COST ES1206 Sub-Working Group "Data Homogenisation": 1st Workshop - Brussels, Belgium

26 Apr 2016, 12:00 → 27 Apr 2016, 17:30 Europe/Brussels

Nadir Room (Royal Observatory of Belgium (ROB))

Eric Pottiaux (Royal Observatory of Belgium (ROB)), Roeland Van Malderen (Royal Meteorological Institute of Belgium)

 

 

The workshop was dedicated to the homogenisation of GNSS-based (long-term) time series in order to use them for studying climate signals (trends, variability, cycles…).

The workshop has taken place on April 26-27, 2016 on the Space Pole where the Royal Observatory of Belgium (ROB), the Royal Meteorological Institute (RMI) and the Belgian Institute for Space Aeronomy (BISA) are located. It was organised in the framework of the COST ES1206 Action "Advanced Global Navigation Satellite Systems tropospheric products for monitoring severe weather events and climate" (GNSS4SWEC) supported by the Solar-Terrestrial Centre of Excellence (STCE).

If you are interested to take part into this activity, please contact us.

Looking forward to hear from you,

Eric Pottiaux and Roeland van Malderen

 

---

For more info about Brussels and the events that occurred in Brussels on March 22, 2016, see also the site visit.brussels.

---

 

Data Homogenisation Workshop in Brussels

• Data Homogenisation Workshop Presentation Bundle
• Minutes of the sub-WG workshop on Homogenisation in Brussels
• sub-WG on Homogenisation workplan after Brussels Workshop

Homogenisation - Metadata Documents Homogenisation - Metadata Template Homogenisation - Metadata Template - Fake Example

Splinter @ Wroclaw - Presentation sub-WG3 Homogenisation - Splinter Session- Wroclaw, Sep 30, 2015

Tuesday, 26 April

14:00 → 14:25 Opening Session

Conveners: Dr Eric Pottiaux (Royal Observatory of Belgium (ROB)), Dr Roeland Van Malderen (Royal Meteorological Institute of Belgium)

14:00 Welcome from the Local Organiser

Speakers: Dr Eric Pottiaux (Royal Observatory of Belgium (ROB)), Dr Roeland Van Malderen (Royal Meteorological Institute of Belgium)

14:10 A reference IWV dataset combining IGS repro1 and ERA-Interim reanalysis for the assessment of homogenization algorithms

This presentation will give insights into the GPS and ERA-Interim IWV datasets provided to WG3 for the assessment of homogenization algorithms: - Processing characteristics of IGS repro1 (and their limitations) - Screening of the GPS ZTD data - Conversion of GPS ZTD to IWV - Computation of ERA-Interim IWV data at the location of GPS stations - Statistical analysis of GPS minus ERA-Interim IWV differences - Screening and validation and of the IWV differences (removal of sites with large representativeness differences) The dataset is limited to 120 global sites with data available for the 1995-2010 period. The IWV data and differences are provided at daily and monthly time resolutions.

Speaker: Dr Olivier Bock (IGN)

14:25 → 16:00 Technical Session on Homogenization Algorithms

This session is dedicated to the description of homogenisation algorithms. Each participant who can give details about the use of such algorithms are invited to contribute to this session.

Conveners: Dr Eric Pottiaux (Royal Observatory of Belgium (ROB)), Dr Roeland Van Malderen (Royal Meteorological Institute of Belgium)

14:25 Statistical method for change-point detection in daily and monthly time-series

Unexpected changes in the time-series are caused by several (usually non-climatic) factors. We applied statistical approach to analyze change-point detection. The method is based on a second correlated series that can be relied on to be correct. In the presented work we also describe the critical values calculation that was used. Beside the main presented method we also applied other technique based on a very similar principles, yet providing different statistical evaluation.

Speaker: Michal Elias (Geodetic Observatory Pecny, Czech Republic)

14:45 Discovering Singular Spectral Analysis and Bootstrapping

During this Academic Year, I studied the approach of Singular Spectral Analysis (SSA). It seems interesting and useful in several applications, included processing of atmospheric data with homogenization purposes. The main idea is to fit original data with SSA and then to compare the approaches of change point detection for original and reconstructed data. Another approach for change point detection is to start with generation pseudo-observables with bootstrapping technique and analyse the cumulative sums of time series, which we got. This method needs the holding of assumption that all values in the original time series have same appearance probability. Some considerations will be presented on the COST ES1206 sub-WG Workshop on Data Homogenisation in Brussels, 26-27 April 2016.

Speaker: Mr Peep Miidla (University of Tartu, Estonia)

15:05 Development of New Segmentation Algorithms for the Homogenization of GNSS IWV Data

Speaker: Dr Olivier Bock (IGN)

15:25 Robust, non-parametric techniques for the identification of change-points in the mean

We present three different tests, based on non-parametric techniques, for the objective identification of discontinuities in the mean (or median) of time series. These change-point tests are able to identify multiple discontinuities and requires no 'metadata' or comparison with neighbouring stations or reference data. These tests are based on (the sum of) the ranks of the values in the sequence. We chose to use the Petitt-Mann-Whitney statistic, which is equivalent to the Wilcoxon rank-sum statistic, but gives a clear procedure for determining critical values. But, for comparison, we also discuss the more commonly used Wilcoxon rank-sum test and the cumulative sum test to detect change points in the mean in a time series.

Speaker: Dr Roeland Van Malderen (Royal Meteorological Institute of Belgium)

15:45 Open Discussion

16:00 → 16:15 Break

16:15 → 18:20 Homogenisation : Status and Expectations of Individual Work

This session is dedicated to the analysis and inter-comparision of the detected breakpoints and homogenised dataset. Each participant having a set of detected breakpoint and/or homogenised dataset is welcome to present his results in this session.

Conveners: Dr Eric Pottiaux (Royal Observatory of Belgium (ROB)), Dr Roeland Van Malderen (Royal Meteorological Institute of Belgium)

16:15 Some remarks on statistical method for change-point detection results

We analyzed the method related to change-point detection. The method is based on a second correlated series that can be relied on to be correct. The preliminary results show that over 30% of analyzed monthly time-series are presented as being inhomogeneous and this also applies to over 65% of daily time-series data.

Speaker: Michal Elias (Geodetic Observatory Pecny)

16:25 Homogenization activities in GFZ Potsdam

The homogenization activities in GFZ Potsdam consist of two different efforts: homogenization of IWV and ZTD time series from GNSS ground-based stations and homogenization of VLBI derived coordinates and troposphere products. The latest studies in the GNSS field were the analysis of 101 stations time series on a global scale using Penalized Maximal t test (PMT). The VLBI effort is an analysis of the impact of different mapping functions on geodetic and tropospheric products from VLBI data analysis.

Speaker: Mr Simeonov Tzvetan (GFZ)

16:35 Homogenisation Activities at IGN France

Speaker: Dr Olivier Bock (IGN)

16:45 Identified break points in the IGS repro 1 reference dataset by nonparametric rank-sum tests

We present the break points that have been identified in the IGS repro 1 IWV series by applying nonparametric rank-sum tests on the daily and monthly IWV differences of the IGS repro 1 reference dataset and the ERA-interim reanalysis output at the station locations. We also applied the same tests on the IGS repro 1 ZTD and IWV time series and compare all the identified break points with each other. Which ones are robust? Which are related to known instrumental changes?

Speaker: Dr Roeland Van Malderen (Royal Meteorological Institute of Belgium)

16:55 Open Discussions

19:00 → 21:00 Conference Dinner

Restaurant website (In french)
http://www.spinnekopke.be/#accueil

Access Map:
https://www.google.be/maps/place/In+'t+Spinnekopke/@50.8483607,4.3435118,15z/data=!4m2!3m1!1s0x0:0xd2dc67f91e6f01f2?sa=X&ved=0CIYBEPwSMApqFQoTCKyXprjm2sYCFclKFAodiMcGOA

Wednesday, 27 April

09:00 → 10:30 Noise and Signals in the time series / Generation of synthetic dataset(s)

This session is dedicated to the following subjects: 1) Noise in the time series, 2) Signal in the time series, and 3) Generation of synthetic dataset(s). Individual presentation will be followed by an open dicussion.

Conveners: Dr Eric Pottiaux (Royal Observatory of Belgium (ROB)), Dr Roeland Van Malderen (Royal Meteorological Institute of Belgium)

09:00 Assessment of Seasonal Signals and Noise in ZTD and IWV time series

Within last 26 years, the Zenith Total Delay (ZTD) or Integrated Water Vapour (IWV) time series have been estimated together with position series during processing or reprocessing of GNSS data. The basic advantage of ZTD and IWV time series analysis is a possible application to monitoring of climate changes. In this kind of studies, trend and its uncertainty indicating a significance are widely interpreted. In our research we model trend, significant seasonal signals and noise properties with Least-Squares Estimation (LSE) and Maximum likelihood Estimation (MLE) for the common data set of daily IWV time series provided by COST ES1206 homogenisation group. We show, that the amplitudes of annual signal are around 20 mm with maxima in summer. The maxima of semi-annual oscillations fall in January and March and are well-correlated for Northern Hemisphere. A detailed analysis of noise properties lead us to decide on the proper noise model that should be used when the uncertainties of seasonal signals and trend are estimated. We examined different processes: from white through power-law ending on autoregressive. We estimated that when the autoregressive process is taken into consideration, the uncertainties of trend are larger of few times when being compared to white noise assumption. The autoregressive model is optimal for IWV data, since it describes the flattening of Power Spectral Densities in low and high frequencies. Moreover, the results of daily IWV analysis were summarized with results of hourly ZTD series that were estimated in a second reprocessing undertaken by University of Luxembourg as part of its activities as an International GNSS Service (IGS) Tide Gauge Benchmark Monitoring (TIGA) analysis centre.

Speaker: Dr Anna Klos (Military University of Technology, Warsaw, Poland)

09:20 Analysis of ZTD and IWV Time Series from GPS Reprocessing Campaign

Seasonal components in the ZTD (Zenith Tropospheric Delay) and IWV (Integrated Water Vapour) time series are important and provide information about weather conditions in a given region. The main purpose of this research is to analyze linear trends and seasonal components of ZTD’s and IWS’s in GPS reprocessing campaign at selected stations in Europe (IGS repro1 dataset-URL1). The seasonal components and linear trends were estimated using the LSE (Least Square Estimation) approach. The frequencies in the ZTD and IWV time series were found and periodograms were prepared. The frequency spectrum was estimated by using combined sine and cosine functions to the time series.

Speaker: Mrs Selma ZENGİN KAZANCI (WG1)

09:40 Homogenisation of tropospheric data with autoregressive noise model

A task of homogenisation of tropospheric data estimated in GNSS processing is based on a proper detection of possible offsets due to changes in hardware or equipment, earthquakes or any other artificial events. This task is aimed at determination of reliable trend and its uncertainty that is used in climate studies since any undetected offset influences the value of trend and increases its error. In this research, we used Monte Carlo simulations of 25 years long data of two different noise types: white and autoregressive. These were chosen, since white noise is being widely assumed in analysis of tropospheric data and autoregressive noise is what tropospheric data follows in reality. The seasonal signals with amplitudes between 5 and 40 mm’s along with few strictly defined offsets were added to simulated series, which were then subjected to homogenisation task. We made blinded tests and detected possible epochs of offsets manually. We found that simulated offsets were easily detected in series with white noise, no influence of seasonal signals on indicated epochs was noticed here. The autoregressive series were much more demanding when offsets had to be determined. We found few epochs, for which no offset was simulated. This was mainly due to strong autocorrelation of data, which brings an artificial trend within.

Speaker: Dr Anna Klos (Military University of Technology, Warsaw, Poland)

10:00 Open Discussion

10:30 → 11:00 Break

11:00 → 12:30 Future Plans and Outcome of our Activities

This session is dedicated to discuss this sub-WG3 future plan and the expected outcomes of our activities. This covers the following topics:

• Exchange format for the homogenized time series and their meta-data (prior and after homogenization)
• What to do with existing a priori meta-data: always correct them? or not ? What's the impact on the trends...
• Inter-compare results (e.g. trends) in the original reference dataset and in the homogenized dataset time series
• Access to deliverable(s)
• Referencing (Acknowledgement, co-authorship, DOI... ?)
• Discussion "Paper in the Special Issue"
• Any link to past/present/future activities ?

Conveners: Dr Eric Pottiaux (Royal Observatory of Belgium (ROB)), Dr Roeland Van Malderen (Royal Meteorological Institute of Belgium)

11:00 Standardization, Exchange Format and Meta-Data

As stated in the Memorandum of Understanding (MoU), the WG3 of the COST Action 1206 (GNSS4SWEC) targets to standardize the methods, exchange format for both the dataset and its associated meta-data. In this presentation, we will present and discuss teh different possibilities to standardize the exchange format for the dataset, the break-points and the associated meta-data within the sub-WG3 'data homogenisation' activities, but also as a final outcome of the WG3 activities.

Speaker: Dr Eric Pottiaux (Royal Observatory of Belgium (ROB))

11:20 Open Dicussions

12:30 → 14:00 Lunch Break

14:00 → 15:00 Expertise from Previous and Ongoing Homogenisation Activities: (Invited Expert: Dr. Enric Aguilar)

Conveners: Dr Eric Pottiaux (Royal Observatory of Belgium (ROB)), Dr Roeland Van Malderen (Royal Meteorological Institute of Belgium)

14:00 Experience from the COST Action 'HOME'

Speaker: Dr Enric Aguilar (Center for Climate Change)

14:30 Open Discussions

15:00 → 17:00 Open Discussions and Closing Session

Conveners: Dr Eric Pottiaux (Royal Observatory of Belgium (ROB)), Dr Roeland Van Malderen (Royal Meteorological Institute of Belgium)