I’ve moved to WordPress. This post can now be found at On The Differences Between Surface And TLT Datasets###################
OVERVIEWThe differences between surface temperature and lower troposphere temperature (TLT) anomalies are examined in this post. The globe is broken down into sectors in an effort to locate the area with the greatest difference between surface temperature and TLT anomalies. The responses of Sea Surface Temperature (SST) and Sea TLT to El Niño and La Niña events are also examined to see if they may cause part of the divergence between the surface temperature and TLT data after the 1997/98 El Nino. The post also presents animated maps of surface temperatures and TLT anomalies and animated maps of the correlations of NINO3.4 SST anomalies and global temperatures to further highlight the differences.
The differences between the infilled surface-based (GISS and NCDC) temperature anomaly datasets and the Lower Troposphere Temperature (RSS and UAH) anomaly datasets create a multitude of comments from both sides of the global climate change debate whenever the datasets are plotted together. While there may be biases created by surface station location, infilling method, Urban Heat Island (UHI) adjustments (or lack thereof), etc., let’s put the land-based discussions aside for this post and consider the possibility that Sea Surface Temperature (SST) might respond differently than the lower troposphere temperatures to El Niño and La Niña events, and that those differences in response might account for some of the divergence.
Satellite-based TLT data has been available since 1979, providing more than 3 decades of data for comparisons to surface data. Unfortunately, the surface temperatures and TLT between 1982 and 1995 are overwhelmed by the dips and rebounds caused by the volcanic eruptions of El Chichon and Mount Pinatubo. Accounting for these in the global data is relatively easy, but we’ll be dividing the globe into sectors in this post and having to account for the volcanic eruptions in each would add unnecessary complexity, so we’ll only look at the period from 1995 to present. Additionally, GISS Deletes Arctic And Southern Ocean Sea Surface Temperature Data in areas with seasonal sea ice and extends adjoining land surface temperature data (with its higher trend) out over the oceans. GISS does this to infill temperature data in the ocean areas with sea ice. The NCDC product does not take this step. To eliminate that difference, the data north of 60N and south of 60S have been excluded from this post.
So basically we’ll be examining the global land and sea surface temperature and lower troposphere temperature data minus the poles from 1995 to present. Figure 1 illustrates the average of the surface temperature datasets (GISTEMP and NCDC) and the average of the TLT datasets (RSS and UAH). We’ll use the averages initially to simplify the illustrations (and to eliminate discussions of any differences between the individual datasets). The TLT anomalies in Figure 1 have been scaled so that the rise in response to the 1997/98 El Niño is approximately the same as the surface data. (And for those wondering why the data ends in May 2010, I’ve used the KMNI Climate Explorer as the source of data for this post, and the latest RSS TLT anomaly data update there is May 2010.) Note also that the title block states that the surface data has been lagged 2 months. In other words, the surface data responded 2 months earlier to the 1997/98 El Niño than the TLT anomalies, so I’ve shifted the surface data back in time by two months. The linear trend of the surface data is more than twice that of the scaled TLT data. The surface data has a linear trend of approximately 1.1 deg C per Century but the trend of the scaled TLT data is only 0.4 deg C per Century. And much of this appears be caused by the failure of the surface data to respond as fully as the TLT data to the 1998/99/00/01 La Niña.
NOTE: The TLT scaling factors and surface data lags vary per subset in the following. The scaling factors and lags are noted in the title blocks.
If we subtract the scaled TLT anomalies from the surface data, Figure 2, we can see that there appears to be an upward shift in the surface data in 1998 that is not present in the TLT data. And after 1998, the surface data rises at a rate that is faster than the TLT anomalies.
LET’S DIVIDE THE GLOBE INTO THREE SECTORS
I’ve divided the globe (minus the poles) into three sectors for the comparisons that follow. The areas are split at 180, 78W, and 80E longitude. To simplify the discussion, I’ve identified them as “Far West”, “Near East and Near West,” and “Far East”, based on their proximity to zero longitude. Refer to Figure 3. The “Far West” data includes the Eastern Pacific and most of North America. The “Near East and Near West” includes the Atlantic and Western Indian Oceans and South America, Africa, Europe, and Western Asia. And the “Far East” data includes the East Indian and West Pacific Oceans and Eastern Asia and Australia.
FAR WEST COMPARISON
A time-series graph of the surface and TLT anomalies for the Far West sector is shown in Figure 4. These datasets capture the Eastern Pacific and most of North America. The surface temperatures have been lagged 5 months to align the leading edges of the responses to the 1997/98 El Niño. Note also that the TLT anomalies have not been scaled, which means the variations in TLT anomalies for the East Pacific and most of North America are comparable to the changes in surface temperature anomalies. Also note that this is the only one of these three sectors where the TLT anomalies are rising faster than the surface temperatures, though the difference in the trends is relatively small. And of course, the linear trends themselves are very small, with the TLT anomalies having a linear trend of less than 0.3 deg C per Century. This means that this sector of the globe does not contribute to the disparity between surface and TLT anomalies; in fact, this area would suppress the difference.
FAR EAST COMPARISON
The Far East data, Figure 5, represents the East Indian and West Pacific Oceans, along with Eastern Asia and Australia. For this part of the globe, the surface data needed to be advanced one month to align the rises of the responses to the 1997/98 El Niño. The surface data does not drop as far as the TLT data during the 1998/99/00/01 La Niña and remains elevated during the 2002/03 El Niño. The two datasets diverge again during the 2007/08 La Niña. These divergences cause the surface data to have a higher trend than the TLT anomalies. The difference in the linear trends, though, is only approximately 0.2 deg per Century. This accounts for little of the disparity in trends shown in Figure 1.
And that means much of the divergence in the two global datasets comes from the Near East-Near West sector, so we’ll concentrate on that sector for the rest of the post.
COMPARISON OF NEAR EAST AND NEAR WEST
For the Near East and Near West portion of the globe, Figure 6, the linear trend for the surface temperature data is approximately 1.9 deg C per Century, while the trend for the scaled TLT data is approximately 0.47 deg C per Century. The surface data has very little drop during the 1998/99/00/01 La Niña.
This is confirmed in Figure 7, which illustrates the difference between the scaled TLT and surface data for the Near East and Near West. Note the significant upward step in the latter half of 1998.
Animation 1 provides maps of the surface temperature and lower troposphere temperature anomalies of the Near East and Near West datasets used in this post. The surface datasets are on the left and the TLT datasets are on the right. The base years for anomalies (1979 to 2009) are the same for all. The maps provide 12-month average temperatures from July to June for the first five years being examined in this post: the ENSO-Neutral year 1996/1997, the El Niño year 1997/1998, and the three La Niña years of 1998/1999, 1999/2000, and 2000/2001. The period of July to June was used since ENSO events typically peak in November, December and January. The animation is provided to show the similarities and differences in the responses to El Niño and La Niña events between the surface-based and TLT datasets.
Figure 8 is the July 1998 to June 1999 cell from the animation. I’ve isolated it because much of the difference between the surface and TLT anomalies (Figure 7) occurs during this period. While there are some differences between land surface temperature and land TLT anomalies during the year from July 1998 to June 1999 in Figure 8, most of the additional warming appears to take place in the sea surface temperature data:
- In the Western North Atlantic,
- In the tropical Eastern South Atlantic in the Benguela and South Equatorial Currents, and
- In the Antarctic Circumpolar Current (ACC) south of South Africa and the western Indian Ocean.
NEAR EAST-NEAR WEST SEA SURFACE AND SEA TLT COMPARISON
So far we’ve compared the average of GISTEMP LOTI and NCDC combined surface data to the average of RSS and UAH TLT anomalies. Unfortunately, the RSS TLT data through the KNMI Climate Explorer is not divided into land and sea TLT subsets. Luckily, the UAH data is, so we’ll have to use it alone in the following comparison.
Figure 9 illustrates the Near East-Near West sea surface and scaled sea TLT data (UAH only) anomalies. The scaled sea TLT linear trend is only 0.27 deg C per Century, while the sea surface temperature (SST) trend is approximately 4 times higher at 1.1 deg C per Century. Much of the divergence occurs during the 1998/99/00/01 La Niña. Then, from 2001 to present, the SST anomalies have climbed higher while the sea TLT anomalies remained relatively flat.
Consider now that the Near East-Near West sea surface area represents approximately 63% of the total surface area for this part of the globe. The divergence between sea surface and sea TLT anomalies for this part of the globe, therefore, represents a significant portion of the overall global difference.
A DISCUSSION OF ENSO EVENTS
During an El Niño, warm water from the surface and from up to 300 meters below the surface of the Pacific Warm Pool sloshes to the east and spreads across the surface of the eastern tropical Pacific. This raises SST there. The rise in SST in the eastern Tropical Pacific causes more moisture than normal to be pumped into the atmosphere through evaporation, which is how the ocean releases most of its heat to the atmosphere. The warm and moist air rises in a process known as convection, and the air cools as it rises. The cooler air can’t hold as much water, so the moisture condenses and it rains, releasing the heat into the atmosphere. (During a La Niña, the waters in the eastern tropical Pacific are cooler than normal, and less heat than normal is released to the atmosphere.)
The warmed air is transported to the east and poleward by atmospheric circulation. This can be seen in the animations of TLT anomalies in the two right-hand cells of Animation 2.
Note: Each of the cells in this animation represents a 12-month average of the temperature anomalies. Again, the surface data are on the left and the TLT data are on the right. The first cell represents the period of Jun 1996 to May 1997. It is followed by the next 12-month period, July 1996 to June 1997, and so on, in sequence. The use of 12-month averages minimizes weather and seasonal noise. I’ve included the eastern Pacific (the Far West sector) in this animation to show the timing of the El Niño in the eastern tropical Pacific and the transition to La Niña. Last, the animation lasts for 36 frames, from the period of July 1996 - June 1997 to the period of June 1999 – May 2000. This captures the evolution and decay of the 1997/98 El Niño plus the first few months of the 1998/99/00/01 La Niña.
It is very apparent in the two left-hand cells of Animation 2 that sea surface temperatures in the tropical North Atlantic rise in response to the El Niño. For a detailed discussion of how this occurs, refer to Wang (2005), "ENSO, Atlantic Climate Variability, And The Walker And Hadley Circulation." Wang (2005) link:
Wang (2005) states, “The Walker and Hadley circulations can serve as a ‘tropospheric bridge’ for transferring the Pacific El Niño SST anomalies to the Atlantic sector and inducing the TNA SST anomalies just at the time of year when the warm pool is developing.”
The warm pool being discussed in Wang (2005) is the Western Hemisphere Warm Pool, which is made up of eastern North Pacific west of Central America, the Gulf of Mexico, the Caribbean and the western tropical North Atlantic. The acronym TNA stands for Tropical North Atlantic.
Wang (2005) continues, “The Hadley weakening results in less subsidence over the subtropical North Atlantic, an associated breakdown of the anticyclone, and a weakening of the NE trades in the TNA. The wind weakening leads to less evaporative surface cooling and entrainment of colder water from below the shallow mixed layer, resulting in positive SST anomalies.”
In other words, an El Niño causes changes in atmospheric circulation which causes a weakening of the trade winds in the tropical North Atlantic. In turn, the weaker trade winds cause less evaporation, resulting in warmer SST. The weakening of the trade winds also causes less cool water from below the surface to be drawn to the surface.
Wang (2005) concludes that paragraph on page 27 with, “The opposite is presumed to occur during boreal winters with unusually cool SSTs in the equatorial Pacific,” meaning that it is assumed the opposite takes place during a La Niña.
Animation 2 shows the differences and similarities between the patterns of surface and TLT anomalies during the evolution and decay of an El Niño. The surface temperature anomalies are responding to changes in atmospheric circulation, while the TLT anomalies are responding to the heat released from the oceans.
For a further discussion of how these patterns are created, refer to Trenberth et al (2002) “Evolution of El Nino–Southern Oscillation and global atmospheric surface temperatures":
Trenberth et al (2002) use a time sequence of correlation maps, their Figure 8, to illustrate the evolution of ENSO events. They write in paragraph 29, “To explore where the heat in the oceans and its effects on the atmosphere actually occur, we have computed correlation and regression maps for several fields at various lags, a subset of which are given here.” And they continue, “This sequence reveals the mean evolution of ENSO and is quite coherent and traceable from about -12- to +12-months lag with N3.4, but becomes quite weak at larger lags. The panels presented in Figure 8 and the [delta] 4-month sequence are chosen to illustrate the evolution in the fewest panels.”
CORRELATION MAP ANIMATIONS
Animating correlation maps of NINO3.4 SST anomalies with surface and TLT anomalies does help to illustrate the evolution and decay in global temperatures in response to ENSO. The areas where there are differences between the surface and TLT datasets also become apparent. And although we are discussing the Near East and Near West sector, we’ll start with an animation of global maps. Keep in mind that the RSS does not present data north of 82.5N or south of 70S, where GISS deletes SST data in these areas and extends land surface data out over the oceans.
Animation 3 shows maps of the correlation of December NINO3.4 SST anomalies with GISS LOTI global surface temperature anomalies in the left-hand cell and RSS global TLT anomalies on the right. GISS data was included because they use satellite-based Reynolds OI.v2 SST data during this period. December NINO3.4 SST anomalies were used because El Niño events normally peak in November-December-January. The animation starts with zero-month lag and continues monthly through a 12-month lag. Like the other animations, there are differences and similarities between the surface and TLT datasets. What caught my eye was the significant positive correlation in the North Atlantic during lags 10 through 12, where they did not appear in the TLT data.
Since the animation can’t be stopped, I’ve included the frames with the 10-, 11-, and 12-month lags as Figures 10 through 12.
I decided to take the animation of the correlation maps one step further. I’ve presented the Atlantic Ocean SST data only in Animation 4, using Reynolds OI.v2 SST data. The correlation maps also use 3-month averages of NINO3.4 and Atlantic SST data. I found the animation quite interesting. During the early lag months, warm waters created by the El Niño appear to migrate from the tropical South Atlantic to the tropical North Atlantic. The warm waters in the tropical North Atlantic then appear to follow ocean currents west into the Caribbean and Gulf of Mexico, before entering the Gulf Stream. Starting around the 9-month lag, warm waters migrate south from the southern tip of Greenland and the Davis Strait. Is that from glacial ice melt in Greenland and Arctic sea ice melt, with the melt caused by the El Niño? They're correlated with NINO3.4 SST anomalies.
That would be a good lead-in animation for a new post on the AMO.
A significant portion of the difference between global surface temperature and TLT anomalies occurs in the North Atlantic Ocean. And the additional warming that takes place appears to be caused by ocean processes correlated with ENSO. Why don’t the TLT anomalies respond to these additional ocean processes?
The data and maps presented in this post are available through the KNMI Climate Explorer:
This post is a tour de force Bob. Lots of skill evident.
I wonder if the TLT is not responding because the energy is effectively stored in the ocean. Only way it can be released in a hurry is via evaporation and precipitation.
That suggests that the speed of the wind is a very important factor in determining sea surface temperature.
Like this analysis, as always. Does my memory serve me well in thinking that you did an article on extracting textual data from the screen, putting it in excel and then converting it to seperate column and row details?
Anonymous: I belive the post you're referring to was this one:
Got it. Thanks Bob
Post a Comment