satellite estimates of precipitation are more prone to biases. Moreover, there is little overlap between regions where an intensification of the annual cycle is identified and regions in which the annual cycle is large.
Previous work (e.g., Durack et al. 2012; Zhang et al. 2007) has identified a human influence on various aspects of the water cycle over the twentieth century. However, the most recent precipitation changes have occurred against a backdrop of the slower-than-expected (compared to models) warming: the much-discussed global warming hiatus (Fyfe and Gillett 2014; Hawkins et al. 2014). This muted global temperature rise has been accompanied by cool temperatures in the equatorial Pacific (Kosaka and Xie 2013) and concomitant intensification of the trade winds (England et al. 2014). There has been vigorous debate over the causes (Huber and Knutti 2014; Kaufmann et al. 2011; Santer et al. 2014; Schmidt et al. 2014), relevance (Johansson et al. 2015), and even statistical existence (Cowtan and Way 2014; Fyfe et al. 2016; Karl et al. 2015) of the hiatus as defined by global mean temperature. However, multiple lines of evidence suggest the recent slowdown in warming resembles a negative phase of the interdecadal Pacific oscillation (IPO; Meehl et al. 2013; Steinman et al. 2015) or sustained La Niña–like conditions (Held 2013). These conditions are attributable to some combination of internal decadal variability, natural forcings (Santer et al. 2014; Schmidt et al. 2014), and anthropogenic warming, and they substantially affect many trends measured over the relatively short satellite era.
How has the seasonality of precipitation changed over the satellite era? How can previous claims of model–observation agreement on a rich-get-richer amplification driven by surface warming be reconciled with the hiatus? These questions require a careful comparison of model-predicted and observed patterns. Here, we perform a formal detection and attribution (D&A) study of the amplitude and phase of the precipitation annual cycle. We consider zonal mean precipitation P because the averaging helps to reduce observational and model uncertainty and, as we shall see, because the zonal mean facilitates physical interpretation of the results.
2. Methods
a. Datasets used
Only two datasets provide long-term measurements of precipitation over both land and ocean. These are the merged satellite–gauge products of the Global Precipitation Climatology Project (GPCP) and the Climate Prediction Center Merged Analysis of Precipitation (CMAP; see below), which provide 2.5° gridded monthly mean data from January 1979 to December 2014. While both datasets provide global coverage, we consider only the region from 50°S to 50°N. The reduced domain eliminates the high latitudes, where the datasets are considered unreliable and will facilitate comparisons (in subsequent work) with the shorter-length TRMM dataset. There are well-known discrepancies between these two satellite datasets, particularly over the oceans (Yin et al. 2004), which result in substantial observational uncertainty. Furthermore, issues with satellite drifts and merging of inhomogeneous data sources may reduce their suitability for long-term trend analysis. Despite these shortcomings, these datasets remain the best global products available. Previous studies have used only GPCP to examine large-scale trends in combined ocean and land precipitation (e.g., Marvel and Bonfils 2013; Polson et al. 2013a). Here, we will show that the two datasets show similar trends in the amplitude and phase of the precipitation annual cycle, enhancing our confidence in the results.
1) GPCP
GPCP (Huffman et al. 1997), version 2.2, begins in 1979 and was last updated in October 2015. It blends retrievals from multiple satellite datasets, sounding observations, and land-based gauges. Post-1988, GPCP incorporates data from the Remote Sensing Systems (RSS) Special Sensor Microwave Imager (SSM/I) over the ocean (Wentz and Spencer 1998).
2) CMAP
NOAA’s CMAP (Xie and Arkin 1997) combines land and low island/atoll gauge data with microwave and infrared observations from polar-orbiting and geostationary satellites. We use the enhanced version, which incorporates the NCEP–NCAR reanalysis to fill gaps in spatial coverage (although this is less relevant for the domain under consideration).
b. Models
We use model simulations from CMIP5 (Taylor et al. 2012). To estimate unforced internal climate variability, we rely on preindustrial (piControl) simulations with no changes in external climate forcings. For comparison with recent observations we use historical simulations with estimated changes in anthropogenic and natural forcings over the period 1860–2005, extended to 2100 by splicing with the corresponding RCP8.5 simulation in which twenty-first-century changes in greenhouse gases and anthropogenic aerosols are prescribed according to the representative concentration pathway 8.5 (ALL+8.5; Van Vuuren et al. 2011).
c. Amplitude and phase time series
(a) MMM 1979–2014 annual mean P (dashed black line), P annual cycle amplitude (area of circle), and phase (color) climatologies. (b) Proportion of total 1979–2014 time series variance explained by the annual cycle in the GPCP (blue) and CMAP (red) datasets and in ALL+8.5 CMIP5 models (orange).
Citation: Journal of Climate 30, 13; 10.1175/JCLI-D-16-0572.1
d. Relative importance of the annual cycle
e. D&A toolkit
In this paper, we will use a standard set of “fingerprinting” and signal detection methods presented in, for example, Santer et al. (2005, 2011, 2013).
1) Fingerprinting
The fingerprint F(ϕ) of climate change is the spatial pattern that characterizes the climate system response to external forcing (Allen and Stott 2003; Gillett et al. 2002; Hegerl et al. 1996; Stott et al. 2000; Tett et al. 2002). Following, for example, Hasselmann (1993) and Santer et al. (2011), we define the fingerprint as the model-predicted response to external forcing. We calculate this as follows:
For each ALL+8.5 model m, year y, and latitude ϕ, we calculate the amplitude Rm(ϕ, y) and phase θm(ϕ, y).
For each model and latitude, we calculate 1979–2014 amplitude and phase climatologies.
We subtract these climatologies to get amplitude and phase anomalies for each latitude and model.
Both amplitude and phase time series are weighted by the annual cycle weighting function W(m, ϕ).
We then average over ensembles and models to obtain MMM amplitude and phase anomaly time series. The averaging process damps internal variability, revealing the forced response.
The amplitude and phase fingerprints are defined as the leading empirical orthogonal function of the 1860–2100 MMM amplitude and phase anomalies. These fingerprints, which will be discussed further in section 2, are similar to the linear trend at each latitude (not shown).
2) Signal
Given a set of amplitude or phase anomalies from observations 
3) Noise
To assess the significance of a signal, we need an understanding of how internal climate variability might project onto the fingerprint. To do this, we calculate weighted amplitude and phase anomalies for the CMIP5 preindustrial control runs. By concatenating the first 200 yr of each piControl simulation and projecting the results onto the fingerprints, we obtain a long (4000 yr) time series pc(y) containing 4000/L nonoverlapping L-length segments. Calculating the best-fit trends in these segments results in a distribution whose standard deviation N(L) provides a measure of noise in the trends: the likelihood of observing a given signal purely due to natural internal variability.
4) Signal to noise
If the dimensionless signal-to-noise ratio (SNR) exceeds 1.64, the signal is considered detectable at 90% confidence relative to our best current multimodel estimates of natural internal variability (Bindoff et al. 2013). We use this two-tailed z test throughout to provide a conservative estimate of significance. Similarly, if the signal-to-noise ratio lies within the 5%–95% confidence interval obtained from CMIP5 ALL+8.5 simulations, the signal is considered compatible with the combination of external forcings present in those simulations. Greenhouse gas forcing is dominant over the satellite era, but these simulations also contain estimates of natural (solar and volcanic) and other anthropogenic (aerosols and ozone depletion) forcings. This means that compatibility with the ALL+8.5 distribution does not imply attribution of the signal to any particular forcing. We note that an observed signal can be compatible with external forcing but too weak to emerge from the background of internal variability, or it can be detectable but not compatible with external forcing (if it is incompatible with both unforced and forced model distributions). The latter case is cause for concern and results from some combination of the following: model failure to realistically capture the forced response, uncertainties in the forcings themselves, model underestimation of internal variability, or observational uncertainties.
f. Results
Modeled and observed amplitude and phase changes
In this paper, we argue that the annual cycle of precipitation provides a useful lens through which to examine current and predicted hydroclimate changes. As an additional benefit, this framework also reduces observational uncertainty due to errors in the long-term satellite datasets. Figure 2a shows 1979–2014 observed and simulated P climatologies. Figures 2b–d show 1979–2014 trends in annual mean P (Fig. 2b), P annual cycle amplitude (Fig. 2c), and P annual cycle phase (Fig. 2d). While CMAP and GPCP broadly agree on the overall P climatology (the spatial correlation between the datasets is greater than 0.96 for the 50°S–50°N domain considered here), they show very different trends in annual mean P (spatial correlation −0.15). By contrast, the spatial correlation between GPCP and CMAP weighted amplitude trends is 0.77 and between weighted phase trends is 0.68. This is perhaps because factors (such as satellite drift) that lead to spurious long-term trends in a satellite dataset are less likely to affect measures based on seasonal differences if both wet and dry seasons are affected by the same biases.
(a) Climatologies, (b) 1979–2014 trends in annual mean P, (c) 1979–2014 trends in P annual cycle amplitude, and (d)1979–2014 trends in P annual cycle phase. Orange shading shows the 5%–95% confidence interval determined from ALL+8.5 CMIP5 models; blue (CMAP) and red (GPCP) lines show observed values.
Citation: Journal of Climate 30, 13; 10.1175/JCLI-D-16-0572.1
The orange envelopes in Fig. 2 show the 5%–95% confidence intervals of these variables obtained from ALL+8.5 CMIP5 model ensembles. Figure 2a shows the severity of the double-ITCZ bias in the models but overall good agreement in the northern tropics and in the midlatitudes. Figure 2b shows that simulated trends in annual mean precipitation are more consistent with GPCP than CMAP, as the latter lies within the model envelope only in the dry subtropics. Figures 2c,d indicate that, while the observed amplitude and phase trends lie within the model envelopes at most latitudes, there are notable areas of model–observation disagreement: observed amplitude trends in the tropics are more negative than in models, and observed phase trends are more positive than in the models at certain subtropical latitudes. To assess the significance of these discrepancies, understand the physical mechanisms that dictate the model trends, and evaluate the performance of climate models over the satellite era, we require a formal comparison using the framework of detection and attribution.
g. Model-derived fingerprints
Figure 3a shows the fingerprint of externally forced changes to the amplitude of the precipitation annual cycle. We note that prior to calculating the multimodel mean, the amplitude time series has been weighted by W(m, ϕ) (Fig. 1b). This pattern explains 89% of the variance in the 1860–2100 MMM amplitude time series. Corroborating previous findings (Chou et al. 2013), the annual cycle amplitude increases almost everywhere, with the largest anomalies in the deep tropics where the base state precipitation is also large. The exceptions to this overall increase are on the equatorward flanks of the storm tracks in both hemispheres. Here, the decrease in amplitude likely reflects the predicted poleward shift of the storm tracks and subsequent reduction in wintertime precipitation on their equatorward edges (Hu and Fu 2007; Previdi and Liepert 2007; Scheff and Frierson 2012). Figure 3b shows the fingerprint of externally forced changes to the phase of the precipitation annual cycle, weighted by W(m, ϕ). This mode explains 72% of the variance in the 1860–2100 MMM phase time series. The fingerprint is characterized by little change or phase delays (later onset of the wet season) at most latitudes, with the exception of advances (earlier wet season onset) on the equatorward flanks of the storm tracks, particularly in the Southern Hemisphere. The phase delays are strongest around 15°N (at the longitude of the African Sahel) and around 45°N and 45°S (the poleward flank of the storm tracks). These results are consistent with the prediction that climate change will result in shorter, more-intense wet seasons in many monsoon regions and hence delay the monsoon onset (Biasutti and Sobel 2009; Biasutti 2013; Seth et al. 2011, 2013). Earlier wet season onsets poleward of 30°N and 30°S (red circles) are likely to be associated with forced circulation shifts that push the storm tracks toward the poles (Scheff and Frierson 2012; Marvel and Bonfils 2013).
Fingerprints of externally forced changes to the P annual cycle (a) amplitude and (b) phase and La Niña composites for the (c) amplitude and (d) phase. The y axis shows the EOF loading in (a),(b), precipitation anomaly (mm day−1) in (c), and phase anomaly (days) in (d).
Citation: Journal of Climate 30, 13; 10.1175/JCLI-D-16-0572.1
h. La Niña composites
Recent studies (e.g., Fyfe and Gillett 2014) have suggested that the global mean temperature trend from circa 1998–2014 may be smaller than predicted by CMIP5 models. This so-called global warming hiatus has been characterized by protracted La Niña–like cooling of the eastern equatorial Pacific (Kosaka and Xie 2013), stronger-than-normal trade winds (England et al. 2014), and distinctive precipitation and SLP patterns (Trenberth et al. 2014). This pattern has been deemed consistent with the IPO (Deser et al. 2004; Kosaka and Xie 2013; Meehl et al. 2013). Naturally occurring La Niña conditions affect the amplitude and phase of zonal mean precipitation. To characterize these effects, we compute La Niña composite patterns as follows: For every 200-yr piControl simulation, we calculate boreal winter (DJF) mean sea surface temperatures averaged over the Niño-3.4 region. We identify years in which this quantity is more than one standard deviation below the mean and designate these as La Niña years. We then calculate amplitude and phase anomalies relative to the piControl temporal mean amplitude and phase at each latitude. Averaging these anomalies over all La Niña years and over all models yields the amplitude and phase composite patterns shown in Figs. 3c and 3d, respectively. The amplitude signal is mostly restricted to the equatorial region and the northern tropics, where it indicates a reduction of the annual cycle amplitude, consistent with a reduction of austral summer and increase in boreal winter zonal mean tropical precipitation during the dry season (Lu et al. 2008). Amplitude weakly increases in the southern subtropics, and changes are negligible elsewhere. Phase changes associated with La Niña, by contrast, are not confined to the tropics. They are positive (earlier) in the latitudes of the southern ITCZ and of the northern monsoons and negative (delayed) in a narrow latitudinal band on the poleward flank of the storm tracks and may be associated with the poleward shifts of the Hadley cell and jets during a La Niña event (Lu et al. 2008).
i. Signal detection results
Figure 4a shows the projections of GPCP and CMAP zonal mean data onto the amplitude fingerprint. The interannual variability is similar in both datasets, and the overall trend is negative. Evidently the similarity between the searched-for fingerprint and the observations has decreased over the satellite era. By contrast, Fig. 4b shows the observed projections onto the phase fingerprint. Here, both GPCP and CMAP increasingly resemble the model-derived fingerprint over time. To assess the significance of these trends, we compare them to the distribution of 36-yr unforced trends that result from the projection of the preindustrial control data onto the fingerprint. We also calculate 1979–2014 trends in the individual ALL+8.5 ensemble members. Figure 5a shows the unforced and ALL+8.5 amplitude signal distributions along with the GPCP and CMAP amplitude signals and the regression slope errors. All have been normalized by the standard deviation of the noise distributions. The ALL+8.5 distribution is shifted to the right of the preindustrial control distribution, and a signal of external forcing is detectable at 90% confidence in approximately 30% of the ALL+8.5 1979–2014 simulations. This, however, is not reflected in the observations. The CMAP signal is detectable at 90% confidence (SNR = −1.8 ± 1.1): it lies in the far-left tail of the noise distribution (Fig. 5a, green bars; see also Table 1). The GPCP signal is negative but not detectable (SNR = −0.2 ± 1.0). However, both GPCP and CMAP amplitude signals are incompatible with the 5%–95% range of the ALL+8.5 distributions, indicating that the recent change in observed precipitation annual cycle amplitude is fundamentally different from the expected responses predicted by the CMIP5 models over the observational period. Figure 5b shows the noise, forced model, and observed phase signals. Again, the ALL+8.5 model distribution is shifted to the right, and the signal is detectable at 90% confidence in 21% of these simulations. Both GPCP (SNR = 2.2 ± 0.8) and CMAP (SNR = 1.3 ± 0.9) show positive signals, but only GPCP is detectable at 90% confidence. The GPCP signal is also consistent with forced CMIP5 models: it lies within the 5%–95% interval calculated from CMIP5 ALL+8.5 distributions (orange bars).
Projections of the observations onto the (a) amplitude and (b) phase fingerprints and best-fit lines obtained through linear OLS regression.
Citation: Journal of Climate 30, 13; 10.1175/JCLI-D-16-0572.1
Signal distributions for the P annual cycle (a) amplitude and (b) phase. All signals have been normalized by the standard deviation of the preindustrial control distribution (green). The orange histograms show the CMIP5 ALL+8.5 forced distributions. The observed signals are shown as vertical lines, and the shading shows the 90% confidence interval on the signals determined from the uncertainty in the regression slope.
Citation: Journal of Climate 30, 13; 10.1175/JCLI-D-16-0572.1
SNRs for the P annual cycle amplitude and phase for various subdomains. Values outside the 5%–95% confidence interval of the CMIP5 preindustrial control distribution are considered detectable and are marked with an asterisk. Values outside the 5%–95% confidence interval of the CMIP5 ALL+8.5 distribution are considered incompatible with the forced response and are marked with a dagger.
1) A protracted La Niña?
Our detection and attribution analysis has revealed that observations over the satellite era increasingly resemble the expected phase fingerprint but are increasingly dissimilar to the expected amplitude fingerprint (in fact, they increasingly resemble the inverse of the expected pattern). To what extent may these behaviors be explained by increasing similarity to a La Niña–like mode of variability?
To gauge the resemblance between the observations and La Niña, we project the observations onto the La Niña composites derived from the piControl simulations (Figs. 3c,d). Figure 6a shows the projection onto the La Niña composite amplitude pattern; Fig. 6b shows the projection onto the composite phase pattern. Both datasets indicate upward trends in both variables, indicating increasing similarity to the La Niña–like patterns.
Projections of the observations onto the La Niña composite (a) amplitude and (b) phase patterns. Also shown are the relationships between observed projections onto the (c) amplitude fingerprint and La Niña composites and (d) phase fingerprint and La Niña composites.
Citation: Journal of Climate 30, 13; 10.1175/JCLI-D-16-0572.1
The relationships between the observed projections onto the fingerprint (Figs. 4a,b) and onto the La Niña composites (Figs. 6a,b) are shown by the scatterplots in Figs. 6c,d. In both CMAP and GPCP, over 90% of the temporal variance in the projection onto the amplitude fingerprint is explained by the projection onto the La Niña composite. This suggests that the observations look less like the model-predicted anthropogenic forcing fingerprint because they look more like a La Niña–type mode of variability. This aligns with previous studies that have characterized the recent global mean temperature slowdown or hiatus as compatible with sustained La Niña–like conditions. We note that this analysis does not address the source of this model–observation mismatch but merely confirms that the observed trends in P amplitude increasingly resemble La Niña.
Figure 6d indicates a very weak relationship between the observed projections onto the phase fingerprint and projections onto the phase La Niña composite. Less than 5% of the variance in phase projection onto the fingerprint is explained by the projection onto La Niña. In this case, similarity to the fingerprint does not appear to result from similarity to a La Niña–like mode of variability.
This apparent discrepancy can be understood by comparing the spatial structure of the amplitude and phase fingerprints (Figs. 3a,b) to the spatial structure of the observed changes (Figs. 2c,d). The observed amplitude changes are large in the tropics, where both the fingerprint and La Niña pattern show large changes (of opposite sign). The observed phase changes, by contrast, are largest outside the tropics, as is the fingerprint. If we exclude latitudes between 10°S and 10°N from the analysis (Table 1), the amplitude signal-to-noise ratios in both GPCP and CMAP are positive (Table 1), and a much smaller percentage (55% for GPCP, 27% for CMAP) of the signal variance is explained by similarity to La Niña. This suggests a hint of forced changes to the amplitude in the extratropics, and Fig. 2c indeed shows that the observed amplitude increases in the subtropics and decreases on the equatorward flanks of the storm tracks, as in the fingerprint. The phase signal-to-noise ratios in the extratropics are also positive (Table 1), and only 10% (GPCP) and 16% (CMAP) of the signal variance is explained by similarity to La Niña. If, by contrast, we consider only the deep tropics (latitudes between 10°S and 10°N, Table 1), the amplitude SNR for GPCP and CMAP are strongly negative. The tropical amplitude trends strongly resemble protracted La Niña conditions, with 98% and 85% of the signal variance explained by projection onto the La Niña composites. The tropics-only phase signals are positive but undetectable above internal variability and largely explained by projection onto La Niña (68% for CMAP, 73% for GPCP).
2) Placing annual cycle D&A results in context
Observed changes to the amplitude of the annual cycle are large in the tropics, do not resemble the fingerprint of external forcing, and are not compatible with model-simulated internal variability at 90% confidence. However, they do resemble the La Niña composite pattern. This suggests that differences between modeled and observed trends during the warming slowdown are not confined to temperature changes. Instead, a signature of the warming hiatus is also expressed in zonal mean precipitation and is congruent with other studies identifying differences between modeled and observed trends in trade wind anomalies (England et al. 2014) and Pacific SSTs (Kosaka and Xie 2013). In other words, while the models and observations do not agree on changes to the precipitation amplitude, this mismatch is consistent with a wide and increasing body of literature on the warming hiatus.
Observed changes to the phase of the annual cycle, by contrast, are not dominated by the tropics and may suggest the emergence of a forced signal. Thus, we obtain categorically different results depending on whether we consider changes to the amplitude or phase of the annual cycle. This difference highlights an important consideration for detection and attribution studies. Precipitation is affected by multiple interlinked physical processes on multiple spatial and temporal scales, and the observations will always reflect some combination of externally forced response and internal climate variability. This means that results will necessarily depend on the analysis method and that a comprehensive understanding of observed and simulated precipitation changes must synthesize multiple analysis frameworks. For example, Marvel and Bonfils (2013) examined changes in zonal mean precipitation, attributing the observed combination of thermodynamic and dynamic precipitation changes to external forcing. Likewise, Polson et al. (2013a) identified an externally forced signal in the contrast between wet and dry regions but allowed the location of such regions to change from season to season and year to year. Using this method, observed drying trends, for example, may result from changes in the location of the dry regions, reductions in water vapor availability, or some combination of the two. Our results are consistent with the conclusions of Lintner et al. (2012), who attributed the mismatch between CMIP3 projections and recent changes in the contrast between wet and dry occurrences to ENSO-related variability.
3. Conclusions
Changes to the amplitude of the precipitation annual cycle have implications for hydrological extremes like droughts and floods (Chou and Lan 2012); changes to the phase can affect growing seasons and crop yields (Sultan et al. 2014). We show that, in addition to their clear societal importance, trends in these quantities are similar in two long-term satellite datasets. A focus on seasonality changes rather than trends in the annual mean therefore reduces observational uncertainty and enhances confidence in the results.
The zonal mean fingerprints used in this study identify changes that are both thermodynamic and dynamic in origin. Increases in the amplitude at most latitudes likely reflect the rich-get-richer mechanism that is prominent over the oceans. The decreases of the amplitude in the subtropics may result from the (dynamical) poleward expansion of the Hadley circulation. Changes in the annual cycle phase are similarly affected by the availability of water vapor [possibly through a temporal variant of the upped-ante mechanism of Neelin et al. (2003)] and changes in atmospheric circulation that affect the position and movement of the storm tracks and ITCZ.
We expect some combination of forced change and the recent climate hiatus to manifest in the observations, but responses to external forcing can be obscured by internal variability. Contrary to previous work, we find no evidence of an observed increase in the range between wet and dry season precipitation that could be attributed to anthropogenic forcing. In fact, our results show that observed changes to the amplitude of the precipitation annual cycle are nearly opposite to those projected by CMIP5 models. These signals are incompatible with CMIP5 forced trends and (in one dataset) incompatible with internal variability. This is largely due to large reductions of the amplitude in the tropics that resemble protracted La Niña–like (or IPO) conditions. Observed zonal mean phase changes, by contrast, are largely characterized by earlier onset of the wet season on the equatorward flanks of the extratropical storm tracks and later onset and little change at most other latitudes. In this sense, they increasingly resemble the fingerprint of externally forced change. This increasing resemblance is not explainable solely by protracted La Niña conditions but rather resembles the change projected by externally forced CMIP5 models. Thus it is the particular latitudinal amplitude and phase dependence of the annual hydrological cycle that exposes the mix of internal variability and external, anthropogenic, and natural forcing that has shaped precipitation seasonality over the satellite era.
Acknowledgments
This research was funded by DOE Grant DE-SC0014423. We acknowledge the World Climate Research Programme’s Working Group on Coupled Modelling, which is responsible for CMIP, and we thank the climate modeling groups for producing and making available their model output. For CMIP the U.S. Department of Energy’s Program for Climate Model Diagnosis and Intercomparison provides coordinating support and led development of software infrastructure in partnership with the Global Organization for Earth System Science Portals. GPCP and CMAP precipitation data are provided by the NOAA/OAR/ESRL PSD, Boulder, Colorado, from their website (http://www.esrl.noaa.gov/psd/). BC was supported by the NASA Modeling, Analysis, and Prediction program under Grant NNX14AB99A.
REFERENCES
Allen, M. R., and W. J. Ingram, 2002: Constraints on future changes in climate and the hydrologic cycle. Nature, 419, 224–232, doi:10.1038/nature01092.
Allen, M. R., and P. Stott, 2003: Estimating signal amplitudes in optimal fingerprinting, part I: Theory. Climate Dyn., 21, 477–491, doi:10.1007/s00382-003-0313-9.
Biasutti, M., 2013: Forced Sahel rainfall trends in the CMIP5 archive. J. Geophys. Res. Atmos., 118, 1613–1623, doi:10.1002/jgrd.50206.
Biasutti, M., and A. H. Sobel, 2009: Delayed Sahel rainfall and global seasonal cycle in a warmer climate. Geophys. Res. Lett., 36, L23707, doi:10.1029/2009GL041303.
Bindoff, N. L., and Coauthors, 2013: Detection and attribution of climate change: From global to regional. Climate Change 2013: The Physical Science Basis, T. F. Stocker et al., Eds., Cambridge University Press, 867–952.
Bony, S., G. Bellon, D. Klocke, S. Sherwood, S. Fermepin, and S. Denvil, 2013: Robust direct effect of carbon dioxide on tropical circulation and regional precipitation. Nat. Geosci., 6, 447–451, doi:10.1038/ngeo1799.
Chou, C., and C.-W. Lan, 2012: Changes in the annual range of precipitation under global warming. J. Climate, 25, 222–235, doi:10.1175/JCLI-D-11-00097.1.
Chou, C., J. D. Neelin, C.-A. Chen, and J.-Y. Tu, 2009: Evaluating the rich-get-richer mechanism in tropical precipitation change under global warming. J. Climate, 22, 1982–2005, doi:10.1175/2008JCLI2471.1.
Chou, C., J. C. Chiang, C.-W. Lan, C.-H. Chung, Y.-C. Liao, and C.-J. Lee, 2013: Increase in the range between wet and dry season precipitation. Nat. Geosci., 6, 263–267, doi:10.1038/ngeo1744.
Cowtan, K., and R. G. Way, 2014: Coverage bias in the HadCRUT4 temperature series and its impact on recent temperature trends. Quart. J. Roy. Meteor. Soc., 140, 1935–1944, doi:10.1002/qj.2297.
Deser, C., A. S. Phillips, and J. W. Hurrell, 2004: Pacific interdecadal climate variability: Linkages between the tropics and the North Pacific during boreal winter since 1900. J. Climate, 17, 3109–3124, doi:10.1175/1520-0442(2004)017<3109:PICVLB>2.0.CO;2.
Durack, P. J., S. E. Wijffels, and R. J. Matear, 2012: Ocean salinities reveal strong global water cycle intensification during 1950 to 2000. Science, 336, 455–458, doi:10.1126/science.1212222.
Dwyer, J. G., M. Biasutti, and A. H. Sobel, 2014: The effect of greenhouse gas–induced changes in SST on the annual cycle of zonal mean tropical precipitation. J. Climate, 27, 4544–4565, doi:10.1175/JCLI-D-13-00216.1.
England, M. H., and Coauthors, 2014: Recent intensification of wind-driven circulation in the Pacific and the ongoing warming hiatus. Nat. Climate Change, 4, 222–227, doi:10.1038/nclimate2106.
Fyfe, J. C., and N. P. Gillett, 2014: Recent observed and simulated warming. Nat. Climate Change, 4, 150–151, doi:10.1038/nclimate2111.
Fyfe, J. C., and Coauthors, 2016: Making sense of the early-2000s warming slowdown. Nat. Climate Change, 6, 224–228, doi:10.1038/nclimate2938.
Gillett, N., F. Zwiers, A. Weaver, G. Hegerl, M. Allen, and P. Stott, 2002: Detecting anthropogenic influence with a multi-model ensemble. Geophys. Res. Lett., 29, 1970, doi:10.1029/2002GL015836.
Hasselmann, K., 1993: Optimal fingerprints for the detection of time-dependent climate change. J. Climate, 6, 1957–1971, doi:10.1175/1520-0442(1993)006<1957:OFFTDO>2.0.CO;2.
Hawkins, E., T. Edwards, and D. McNeall, 2014: Pause for thought. Nat. Climate Change, 4, 154–156, doi:10.1038/nclimate2150.
Hegerl, G. C., H. von Storch, K. Hasselmann, B. D. Santer, U. Cubasch, and P. D. Jones, 1996: Detecting greenhouse-gas-induced climate change with an optimal fingerprint method. J. Climate, 9, 2281–2306, doi:10.1175/1520-0442(1996)009<2281:DGGICC>2.0.CO;2.
Hegerl, G. C., and Coauthors, 2015: Challenges in quantifying changes in the global water cycle. Bull. Amer. Meteor. Soc., 96, 1097–1115, doi:10.1175/BAMS-D-13-00212.1.
Held, I. M., 2013: Climate science: The cause of the pause. Nature, 501, 318–319, doi:10.1038/501318a.
Held, I. M., and B. J. Soden, 2006: Robust responses of the hydrological cycle to global warming. J. Climate, 19, 5686–5699, doi:10.1175/JCLI3990.1.
Hu, Y., and Q. Fu, 2007: Observed poleward expansion of the Hadley circulation since 1979. Atmos. Chem. Phys., 7, 5229–5236, doi:10.5194/acp-7-5229-2007.
Huber, M., and R. Knutti, 2014: Natural variability, radiative forcing and climate response in the recent hiatus reconciled. Nat. Geosci., 7, 651–656, doi:10.1038/ngeo2228.
Huffman, G. J., and Coauthors, 1997: The Global Precipitation Climatology Project (GPCP) combined precipitation dataset. Bull. Amer. Meteor. Soc., 78, 5–20, doi:10.1175/1520-0477(1997)078<0005:TGPCPG>2.0.CO;2.
Johansson, D. J., B. C. O. Neill, C. Tebaldi, and O. Häggström, 2015: Equilibrium climate sensitivity in light of observations over the warming hiatus. Nat. Climate Change, 5, 449–453, doi:10.1038/nclimate2573.
Karl, T. R., and Coauthors, 2015: Possible artifacts of data biases in the recent global surface warming hiatus. Science, 348, 1469–1472, doi:10.1126/science.aaa5632.
Kaufmann, R. K., H. Kauppi, M. L. Mann, and J. H. Stock, 2011: Reconciling anthropogenic climate change with observed temperature 1998–2008. Proc. Natl. Acad. Sci. USA, 108, 11 790–11 793, doi:10.1073/pnas.1102467108.
Kosaka, Y., and S.-P. Xie, 2013: Recent global-warming hiatus tied to equatorial Pacific surface cooling. Nature, 501, 403–407, doi:10.1038/nature12534.
Levy, A. A., W. Ingram, M. Jenkinson, C. Huntingford, F. Hugo Lambert, and M. Allen, 2013: Can correcting feature location in simulated mean climate improve agreement on projected changes? Geophys. Res. Lett., 40, 354–358, doi:10.1002/2012GL053964.
Lintner, B. R., M. Biasutti, N. S. Diffenbaugh, J.-E. Lee, M. J. Niznik, and K. L. Findell, 2012: Amplification of wet and dry month occurrence over tropical land regions in response to global warming. J. Geophys. Res., 117, D11106, doi:10.1029/2012JD017499.
Lu, J., G. Chen, and D. M. Frierson, 2008: Response of the zonal mean atmospheric circulation to El Niño versus global warming. J. Climate, 21, 5835–5851, doi:10.1175/2008JCLI2200.1.
Marvel, K., and C. Bonfils, 2013: Identifying external influences on global precipitation. Proc. Natl. Acad. Sci. USA, 110, 19 301–19 306, doi:10.1073/pnas.1314382110.
Meehl, G. A., A. Hu, J. M. Arblaster, J. Fasullo, and K. E. Trenberth, 2013: Externally forced and internally generated decadal climate variability associated with the interdecadal Pacific oscillation. J. Climate, 26, 7298–7310, doi:10.1175/JCLI-D-12-00548.1.
Min, S.-K., X. Zhang, and F. Zwiers, 2008: Human-induced Arctic moistening. Science, 320, 518–520, doi:10.1126/science.1153468.
Neelin, J., C. Chou, and H. Su, 2003: Tropical drought regions in global warming and El Niño teleconnections. Geophys. Res. Lett., 30, 2275, doi:10.1029/2003GL018625.
Polson, D., G. C. Hegerl, R. Allan, and B. B. Sarojini, 2013a: Have greenhouse gases intensified the contrast between wet and dry regions? Geophys. Res. Lett., 40, 4783–4787, doi:10.1002/grl.50923.
Polson, D., G. C. Hegerl, X. Zhang, and T. J. Osborn, 2013b: Causes of robust seasonal land precipitation changes. J. Climate, 26, 6679–6697, doi:10.1175/JCLI-D-12-00474.1.
Previdi, M., and B. G. Liepert, 2007: Annular modes and Hadley cell expansion under global warming. Geophys. Res. Lett., 34, L22701, doi:10.1029/2007GL031243.
Santer, B. D., and Coauthors, 2005: Amplification of surface temperature trends and variability in the tropical atmosphere. Science, 309, 1551–1556, doi:10.1126/science.1114867.
Santer, B. D., and Coauthors, 2007: Identification of human-induced changes in atmospheric moisture content. Proc. Natl. Acad. Sci. USA, 104, 15 248–15 253, doi:10.1073/pnas.0702872104.
Santer, B. D., and Coauthors, 2011: Separating signal and noise in atmospheric temperature changes: The importance of timescale. J. Geophys. Res., 116, D22105, doi:10.1029/2011JD016263.
Santer, B. D., and Coauthors, 2013: Identifying human influences on atmospheric temperature. Proc. Natl. Acad. Sci. USA, 110, 26–33, doi:10.1073/pnas.1210514109.
Santer, B. D., and Coauthors, 2014: Volcanic contribution to decadal changes in tropospheric temperature. Nat. Geosci., 7, 185–189, doi:10.1038/ngeo2098.
Sarojini, B. B., P. A. Stott, and E. Black, 2016: Detection and attribution of human influence on regional precipitation. Nat. Climate Change, 6, 669–675, doi:10.1038/nclimate2976.
Scheff, J., and D. M. Frierson, 2012: Robust future precipitation declines in CMIP5 largely reflect the poleward expansion of model subtropical dry zones. Geophys. Res. Lett., 39, L18704, doi:10.1029/2012GL052910.
Schmidt, G. A., D. T. Shindell, and K. Tsigaridis, 2014: Reconciling warming trends. Nat. Geosci., 7, 158–160, doi:10.1038/ngeo2105.
Schneider, T., T. Bischoff, and G. H. Haug, 2014: Migrations and dynamics of the intertropical convergence zone. Nature, 513, 45–53, doi:10.1038/nature13636.
Seager, R., N. Naik, and G. A. Vecchi, 2010: Thermodynamic and dynamic mechanisms for large-scale changes in the hydrological cycle in response to global warming. J. Climate, 23, 4651–4668, doi:10.1175/2010JCLI3655.1.
Seth, A., S. A. Rauscher, M. Rojas, A. Giannini, and S. J. Camargo, 2011: Enhanced spring convective barrier for monsoons in a warmer world? Climatic Change, 104, 403–414, doi:10.1007/s10584-010-9973-8.
Seth, A., S. A. Rauscher, M. Biasutti, A. Giannini, S. J. Camargo, and M. Rojas, 2013: CMIP5 projected changes in the annual cycle of precipitation in monsoon regions. J. Climate, 26, 7328–7351, doi:10.1175/JCLI-D-12-00726.1.
Steinman, B. A., M. E. Mann, and S. K. Miller, 2015: Atlantic and Pacific multidecadal oscillations and Northern Hemisphere temperatures. Science, 347, 988–991, doi:10.1126/science.1257856.
Stine, A., P. Huybers, and I. Fung, 2009: Changes in the phase of the annual cycle of surface temperature. Nature, 457, 435–440, doi:10.1038/nature07675.
Stott, P. A., S. Tett, G. Jones, M. Allen, J. Mitchell, and G. Jenkins, 2000: External control of 20th century temperature by natural and anthropogenic forcings. Science, 290, 2133–2137, doi:10.1126/science.290.5499.2133.
Sultan, B., K. Guan, M. Kouressy, M. Biasutti, C. Piani, G. Hammer, G. McLean, and D. Lobell, 2014: Robust features of future climate change impacts on sorghum yields in West Africa. Environ. Res. Lett., 9, 104006, doi:10.1088/1748-9326/9/10/104006.
Tan, P.-H., C. Chou, and J.-Y. Tu, 2008: Mechanisms of global warming impacts on robustness of tropical precipitation asymmetry. J. Climate, 21, 5585–5602, doi:10.1175/2008JCLI2154.1.
Taylor, K. E., R. J. Stouffer, and G. A. Meehl, 2012: An overview of CMIP5 and the experiment design. Bull. Amer. Meteor. Soc., 93, 485–498, doi:10.1175/BAMS-D-11-00094.1.
Tett, S. F., and Coauthors, 2002: Estimation of natural and anthropogenic contributions to twentieth century temperature change. J. Geophys. Res., 107, 4306, doi:10.1029/2000JD000028.
Trenberth, K. E., J. T. Fasullo, G. Branstator, and A. S. Phillips, 2014: Seasonal aspects of the recent pause in surface warming. Nat. Climate Change, 4, 911–916, doi:10.1038/nclimate2341.
Van Vuuren, D. P., and Coauthors, 2011: The representative concentration pathways: An overview. Climatic Change, 109, 5–31, doi:10.1007/s10584-011-0148-z.
Wentz, F. J., and R. W. Spencer, 1998: SSM/I rain retrievals within a unified all-weather ocean algorithm. J. Atmos. Sci., 55, 1613–1627, doi:10.1175/1520-0469(1998)055<1613:SIRRWA>2.0.CO;2.
Xie, P., and P. A. Arkin, 1997: Global precipitation: A 17-year monthly analysis based on gauge observations, satellite estimates, and numerical model outputs. Bull. Amer. Meteor. Soc., 78, 2539, doi:10.1175/1520-0477(1997)078<2539:GPAYMA>2.0.CO;2.
Yin, X., A. Gruber, and P. Arkin, 2004: Comparison of the GPCP and CMAP merged gauge–satellite monthly precipitation products for the period 1979–2001. J. Hydrometeor., 5, 1207–1222, doi:10.1175/JHM-392.1.
Zhang, X., F. W. Zwiers, G. C. Hegerl, F. H. Lambert, N. P. Gillett, S. Solomon, P. A. Stott, and T. Nozawa, 2007: Detection of human influence on twentieth-century precipitation trends. Nature, 448, 461–465, doi:10.1038/nature06025.