Understanding sudden changes in cloud amount: The Southern Annular Mode and South American weather fluctuations

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 117,, doi: /2012jd017626, 2012 Understanding sudden changes in cloud amount: The Southern Annular Mode and South American weather fluctuations Benjamin A. Laken 1,2 and Enric Pallé 1,2 Received 14 February 2012; revised 10 April 2012; accepted 25 May 2012; published 4 July [1] This work investigates the cause and effects of extreme changes in synoptic-scale cloud cover operating at daily timescales using a variety of satellite-based and reanalysis data sets. It is found that the largest sudden increases detected in globally averaged cloud cover over the last ten years of satellite-based observations occur following positively correlated shifts in the phase of the Southern Annular Mode (SAM) index. The associated pressure anomalies are found to generate frontal cloud formation over large areas of the South American continent, increasing regional cloud cover by up to 20%; these changes are correlated to statistically significant reductions in local temperatures of approximately 2.5 C with a +1 day time lag, indicating the SAM index is associated with large scale weather fluctuations over South America. Citation: Laken, B. A., and E. Pallé (2012), Understanding sudden changes in cloud amount: The Southern Annular Mode and South American weather fluctuations, J. Geophys. Res., 117,, doi: /2012jd Introduction [2] Due to the strong forcing exerted by clouds on Earth s radiation balance, even slight variations in cloud cover have the capacity to significantly influence Earth s climate [Ramanathan et al., 1989]. It is largely unknown how global cloud properties will respond to anthropogenic induced climate changes, and consequently it is unclear how future changes in global cloud properties will act to enhance or mitigate the effects of global warming; making clouds one of the largest sources of uncertainty in our understanding of climate change and a point of considerable debate [Cess et al., 1989; Clement et al., 2009; Dessler, 2010; Spencer and Braswell, 2010]. [3] Although global cloud cover has been recorded by satellite-based irradiance measurements for around 30 years, our understanding of this highly variable component of the atmosphere is still incomplete. This is partially due to the limitations of satellite cloud detections, which are restricted by their top-down view of cloud cover, and so have difficulty in reliably determining obscured lower clouds; consequently, cloud trends show strong artificial anti-correlated behaviors between high and low tropospheric levels, which make determining the nature of global cloud properties in recent decades highly problematic [Norris, 2001; Pallé, 2005; Evan et al., 2007]. Indeed comparisons between 1 Instituto de Astrofísica de Canarias, La Laguna, Spain. 2 Department of Astrophysics, Faculty of Physics, Universidad de La Laguna, La Laguna, Spain. Corresponding author: B. A. Laken, Instituto de Astrofísica de Canarias, Via Lactea s/n, E La Laguna, Tenerife, Spain. (blaken@iac.es) American Geophysical Union. All Rights Reserved /12/2012JD independent cloud measuring programs provide unclear and even conflicting results [Cess et al., 1996; Dai et al., 2006]. [4] In order to better understand the relationship between cloud changes and climate, we focus on an investigation on extreme global scale cloud changes at daily timescales and examine the preceding and following state of the atmosphere. We base this study on satellite-detected cloud data from During this time simultaneous measurements exist between two comparable, yet independent, global scale cloud monitoring programs: the International Satellite Cloud Climatology Project (ISCCP) [Rossow and Schiffer, 1991], and the more modern MODerate Resolution Imaging Spectroradiometer (MODIS) project [King et al., 1992; Platnick et al., 2003]. 2. Data Sets [5] The ISCCP and MODIS data sets both provide a global sampling of cloud cover over a long time period (currently for ISCCP and for MODIS). These programs provide estimates of many similar properties related to clouds, however they use different retrieval methods, and often arrive at different results. A brief description of these data sets will now be considered here; for a more complete comparison of these data sets, see Pincus et al. [2012] and references therein. [6] ISCCP data are constructed from intercalibrated radiance measurements taken at infrared (IR) (11 mm) and visible (VIS) (0.6 mm) wavelengths, from a network of geostationary and polar orbiting weather satellites. ISCCP determinations of cloudiness are based on a comparison of observed radiance measurements against estimates of clearsky values, wherein pixels identified as being both colder and brighter than clear-sky pixel values are flagged as cloudy. This work uses the D1 mean IR cloud amount product, which is 1of7

2 provided at 3 h time intervals over a resolution global grid. [7] The inclusion of geostationary satellites in ISCCP data gives it an advantage of achieving uniform and high temporal resolution sampling and should provide a good basis for calibration adjustments. However, this approach results in the inclusion of numerous artifacts associated with the geostationary satellite footprints; these result from the dependence of cloud detections on the satellite viewing angle, with cloud amount appearing to increase as viewing angle increases [Evan et al., 2007], leading to discontinuities in the data over regions of satellite overlap [Stordal et al., 2005]. Indeed, this problem has also likely resulted in the inclusion of spurious long-term trends in the ISCCP data, as over time the number of geostationary satellites providing measurements to the ISCCP data has increased, resulting in greater overlap between the geostationary satellite footprints, and thus a general decrease in viewing angle; as measurements from the limbs of the geostationary footprint regions are preferentially excluded as the overlap/coverage between the satellite footprints increases with the addition of new geostationary instruments. [8] MODIS observations are made from two polar orbiting satellites, one on board the Terra EOS platform (operational since 2000), and the second on board the Aqua EOS platform (operational since 2002). Together these provide near global measurements over an approximately 24 hr period. MODIS has a high spectral resolution, operating in 36 different bands, which enables the retrieval of cloud properties in a different manner to ISCCP. The higher spectral resolution also allows for the simultaneous retrievals of additional atmospheric properties such as particle size and thermodynamic phase. The MODIS data used in this work belongs to collection 5.1 version 3, and is provided globally at a 1 1 resolution. [9] Some key differences between the methods MODIS and ISCCP use to estimate cloud properties of relevance to this investigation are now briefly considered: (1) the amount of cloud retrieved at middle cloud levels. MODIS uses a more accurate technique for determining cloud top pressure than ISCCP known as CO 2 slicing [Menzel et al., 1983]. The method employed by ISCCP is prone to incorrectly interpreting pressure in certain cloud situations leading to an overassignment of middle level cloud cover. (2) ISCCP adjusts the nighttime IR cloud retrievals based on daytime (VIS + IR) retrievals, whereas MODIS gives only the simple daily mean cloud retrieval with no further adjustments applied. (3) The thresholds upon which regions of fractional cloud cover are determined to be cloudy/non-cloudy differ, although MODIS and ISCCP often produce similar results their cloud detections are biased in different manners. ISCCP thresholds were partly calibrated to offset over-estimates of cloud amounts resulting from low-spatial resolutions [Rossow and Garder, 1993]. Whereas, MODIS frequently excludes pixels near cloud edges of optically thin cloud [Pincus et al., 2012], while overestimating cloud fraction in areas of low-level sub-pixel scale cloud [Zhao and Di Girolamo, 2006]. 4) Interpretation of regions of uncertain cloud detections are treated differently: ISCCP includes these pixels as cloudy, whereas MODIS excludes such pixels from its final estimate of cloud amount. [10] In addition to the cloud data sets, this work also uses the Southern Annular Mode Index (hereafter referred to as the SAM Index), surface level (SL) air temperature changes, SL pressure and SL wind data. The SAM index is defined as the difference in the normalized daily zonal-mean sea level pressure between 40 S and 70 S[Nan and Li, 2003], and describes hemispheric-scale patterns of climate variability characterized by north south shifts in atmospheric mass between the polar and middle latitude regions [Thompson and Wallace, 2000]. The daily mean surface level (SL) air temperature, pressure and wind (meridional and zonal) data are sourced from the NCEP/NCAR reanalysis program [Kalnay et al., 1996]. 3. Methodology [11] Our experiment is designed to isolate extreme changes in cloud cover over short (daily) timescales. To accomplish this we selected the largest sudden increases in globally averaged, daily timescale cloud cover as the subjects of an epoch-superpositional (composite) study. To isolate extreme changes in daily timescale cloud we first calculated the daily cloud anomalies over the period, by subtracting the averaged cloud cover of each day in the sample period from an average of three days, beginning five days earlier. The gap between the averaging period and key day value is required to account for autocorrelation in the data set. The size of the averaging period and the time distance from the key date also constrain the type of cloud changes we are observing, restricting them to synoptic-scale changes. These changes likely originate from the waxing and waning of regional-scale cloud systems, minimizing the contribution from day-to-day low-level variability. The calculated daily anomalies were then ranked according to their magnitude, and the 0.95 percentile cloud increases identified and used as the key composite date. This methodology produced a composite with a sample size of n = 135, no seasonal bias was identified in this sample. [12] Throughout this work the statistical significance of our local (pixel) scale observations are evaluated by means of Monte Carlo (MC) techniques, using a resampling method to generate 1,000 randomized composites from the real data sets from which to evaluate significance. The statistical significance is evaluated at the 95th two-tailed percentile level. Additionally, where the data are presented as a composite time series, confidence intervals are given to both the 95th and 99th two-tailed percentile level, based on 100,000 MC simulations. For the time series data, the values are given as anomalies, where each data point is a daily mean subtracted from a 21-day running average (centered on the day in question). 4. Results [13] The globally averaged cloud cover increase identified by the composite sample is shown over a 40 day period in Figure 1. Cloud cover variations between the ISCCP and MODIS data sets show a close and statistically significant correlation over this period (r = 0.83, p = <0.01). However, ISCCP data detects a cloud increase of 1.76% while MODIS sees a smaller increase of 1.06% (increase value quoted is the key date anomaly). ISCCP and MODIS data both show high levels of statistical significance well above the 99th percentile (two-tailed) confidence intervals calculated 2of7

3 Figure 1. Composite of cloud increases: (a) ISCCP and (b) MODIS cloud cover anomalies (%), over a 40 day composite period surrounding a key date of the most extreme (>0.95 percentile) ISCCP detected cloud increases in daily timescale cloud (n = 135) from Solid line shows mean value, while dashed and dotted lines show 95th and 99th percentile confidence intervals respectively, based on 100,000 Monte Carlo simulations. from MC simulations; such high significance levels are as expected, as the composites reflect the most extreme increases in daily timescale cloud cover. [14] Figure 2 presents a local scale analysis of the statistically significant cloud cover anomalies on the key date from the ISCCP and MODIS data. We find that ISCCP detects large regions of significant cloud anomalies, whereas MODIS detects far less extensive locally significant changes. The correspondence between the areas identified by both instruments is good, with clearly recognizable features identified by both data sets. The agreement between the independent ISCCP and MODIS data sets enables us to confidently establish the validity of the cloud anomalies. Of most note are the cloud cover increases occurring over the South American region; these are the most extensive, and highest magnitude (approximately +20%) cloud anomalies identified in both data sets. [15] To attempt to understand where and why the ISCCP and MODIS data sets disagree on the significance of the cloud changes, we have analyzed the correlations between the data sets at an individual pixel level, over a 10 day period surrounding the key date; the results are displayed in Figure 3. It shows that although the MODIS data are not as widely significant as the ISCCP data, the data sets broadly agreed on the cloud changes that were occurring, suggesting that ISCCP observes the same cloud changes as MODIS, however these changes appear larger and more statistically significant: this is also indicated by comparing the magnitude of the globally averaged anomalous day 0 cloud cover changes in Figure 1. The agreement between the data sets was found to be best over low and middle latitudes, and tended to become weaker at progressively higher latitudes. [16] In particular, correlations between ISCCP and MODIS data were found to be weak over widespread areas of the Southern Oceans. The poorest correspondence was found at high latitudes, with the Antarctic and Arctic regions both showing areas of strong anticorrelation. Some of these regions contributed to the pixels identified as statistically significant cloud anomalies by ISCCP (Figure 2a). Such features are likely a result of the ISCCP detection algorithms mistaking temperature changes for clouds over high latitudes, i.e., a cloud decrease over an ice-covered surface results in a cooling, which the MODIS may correctly detect, but the ISCCP may mistake as an increase in cloud [Rossow and Schiffer, 1999], thereby resulting in the complete disagreement observed between the data sets over such regions. [17] The large magnitude cloud cover increases reliably identified over the South American region are the source of the majority of the detected global cloud increases. To understand the origin and development of these cloud cover anomalies we examine tropospheric properties immediately prior to and during the key date of the composite: surface level vector wind, and pressure are presented for days 5, 2 and 0 of the composite in Figure 4. These data show a strong shift in pressure over this period. High-pressure Figure 2. Locally statistically significant cloud cover anomalies (at the 0.95 two-tailed confidence level) for (a) ISCCP and (b) MODIS data sets. Composite samples based on the largest ( 0.95) percentile increases in globally averaged daily timescale cloud cover detected by ISCCP from (n = 135). 3of7

4 Figure 3. Local correlation coefficient (r) values obtained between a comparison of ISCCP and MODIS cloud amount over a 10 day period surrounding the key composite dates. anomalies (of approximately 7 mb) are observed to the southwest of the South American continent on day 5. Over a six-day period the high pressure is shifted in a northeasterly direction over the South American continent, and is replaced by low-pressure anomalies (of approximately 7 mb). As the pressure advances northeasterly increases in cloud cover are observed at the edges of the high pressure region. It is this frontal action that generates the observed South American cloud increase. [18] Anomalous surface level air temperature decreases (of as low as 2.5 C locally) are observed in association with the South American pressure/cloud cover changes (Figure 5a). The statistically significant maximal temperature reductions are found to lag behind the cloud cover increases by approximately one day (Figure 5b). Over the region of 15 S 40 S, 80 W 40 W, surface level air temperatures were reduced by 0.74 C between day 6 and +1, this change was found to be highly statistically significant (>99th percentile level). Figure 4. Temporal evolution of an strong increase in cloud cover over the South American region over observed on days (a and b) 5, (c and d) 2, and (e and f) 0 of the composite sample. Figures 4a, 4c, and 4e show surface level pressure anomalies, with vector wind anomalies overlaid. Figures 4b, 4d, and 4f show the co-temporal ISCCP detected cloud cover anomalies. 4of7

5 Figure 5. (a) South American air temperatures and cloud cover NCEP/NCAR reanalysis surface level (SL) air temperature anomalies ( C) on day 0 of the composite over the South American region. (b) A time series of regional SL air temperature anomalies (15 S 40 S, 80 W 40 W) over a 20 day composite period. The dashed and dotted lines indicating the 95th and 99th percentile confidence intervals respectively, calculated from 100,000 Monte Carlo simulations. Air temperature shows a statistically significant decline, reaching a maximal reduction one day after the cloud increase. [19] These observations suggests a chain of events whereby repetitive (northeastward) shifts in high pressure from high- Southern latitudes across the South American region generate frontal cloud, which in turn reduces surface level air temperatures across the region. However, it is unclear if the majority of the temperature reductions are due to advection of cold southerly air generated by the frontal activity (clearly observed in Figure 4), or the direct result of radiative changes from the cloud increase. 5. Discussion [20] The SAM index provides a measure of atmospheric variability across the Southern hemisphere, characterizing atmospheric variability not associated with the seasonal cycle. The state of the SAM is largely connected to Antarctic circulation: a strong (weak) Antarctic circumpolar vortex corresponds to a more positive (negative) SAM index value, as a strong (weak) circumpolar vortex induces widespread cooling (warming) over high latitude regions, via interactions with polar temperature inversions [van den Broeke and van Lipzig, 2004]. Positive (negative) shifts in the SAM are connected to north (south) shifts in high pressure between high and middle latitude regions. [21] Previous studies have noted a wide range of associations between climate fluctuations and the state of the SAM across the Southern Hemisphere. Evidence from a variety of sources including reanalysis, station observations, and climate models indicate that a positive SAM index is linked to significant temperature and precipitation changes, including: cool and wet conditions over large areas of Australia, and warm and dry conditions over New Zealand, Tasmania and South America [Cai and Watterson, 2002; Gillett et al., 2006; Sen Gupta and England, 2006]. [22] We find that during our composite, the SAM index shows statistically significant (>99th percentile) high magnitude correlated shifts one day prior to the significant cloud changes (Figure 6). This supports our previously outlined scenario, that northerly shifts in atmospheric mass from high to middle latitudes (as indicated by the SAM index) are responsible for the generation of the cloud cover increases, which in turn are associated with reductions in regional temperatures (either via a radiative cooling effect of the clouds, or via the southerly advection of cold air, or a combination of the both). 5of7

6 Figure 6. Globally averaged changes in MODIS-detected cloud cover (red line) compared to the Southern Annular mode (SAM) index anomalies (black line) over a 40 day composite period surrounding the key date of cloud increase. Dashed and dotted lines show the 95th and 99th percentile confidence intervals respectively for the SAM index anomalies, calculated from 100,000 Monte Carlo simulations. Highly statistically significant reductions in the SAM index are observed one day in advance of MODIS cloud changes. [23] Over the years we find that there is no significant trend in the number of extreme events per year, nor do we find any trend in the number of extreme daily timescale fluctuations in the SAM index, however, we note that it is difficult to draw any conclusions about trends over such a short time period. The SAM has demonstrated a significant trend toward the positive phase over recent decades [Marshall, 2003]. This may potentially be due to polar vortex intensification related to ozone driven photochemical reactions; although ozone emissions are stabilizing, it is predicted that polar vortex intensification will likely continue over the foreseeable future as a result of greenhouse gas emissions, and thus present-day SAM trends may also continue into the future [Arblaster and Meehl, 2006]. If so, it is possible that as the SAM shifts toward more positive values the occurrence of the atmospheric variations described in this work may become more frequent. [24] Acknowledgments. The authors acknowledge the NCEP Reanalysis Project data, provided by the NOAA/OAR/ERSL PSD, Boulder, Colorado, USA from the ISCCP D1 data, available from the ISCCP Web site at maintained by the ISCCP research group at the NASA Goddard Institute for Space Studies; the MODIS data, obtained from the NASA website and the SAM/NAM Index data available from NAM-SAM-NAO/SAM(Daily)-AAO.htm and data-nam-sam-nao/nam-ao.htm. The authors acknowledge support from the Spanish MICIIN, grant CGL References Arblaster, J. M., and G. A. Meehl (2006), Contributions of external forcings to Southern Annular Mode trends, J. Clim., 19(12), , doi: / JCLI Cai, W., and I. Watterson (2002), Modes of interannual variability of the Southern Hemisphere circulation simulated by the CSIRO model, J. Clim., 15(10), , doi: / (2002)015<1159:moivot> 2.0.CO;2. Cess, R. D., et al. (1989), Interpretation of cloud-climate feedbacks as produced by 14 atmospheric general circulation models, Science, 245(4917), , doi: /science Cess, R. D., M. H. Zhang, Y. Zhou, X. Jing, and V. Dvortsov (1996), Absorption of solar radiation by clouds: Interpretations of satellite, surface, and aircraft measurements, J. Geophys. Res., 101(D18), , doi:1029/96jd Clement, A. C., R. Burgman, and J. R. Norris (2009), Observational and model evidence for positive low-level cloud feedback, Science, 325(5939), , doi: /science Dai, A., T. Karl, B. Sun, and K. E. Trenbeth (2006), Recent trends in cloudiness over the United States: A tale of monitoring inadequacies, Bull. Am. Meteorol. Soc., 87(5), , doi: /bams Dessler, A. E. (2010), A determination of the cloud feedback from climate variations over the past decade, Science, 330(6010), , doi: / science Evan, A. T., A. K. Heidinger, and D. J. Vimont (2007), Arguments against a physical long-term trend in global ISCCP cloud amounts, Geophys. Res. Lett., 34, L04701, doi: /2006gl Gillett, N., T. Kell, and P. Jones (2006), Regional climate impacts of the Southern Annular Mode, Geophys. Res. Lett., 33, L23704, doi: / 2006GL Kalnay, E., et al. (1996), The NCEP/NCAR 40-Year Reanalysis Project, Bull. Am. Meteorol. Soc., 77(3), , doi: / (1996) 077<0437:TNYRP>2.0.CO;2. King, M., Y. Kaufman, W. Menzel, and D. Tanre (1992), Remote sensing of cloud, aerosol, and water vapour properties from the Moderate Resolution Imaging Spectrometer (MODIS), IEEE Trans. Geosci. Remote Sens., 30(1), 2 27, doi: / Marshall, G. J. (2003), Trends in the Southern Annular Mode from observations and reanalyses, J. Clim., 16(24), , doi: / (2003)016<4134:titsam>2.0.co;2. Menzel, W., W. Smith, and T. Stewart (1983), Improved cloud motion wind vector and altitude assignment using VAS, J. Appl. Meteorol., 22(3), , doi: / (1983)022<0377:icmwva>2.0.co;2. Nan, S., and J. Li (2003), The relationship between summer precipitation in the Yangtze River valley and the boreal sprint Southern Hemisphere Annular Mode, Geophys. Res. Lett., 30(24), 2266, doi: /2003gl Norris, J. (2001), What can cloud observations tell us about climate variability?, Space Sci. Rev., 94(1 2), , doi: /a: Pallé, E. (2005), Possible satellite perspective effects on the reported correlations between solar activity and clouds, Geophys. Res. Lett., 32, L03802, doi: /2004gl Pincus, R., S. Platnick, S. Ackerman, R. Hemler, and R. Hofmann (2012), Reconciling simulated and observed views of clouds: MODIS, ISCCP, and the limits of instrument simulators, J. Clim., doi: /jcli-d , in press. Platnick, S., M. King, S. Ackerman, W. Menzel, B. Baum, J. Riedi, and R. Frey (2003), The MODIS cloud products: Algorithms and examples from Terra, IEEE Trans. Geosci. Remote Sens., 41(2), , doi: /tgrs Ramanathan, V., R. Cess, E. Harrison, P. Minnis, B. Barkstrom, E. Ahmad, and D. Hartmann (1989), Cloud-radiative forcing and climate: Results from the Earth radiation budget experiment, Science, 243(4887), 57 63, doi: /science Rossow, W., and L. Garder (1993), Validation of ISCCP cloud detections, J. Clim., 6(12), , doi: / (1993)006<2370: VOICD>2.0.CO;2. Rossow, W., and R. Schiffer (1991), ISCCP cloud data products, Bull. Am. Meteorol. Soc., 72(1), 2 20, doi: / (1991)072<0002: ICDP>2.0.CO;2. Rossow, W., and R. Schiffer (1999), Advances in understanding clouds from ISCCP, Bull. Am. Meteorol. Soc., 80(11), , doi: / (1999)080<2261:aiucfi>2.0.co;2. Sen Gupta, A., and M. England (2006), Coupled ocean-atmosphere-ice response to variations in the Southern Annular Mode, J. Clim., 19(18), , doi: /jcli Spencer, R., and W. Braswell (2010), On the diagnosis of radiative feedback in the presence of unknown radiative forcing, J. Geophys. Res., 115, D16109, doi: /2009jd Stordal, F., G. Myhre, E. Stordal, W. Rossow, D. Lee, D. Arlander, and T. Svendby (2005), Is there a trend in cirrus cloud cover due to aircraft traffic?, Atmos. Chem. Phys., 5(8), , doi: /acp of7

7 Thompson, D., and J. Wallace (2000), Annular modes in the extratropical circulation. Part I: Month-to-month variability, J. Clim., 13(5), , doi: / (2000)013<1000:amitec>2.0.co;2. van den Broeke, M. R., and N. P. van Lipzig (2004), Changes in Antarctic temperature, wind and precipitation in response to the Antarctic Oscillation, Ann. Glaciol., 39, , doi: / Zhao, G., and L. Di Girolamo (2006), Cloud fraction errors for trade wind cumuli from EOS-Terra instruments, Geophys. Res. Lett., 33, L20802, doi: /2006gl of7