Good afternoon everyone, the topic of my talk today is using ground-based GNSS data to estimate long term trends in the atmospheric water vapour content.
Here comes some movitvation for the study. Water vapor is a very efficient greenhouse gas. An increase of 20% of the water vapor content in the tropics has a larger impact than a doubling of the carbon dioxide concentration. Water vapor is also one of the most important climate feedback process. An increase of temperature will result in an increase in water vapor in the atmosphere since warm air can contain more water vapor than the cold air. Therefore, long-trend of the atmospheric water vapor can be used as an independent data source to detect global warming. In order to determine a long-term trend, accurate observations with long-term stability is necessary, Meanwhile, a high spatial density of measurements is desired for a global monitoring. GPS techniques, therefore, has a prmary advantage for such applications in terms of stability and surface spatial resolution.
The principle observes of GPS are the differences in arriving time of signal from satellites. Since time is a physical parameter, it can be measured with high accuracy principely under all weather conditions, also over long time periods. Additionally, the spatial resolution of GPS network is continuously improving globally and locally.
As you all know, that the neutral atmosphere is the lower part of the atmosphere. The GNSS signals are taking longer time propagating in the neutral atmosphere than in the vacuum, partly due to the presence of water vapor. The excess propagation path are therefore called as delays. For geodesists, of course, those delays are bad to see because they are one of errors for the GNSS. However, meteorologists are happy to know these delay because these delays give us the information about the atmospheric water vapour. So now, errors become signals.
So how can we measure the water vapour using GNSS signal? Here is the flow chart to show the whole procedure. First, observations are put into analysis using GNSS processing software, like GIPSY. One of outputs from the GIPSY processing is ZTD which actually is the delay in the neutral atmosphere as I mentioned before, but refer to the zenith direction. The ZTD consists of two part of, one is the hydrostatic delay, which can be estimated with very high accuracy if we know the ground pressure. The rest part is the wet delay, which is directly related to the integrated water vapour through a parameter Q , which is depending on the location and also on the different season varying. It around 6.5 in the area for my study.
After getting the IWV time series, it is possible to estimate the long term trend of it. This can be done by fitting IWV values into a model which considers the periodic variations of the IWV. The figure to the right shows the time series of estimated IWV for the site of Kiruna at north of Sweden, and the figure to the right is the corresponding the results shown in frequency domain, where we can see how many terms in the model is necessary to describe the periodic variations of the IWV. We can see clearly see a peak for the annual perido, and also one much smaller at the semiannual period. No other significant peaks are seen indicating that it is sufficient to use an annual and a semiannual terms in the model here. In this model, coefficient A stands for the IWV trend.
Here are some examples showing the original time series of IWV and the fitted models for three GPS sites. The green lines represent the seasonal fit of the model and red lines are IWV trends. We can see that the fit follows the seasonal pattern well. I should also point out here, since we only consider annual and semi-annual varaitions in the model, there are deviations between the original IWV and model fit due to some shorter-term variations of the IWV, like in few days, which are not described by the model. The figure shows the covariance of residuals after fit the model to the original IWV, we can clearly see those residuals are correlated to each other over several days. Therefore, when we calculated the uncertainty of the estimated IWV trends, We cannot simple assume that the model errors are uncorrelated behaving like a white noise, which will make us underestimate the uncertainty of the estimated IWV trend. In order to get more realiable uncertainties of the trend estimation, we have to take those temporal correlations into accoun.
So, in my current study. I have processed GPS data obtained from 21 sites in Sweden and 12 sites from Finland. The IWV trends are estimated for a time period from 21th Nov. 1996 to 20 th Nov. 2010. The figure shows the geographical pattern of these trends. We can that the IWV trends for this 14 years are in the range from -0.5 to 0.5 kg/m^2/decade. After taking the temporal correlation into account, we have the uncertainties in the IWV trends approximately 0.35 kg/m^2/decade for all sites.
The sensitivity of the IWV trends to the selected time period is shown in this two figures. The figure to left show the IWV trends obtained from the first 13 years, from 1996 to 2009, and the right figure shows the result by using the last 13 years, from 1997 to 2010. You can immediately notice some differences between two figures. Like in these area (ARJ0, VIL0, OSTO), we have positive trends for the first 13 years, but negative values from the other. The reason for those differences is likely that a 13 year period is two short. A very dry or wet year either in the beginning or the end of the period will significantly effect the estimated trend.
A similar effect, in terms of sensitivity, can also be seen in these two figures where IWV trends were obtained using data only from the summer season and from the winter season. The difference are larger than what we have just seen from last slide, and the summer season have much larger trends compared to the winter season. What we are seeing here probably is a true seasonal difference in IWV trends, which however also could be an effect caused by the different time periods used for trends estimation as we see from the last slide. This can be validated by using even longer time series for trend estimation.
Except only looking the GPS derived IWV trends, I also made trend comparison between the GPS and Raidosonde data. The map show location of the GPS sites and radiosonde sites for the comparison where starts represent GPS sites and dots are radiosonde sites. The IWV trends estiamted from 13 GPS sites were compared to the trends obtained from 9 radiosonde sites.
The results are showed in this table. The maximum distance between the GPS site and the radiosonde site in a comparison is around 120 km. The GPS and RS trends are shown in these two columns, and the maximum difference in trends is seen from this comparison (KIVE) about 0.5 kg/m^2/decade, which is still comparable to the uncertainty of the trend (0.35 kg/m^2/decade). If plotting the GPS trends again the RS trends, we can see a good agreement with a correlation coefficents about 0.68.
So, some conclusions.
MONITORING LONG TERM VARIABILITY IN THE ATMOSPHERIC WATER VAPOUR CONTENT USING GROUND-BASED GPS RECEIVER NETWORKS
Monitoring Long Term Variability in the Atmospheric Water Vapor Content Using Ground-Based GPS Receiver Networks Tong Ning and Gunnar Elgered Department of Earth and Space Sciences Chalmers University of Technology Onsala Space Observatory, Sweden
Motivation <ul><li>Water vapor is a very important greenhouse gas. </li></ul><ul><li>Water vapor is one of the most important climate feedback process. </li></ul><ul><li>Long-term trends of the atmospheric water vapor content can be used as an independent data source to detect global warming. </li></ul><ul><li>Accurate observations with long-term stability is important for trend estimations. </li></ul><ul><li>A high spatial density of measurements is desired. </li></ul>
GPS can work under in principle all weather conditions with increasing spatial resolution locally and globally. Global: the number of stations from the permanent International Global Navigation Satellite Systems (GNSS) Service (IGS), formerly the International GPS Service, is now (July 2011) globally over 360. Local network from Sweden: <ul><li>SWEPOS has been in operation since 1993 with 21 geodetic quality stations (stars). </li></ul><ul><li>More than 170 stations, 1200 km from north to south, and 400 km from east to west, with an average site separation of approximately 70 km. </li></ul>GPS networks
neutral atmosphere Errors to geodesists Signals to meteorologists Measuring water vapor using GPS
Measuring water vapor using GPS (continued) <ul><li>Use GPS processing software, e.g. GIPSY 5.0 applying antenna phase center corrections and an elevation cut-off angle of 10 degrees </li></ul><ul><li>Solve for station coordinates, clock errors, Zenith Total Delay (ZTD), etc. </li></ul><ul><li>ZTD=Zenith Hydrostatic Delay (ZHD) +… </li></ul><ul><li>Zenith Wet delay (ZWD) </li></ul><ul><ul><li>ZHD can be estimated if surface pressure is known. </li></ul></ul><ul><ul><li>ZWD is related to the Integrated Water Vapor ( I WV) content of the atmosphere: </li></ul></ul><ul><ul><li>ZWD (mm) =Q • IWV (kg/m 2 ) </li></ul></ul><ul><ul><li>where Q ≈ 6.5 (depending on location and season) </li></ul></ul>
Estimating IWV trends The IWV has been obtained from we make a fit to the model: where t is the time in years and the coefficients I 0 , A, B, C, D, E are estimated. Both annual and semi-annual terms are used to model the seasonal variations.
IWV trends for some GPS sites Latitude: 66.32 o Trend: -0.27 kg/m 2 /decade Latitude: 62.23 o Trend: 0.08 kg/m 2 /decade Latitude: 56.09 o Trend: 0.17 kg/m 2 /decade
IWV trends over Sweden and Finland <ul><li>21 sites from Sweden and 12 sites from Finland </li></ul><ul><li>IWV trends in kg/m 2 /decade. </li></ul><ul><li>Analysis period: </li></ul><ul><li>November 21, 1996 – November 20, 2010. </li></ul><ul><li>The uncertainties in the trends are estimated to </li></ul><ul><li>~0.35 kg/m 2 /decade (taking the temporal correlation into account) </li></ul>
Sensitivity of the trends to different time periods 21 Nov. 1996 – 20 Nov. 2009 21 Nov. 1997 – 20 Nov. 2010
Conclusions <ul><li>IWV trends estimated from GPS vary from –0.3 to +0.5 kg/m 2 /decade over Sweden and Finland for the last 14 years. </li></ul><ul><li>Uncertainties in the trends are ~0.35 kg/m 2 /decade (taking temporal correlations into account) </li></ul><ul><li>Trends are (as expected) sensitive to the specific time period investigated (due to the short period of data available). </li></ul><ul><li>Good agreement — correlation coefficient of 0.68 — with the trends from radiosondes launched nearby. </li></ul>