Portal de Periódicos da CAPES

Reply to comment by John R. Lanzante on “Trends in the temperature and water vapor content of the tropical lower stratosphere: Sea surface connection”

2009; American Geophysical Union; Volume: 114; Issue: D12 Linguagem: Inglês

10.1029/2008jd011265

2156-2202

K. H. Rosenlof, George C. Reid,

Atmospheric and Environmental Gas Dynamics

[1] In the comment by Lanzante [2009] (hereinafter referred to as L09) the paper by Rosenlof and Reid [2008] (hereinafter referred to as RR08) is criticized for using data from the Integrated Global Radiosonde Archive (IGRA) [Durre et al., 2006] without paying sufficient attention to the effects of changing instrumentation and data treatment at individual radiosonde stations. L09 refers to several “homogenized” radiosonde data sets in the literature that have attempted to take account of these problems, using the metadata records from the National Climatic Data Center and other temperature data obtained by remote sensing techniques, and he attempts to show that the unusual behavior of temperature variations in the tropopause and lower stratospheric regions of the tropical atmosphere since about 1990 is largely due to artificial discontinuities in the archived data. L09 points out that the mid-1990s coincided roughly with the time period in which several radiosonde stations changed their basic instrument types from VIZ to Vaisala, thereby introducing a possible discontinuity into the archived data. [2] We should first point out that all of the existing homogenized data sets apply only to the mandatory pressure levels of radiosonde analysis. A major theme of RR08 dealt with the tropopause and cold point regions, whose properties are not available from those data sets. Since cold point temperatures control the amount of water vapor entering the stratosphere, and since water vapor is an important contributor to the radiative balance of the stratosphere, as well as being the atmosphere's major greenhouse gas, it is clear that the homogenized data sets in their present form are not suitable for use in our study. [3] As a minor point, the “mandatory” levels of radiosonde analysis used by RR08 were 100 hPa and 70 hPa, while L09 shows data only from the 50-hPa level. While this may seem an unimportant difference, the 50- and 70-hPa levels often show significant differences, an outstanding example being illustrated in Figure 4 of RR08, in which the warming event in 2000 is strong at 70 hPa and weak at 50 hPa. The vertical stability of the stratosphere allows for the existence of layered structure of this kind. [4] As far as the radiosonde data are concerned, we do not deny that there have been instrumental changes, and these have potentially amplified the magnitude of stratospheric temperature trends, as noted by a number of authors. However, the absolute magnitude of the trend is actually not our interest here. The more important point is that the character of the temperature anomaly record changed. In Figure 2 of RR08, we note very little cooling in the lower stratosphere (and at the cold point, which is significant for the input of water vapor into the stratosphere) throughout the 1970s and 1980s, with a strong QBO-related variation, consistent with what would be expected in the tropics from ozone depletion and greenhouse gas changes. A more substantial tropical tropopause/lower stratosphere cooling started in the early 1990s, and has continued to the present. During the 1990s cooling period, there was a significant muting of the QBO signal, with a transient return during 1999–2000. There is a dramatic, step function like cooling event at the end of 2001 that is mirrored in changes in stratospheric water vapor as discussed in RR08. The actual temperature decrease associated with the 1990s cooling was about 3 K in the IGRA records at the 70-hPa level as opposed to about 1.5 K in the 50-hPa homogenized records, as illustrated in Figures 1 and 2 of L09. (Note that the Vaisala web site states that the accuracy for a current RS-92 radiosonde is on the order of 0.5°C, with a precision that varies from 0.2° to 0.5°C depending on altitude.) The difference between the 70-hPa IGRA record (3 K over the 1990s) and the 50-hPa homogenized record shown in L09 (1.5 K) may well be caused, at least in part, by instrumental and data-handling changes. The difference in pressure level may also be important, but in either case, the conclusions in RR08 are not affected. What we find is that there is little trend for a long period of time, then a period of increasing downward trend in the lower stratosphere. Figure 1 of L09 for the time series of 50-hPa temperature anomalies at Koror shows a sharp decrease at about 1993 that is sustained for the remainder of the record. In the compressed time scale of L09's Figure 1, the drop in 1993 looks suspiciously like a discontinuity, implying an instrument change. The IGRA record on a monthly time scale, however, shows that the temperature decrease at 70-hPa occurred monotonically over a period of about a year and is far from discontinuous. The same is true of all the warm pool stations we examined. [5] We did indeed examine metadata for the sondes of interest (from the IGRA Web page at http://www1.ncdc.noaa.gov/pub/data/igra/igra-metadata.txt). For Koror, the metadata is detailed in Table 1. [6] Figure 1 shows Figure 3 from RR08 for the 70-hPa level at Koror, annotated with the changes noted above. Also included are anomalies from the homogenized data set described by Sherwood et al. [2008]. Radiosonde model changes are indicated by lines with long dashes, computer and algorithm changes are noted by lines with short dashes. The only change that occurs during the cooling period of the 1990s we are interested in is the one in 1995, which does not appear to coincide with a change in the 70-hPa temperature trace that would change the sense of the sign we are discussing here. If anything, the introduction of the Vaisala sonde in 1995 may have produced an upward jump and decreased the magnitude of the trend of interest, although the Sherwood et al. [2008] corrected data do not show any significant change there. The QBO features discussed previously are present in both temperatures traces. It is also worth pointing out that the IGRA data for Majuro shows a decrease in temperature at the 70-hPa level similar to the decrease at Koror between 1993 and 1999 (see Figure 2 of RR08), but the metadata for Majuro shows no sonde change in 1995, in contrast to Koror. [7] Quoting from L09 “Another feature considered by RR08 at their warm pool stations is the abrupt rise in temperature of about 4 K in early 1999, followed by an abrupt drop of about 6 K about 2 years later. While it is not as easy to attribute this feature to artificial causes, there is good reason to be suspicious. The station history metadata indicate a computer system change at all six of their select stations in early 1999.” [8] Although L09 tries hard to show that this feature could be artificial, its reality is unquestionable. Apart from other considerations, it is difficult to see how a computer system change could produce an anomaly of this magnitude or shape. However, Figure 2 clearly shows that the feature is present at radiosonde stations far from the six mentioned above, including Bogotá, Ascension, and Singapore, and also in a weaker form at Darwin. It is also present in the Sherwood et al. [2008] data set at Koror plotted in Figure 1. The geographical separation of these stations suggests that it is in fact present throughout the global tropics. The feature is of interest since (1) as we suggested in RR08 it seems to represent a temporary return of the QBO signal, which had been very weak since the early 1990s, and (2) it appeared to initiate the even colder period of the early 21st century, when temperatures reached the lowest point of the entire 40-year period, and when the anticorrelation with SST anomalies became clear. L09 suggests that “procedural changes” could be responsible for this event, but its appearance as a major feature at many independent locations around the global tropics, and its occurrence immediately preceding the coldest known period in the tropical lower stratosphere make this explanation untenable. [9] Quoting again from L09 “In examining Figure 7 of RR08, which shows the temperature difference between the time periods 2001–2003 and 1995–1997, it is curious that by far the largest differences in the tropical zone occur in the region dominated by the radiosonde stations having suspect continuity at this time.” [10] A curious switch has taken place here. The time period of possible suspect continuity was previously placed in the mid-1990s but is now extended to the period around the turn of the century. If one takes L09's criticism at face value, one would have to assume that the radiosonde stations of the western tropical Pacific have been in a constant state of turmoil since the early 1990s, after having apparently been stable and well behaved for at least the preceding 30 years. As we stated in RR08, the strong stratospheric cooling was located over the Indian Ocean/Indonesian “maritime continent,” where SSTs are the warmest in the world, and deep tropospheric convection is an everyday feature. If deep tropospheric convection is the link between the sea surface and the lower stratosphere, this is the region where the connection is likely to be strongest. [11] The reality of the cooling of 2000 is shown by Figure 3, which plots cold point temperature anomalies as derived from the Majuro radiosonde record and the MSU TLS (Temperature Lower Stratosphere) channel matched to Majuro's location. Also shown is a completely independent record of zonally averaged tropical water vapor at the 82-hPa level derived from the HALOE instrument on the UARS satellite. The drop in water vapor content at the end 2000 (on the order of 0.4 ppmv as noted in RR08) coincides with the cold point temperature drop as well, with a time delay as noted in RR08, and is a result of the freeze-drying effect of the extremely cold tropopause region. The precision of the HALOE water vapor measurements is estimated by Harries et al. [1996] to be a few percent, and the change at the end of 2000 amounts to a 15% change, so is a significant drop in tropical stratospheric water vapor. The time delay represents the time needed to transport the water vapor from the cold point to the 82-hPa level. Again there can be little doubt of the reality of the cooling event. It is also quite clear in both NCAR/NCEP and UKMO zonally averaged assimilated temperatures, shown in Figures 4 and 11 of RR08. [12] In summary, the central criticism of the L09 is that the conclusions presented in RR08 are suspect at the least, and possibly worthless because little attention has been paid to purported artificial features introduced into the data by changes in radiosonde operation during the 1990s. Some of L09's points could readily have been resolved by a careful reading of RR08, and we have shown above that others are irrelevant to our conclusions, which do not rest on the magnitude of the effects we discuss, but simply on their existence, and their correlation with other geophysical parameters such as sea surface temperatures and stratospheric water. Furthermore, as mentioned above, the study reported in RR08 could not have been carried out with any of the homogenized data sets favored by L09, since they do not report the important cold point heights and temperatures. The assumption is made throughout L09 that the homogenized radiosonde data is absolutely correct, while the high-quality Integrated Global Radiosonde Archive, which has been widely used, together with its predecessor (CARDS), is in serious error. When five independent radiosonde time series from five stations in the western tropical Pacific all show similar dramatic changes, using the logic expressed in L09, one is forced to the unlikely conclusion that all stations changed instruments or data procedures at exactly the same time. Yet, the metadata available with the IGRA sonde data does not indicate that to be the case. As one example, the metadata for Koror shows that a VIZ to Vaisala instrument change occurred on 1 December 1995, while at Singapore, Vaisala instrumentation was used throughout the 1990s. Both stations (see Figure 2) show the decrease over the 1990s that was one of the topics in RR08, where the sign of the feature is our interest, as opposed to the absolute magnitude of the trend. The available metadata does not have entries for 2001 for the stations considered in RR08, but we do have the added information from the HALOE water vapor retrievals that there was a dramatic change at that time. This change is reflected in changes in both assimilated temperatures and the IGRA data, and the magnitudes, as noted in RR08, are consistent. Looking at a single geophysical quantity and a single station may flag the feature as suspicious. However, our point is that the consistency between different independently measured geophysical quantities supports its validity. One should not attribute all apparent discontinuities in geophysical temperature records as unphysical. We should keep in mind that such discontinuities can be valid solutions to the Navier-Stokes equations, and are actually of great interest in a world with a changing climate. [13] The quality of the data record being used for long-term trend analysis, and the existence of obvious instrumentation differences must clearly be considered. The use of multiple data sets is an important factor in the interpretation of changes, and in deciding whether changes are physical or not. The “homogenization” of the radiosonde record is a valuable effort, and a necessary step in the evaluation of long-term trends at certain fixed pressure levels. It should not, however, be allowed to mask real physical changes that have been taking place in the tropical atmosphere, or to label them as “almost certainly artificial,” as L09 has done. After all, even his Figure 1 shows a clear cooling in the mid-1990s in the homogenized records, although with only half the magnitude of the raw radiosonde data. In fact there is good reason to be suspicious of the magnitude of the correction that has been applied, since the example we discuss above of Koror shows little or no variation occurring at the time of the instrument change, undermining L09's attribution of the change to the fact that all of the warm pool stations made instrument changes at roughly the same time. Clearly there is a need for further study of these problems, since the radiosonde archive is an invaluable historical record of the Earth's climate. For the present case, however, we conclude that the conclusions of RR08 are robust and valid, and that the criticisms of L09 are largely irrelevant. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.