Global Navigation Satellite System (GNSS) for Disaster Mitigation

Global Navigation Satellite System (GNSS) for Disaster Mitigation By Chathura H. Wickramasinghe Geoinformatics Center Asian Institute of Technology

Establish in 1959 as a Post Graduate School Catering for higher education in Asia 1,967 students from 46 countries 18,020 alumni from 74 countries 22,000 short-term trainees from 71 countries 120 faculty from 26 countries

AIT Academic Structure School of Engineering and Technology Computer Science Design and Manufacturing Engineering Industrial Engineering and Management Information Management Remote Sensing and Geographic Information Systems Telecommunications Information and Communications Technologies Mechatronics Microelectronics Construction, Engineering and Infrastructure Management Geotechnical and Geoenvironmental Engineering Structural Engineering Transportation Engineering Water Engineering and Management GIC School of Environment Resources and Development Agricultural Systems and Engineering Aquaculture and Aquatic Resources Management Energy Environmental Engineering and Management Food Engineering and Bioprocess Technology Gender and Development Studies Natural Resources Management Pulp and Paper Technology Regional and Rural Development Planning Urban Environmental Management School of Management International Business International Public Management Management of Technology Service Marketing and Technology Executive MBA EMBA Bangkok EMBA Vietnam EMBA-HRM AIT Extension (Non-degree training, consultancy and services) Agriculture, Resources and Rural Development Business Management Development Management Education and Training Development Environment, Infrastructure and Urban Development Information and Communications Technology

. http://www.geoinfo.ait.ac.th Director : Dr Lal Samarakoon MODIS & NOAA Receiving station. QZSS Base station QZSS Base Station

Improving Real-Time PPP positioning by combining QZSS-LEX Massage and IGS Ultra-rapid products Mr. Chathura H Wickramasinghe Dr. Lal Samarkon Dr. Taravudh Tipdecho Dr. Manukid Parnichkun Geoinformatics Center Asian Institute of Technology

Content Multi GNSS Introduction QZSS Real Time Positioning Methods Research Question & Objectives PPP Positioning sensitivity Methodology Results Conclusion & recommendation

Introduction to Multi GNSS Global Navigation Satellite System(GNSS) GPS(USA), GLONASS(RUSSIA), Galileo(EU), Compass(CHINA) : Global coverage IRNSS(INDIA): Regional coverage WASS(USA), EGNOS(EU),QZSS(JAPAN) : GPS Augmentation system GNSS Satellite availability density for 24 hours when all the systems are fully functioning. More than 30 satellites are visible Asia Oceania Region GNSS GPS GLONASS Galileo COMPASS IRNSS QZSS Source : http://www.multignss.asia/

QZSS (Quasi-Zenith Satellite System) Japanese GPS Augmentation system. The consolation is to consist of 3 QZS satellites, First satellite was launched on 11 September 2011, Full operation is expected in 2013. Aims to improve GPS accuracy in Japan Urban environment Availability enhancement Near Zenith Orbit over Japan Performance improvement L1-SAIF and experimental LEX QZS transmit signals L1C/A,L1C, L2C, L5, L1-SAIF and LEX Source: http://qz-vision.jaxa.jp/use/isqzss/index_e.html

Post processing PPP accuracy is comparable with the DGPS accuracy (Kouba 2001), Real time positioning DGPS-RTK (Differential GPS-Real Time Kinematic) Uses an base station to calculate the error(differential correction) PPP(Precise Point Positioning) International GNSS Services (IGS) provides precise GPS Satellite ephemeris, clocks and Ionospheric correction data. Segment Standard deviation Errors source (m) Space Satellite clock stability 3.0 Control Ephemeris prediction 4.2 errors Ionospheric delay 5.0 Tropospheric delay 1.5 User Receiver noise and 1.5 resolution Multipath 2.5 Other 0.5 Table 2-2 GPS positioning error budget(kaplan 1996) Accuracy of PPP based positioning will depend the accuracy of the model

PPP vs DGPS-rtk PPP No need for base station Not limited to base line distance No ground infrastructure necessary No cost for country level implementation Need GSM coverage or LEX receiver DGPS-RTK Need Base station Limited to 100km of base line Need ground infrastructure High cost is necessary for country level network Need radio receiver or GSM module Automated Farming Self driving Car Auto landing Ground truth

Real time ppp There area two methods IGS(International GNSS services) Ultra rapid products GSM receiver: Good Internet connection Experimental QZS LEX signal QZSS LEX RF receiver Comparison of the two methods QZSS Limited to Satellite visibility Ideal for rural areas & urban areas Precise navigation and location in the sea Land side monitoring in hilly areas Precise Navigation for Aviation IGS Limited to GSM coverage Suited only for urban areas Not possible Not possible Not possible

Research Question? How accurate are the QZSS LEX correction data? Applicability of the LEX data for Thailand How can the LEX and IGS data be combined, to improve Realtime PPP positioning accuracy?

PPP Positioning accuracy Comparing PPP data products based positioning IGS Final products (post processing): IGS Ultra Rapid Products (real time) : LEX data (real time)

PPP Positioning accuracy CONT..

Objective Improve the accuracy of LEX PPP correction data by fusing with IGS PPP data products, to improve accuracy of real time positioning. Thus providing accurate positioning on land or sea without the limitation of IGS data.

Scope Research will only focus on the GPS ephemeris and clock errors and will consider that there are very little or no multipath errors in the observations. For the evaluation of QZS LEX massage data, IGS final product will be considered as the most accurate data for GPS satellite ephemeris and clocks(griffiths and Ray 2009).

PPP sensitivity base on SV ephemeris & clocks How the accuracy of Ephemeris and Clocks will effect the positioning IGS Final product data was used. Constant error was added to the data. Clock correction accuracy has high sensitivity compared to Ephemeris Positioning error has exponential relationship

PPP sensitivity base on SV ephemeris & clocks cont.

Main Methodology Decode LEX data Analyze LEX & IGS Ephemeris and Clock correction data. Develop New Algorithm to fuse the data. Testing the New Algorithm

Decoding QZSS LEX massage LEX Massage In binary format Extracted Massage ID 10 data from LEX Massage

Extracted IGS data IGS data products are freely available online in SP3 format. http://igscb.jpl.nasa.gov/components/compindex.html IGS sp3 format New format

Analysis of LEX data Compare LEX SV (space vehicle) ephemeris and clock correction data with IGS products Daily error for all 30 GPS satellites Absolute Error for period of one week for all 30 GPS satellites PPP positioning analysis LEX based positioning LEX Corrected positioning IGS based positioning

IGS & LEX ephemeris data correlation Correlation coefficient between IGS Final products and LEX data quantifies the relationship between the two data sets. Ephemeris & Clock correction correlation coefficient Satellite SV X SV Y SV Z direction direction direction SV Clock 16 0.98 0.98 0.98 0.21 02 0.98 0.92 0.89 0.04 12 0.96 0.98 0.95 0.02 28 0.98 0.98 0.98 0.15 Correlation High Low

LEX data ephemeris Error LEX X- axis error for SV 06 on Day 01 LEX data average ephemeris error is 1m

LEX axis Error variation throughout one week (SV-6 X) How the LEX X axis error varies through the week High Correlating Ephemeris error correlation coefficient SV X- axis SV Y - SV Z -axis axis Day 1 (6 th May ) 0.985 0.993 0.984 Day 2 (7 th May ) 0.986 0.996 0.986 Day 3(8 th May ) 0.987 0.994 0.981 Day 4(9 th May ) 0.982 0.991 0.981 Day 5(10 th May ) 0.975 0.989 0.969 Day 6(11 th May ) 0.951 0.983 0.951

LEX SV-6 Y axis Error variation throughout one week

LEX SV-6 Z axis error variation throughout one week Ephemeris error in LEX data is systematic and vary with time. Correction at the beginning of the week can be applied to the rest of the week

Polynomial fitting to remove the error Using Polynomial fitting to remove the error 14 th Order Polynomial for X,Y directions and 12 th Order for Z axis was analyzed as the best fit for all the satellites. First day error is used to calculate the coefficient values of the polynomial X =-1.37 10 13 t 11 +.363 10 11 t 10 2.081096 10 8 t 9 + 3.0215338 10 7 t 8 + 3.22442312 10 6 t 7 2.419875709 10 6 t 6 + 1.84744912 10 5 t 5 3.1976905952 10 4 t 4 + 1.77068451 10 5 t 3 1.321799620 10 4 t 2 2.051067296 10 4 t + 2.221375947762 10 3 X = poly fit equaiton forsv6 X axis t = daily time in seconds starting at 00.00 ours daily X 10-3 X 10-4 Error after Polynomial fitting

Wavelet Analysis First approximation of Wavelet analysis was used to reduce the error even further Error Wavelet first approximation

Improvement in ephemeris error day 3: x-axis ephemeris error in SV 6 & 8 SV 6 Day 3 X axis error Poly fit + Wavelet has failed to improve the ephemeris SV 8 Day 3 X axis error Over all error is reduction is 0.4m

IMPROVEMENT IN EPHEMERIS ERROR DAY 3: Y-AXIS EPHEMERIS ERROR IN SV 6

IMPROVEMENT IN EPHEMERIS ERROR DAY 3: Z-AXIS EPHEMERIS ERROR IN SV 6 Poly fit Has failed to remove any errors More depth analysis of SV data must be carried out Because of the time constrain it was decided to use poly fit+ Wavelet

Five day absolute mean error value for Y axis To evaluate the performance of the new method through the week Error in LEX data due to outliers Error in correction Unable to correct outlier Over all 78% of satellites improved in the X, Y axis and 82% in the Z axis High Improvement in the ephemeris

Five day absolute mean error value for x, z axis

Single day improvement on selected satellites 7 th of May 2012 X direction absolute mean error improvement shows the error was reduced to 0.2m from 0.8m, Improvement of 24%. < 0.8 m < 0.2 m

Ephemeris improvement: quantitative Five day absolute mean error for selected satellites ephemeris show the correction method was effective in reducing the error through the week. Five day absolute mean error value in X direction (km) SV LEX Poly fit Poly fit + Wavelet 02 23.033 5.023 x10-4 4.764 x10-4 03 3.1591x0-3 7.773 x10-4 5.11 x10-4 08 6.362 0.338 4.932 x10-4 13 1.324x10-3 4.341 x10-4 4.318 x10-4 20 1.520x10-3 4.295 x10-4 4.259 x10-4 Very low values after correction Correction method was able to reduce the error to around 0.4m compared to 1m error in LEX data.

Clock correction error variation Low correlation between the IGS and LEX clock is due to the fact the LEX clock correction is an constant value rather than the actual changing value Gradually changing clock correction Constant value for LEX clock correction

Clock correction New equation was developed for clock correction Based on the analysis of the LEX Clock correction data c = C IGS + g t c = SV clock correction C IGS = Initial IGS clock correction g = Calculated IGS clock drifting gradient t = Obervation time from C IGS initial epoc SV 00.00 to 01.00 hours Day 3 IGS Clock error drift (microseconds) 11.30-12.30 hours 23.00-24.00 hours 02-0.0097-0.0096-0.0108 16 0.0105 0.0099 0.0109 20-0.0004-0.0005 0.0009 28-0.0299-0.0291-0.0328 Clock correction has almost the same gradient through the day

New Clock correction Accuracy Results show the new method has kept an a accuracy of 0.01 microseconds After 6 days. Where as LEX error is one microsecond SV LEX Clock Correction error :Day 04 (Microseconds ) Clock correction accuracy comparison New Method error : Day 4 (Microseconds) Clock correction value on 11 th (Day 6) May 2012 @ 1:00:00h (Microseconds) IGS New method LEX 02 0.3651 0.0009 389.768 389.777 390 16 0.0393 0.0003-234.436-234.434-234 20 0.3107 0.0007 63.343 63.334 63 28 Starting 0.4068 day for the week is 60.0019 th May 2012 142.688 142.687 143

New Algorithm for LEX data correction LEX Data Ephemeris Data IGS data Post process Calculate Error IGS Clock Correction Outlier remover Generate error Polynomial Error Correction Generate Wavelet Approximation SV clock drift Clock Correction drift calculation Real Time Process Ephemeris Data Real time LEX Correction New SV Ephemeris & Clocks PPP Observation time Observation data NEW POSITION

Positioning improvement cont... SV 10 12 14 17 21 30 has not been taken for analysis as they did not improve accuracy In the ephemeris Day 2 Corrected LEX and LEX based PPP positioning comparison Corrected LEX LEX Root Mean Square Error Correct ed (m) LEX (m) E 0.07 0.16 N 0.07 0.10 U 0.77 0.82 2D 0.21 0.48 12 hour period with 15 munities interval

Positioning improvement cont... Day 3 Corrected LEX and LEX based PPP positioning Corrected LEX LEX Root Mean Square Error Correct ed(m) LEX (m) E 0.10 0.20 N 0.11 0.08 U 1.68 0.88 2D 0.31 0.44

Positioning improvement cont... Day 4 Corrected LEX and LEX based PPP positioning Corrected LEX LEX Corrected LEX LEX Root Mean Square Error Correct LEX ed (m) (m) E 0.10 0.1 N 0.13 0.04 U 1.05 0.54 2D 0.32 0.22 12 hour period with 15 munities interval

Positioning improvement cont... Day 5 Corrected LEX and LEX based PPP positioning Corrected LEX LEX Root Mean Square Error Correct ed (m) LEX (m) E 0.04 0.22 N 0.10 0.10 U 0.16 0.63 2D 0.22 0.29 12 hour period with 15 munities interval

Positioning: LEX CORRECTED & IGS Day 5 IGS Ultra rapid and Corrected LEX based PPP positioning IGS Ultra rapid Corrected LEX Root Mean Square Error Corre cted (m) IGS (m) LEX (m) E 0.04 0.03 0.22 N 0.10 0.01 0.10 U 0.16 0.03 0.63 2 D 0.22 0.07 0.29 12 hour period with 15 munities interval

Positioning improvement cont... Day 6 Corrected LEX and LEX based PPP positioning comparison for Day 6 Corrected LEX LEX Root Mean Square Error Correct ed (m) LEX (m) E 0.06 0.23 N 0.06 0.01 U 0.14 0.3 2D 0.18 0.46 12 hour period with 15 munities interval

Positioning accuracy improvement Day 5 IGS Ultra rapid, LEX and LEX Corrected positioning error Root Mean Square (m) Standard Deviation (m) Position IGS LEX LEX IGS Ultra LEX LEX Ultra Rapid Corrected Rapid Corrected Nothing 0.034 0.044 0.094 0.009 0.013 0.021 Easting 0.006 0.100 0.225 0.003 0.046 0.009 2D 0.069 0.219 0.487 - Elevation 0.026 0.163 0.630 0.007 0.110 0.050 New algorithm has improved the 2D positioning accuracy by 0.15m It improve the Elevation accuracy with error of 0.47m Weekly average positioning errors for LEX and LEX Corrected data Positioning error RMS (M) LEX Corrected LEX Day 2D Height 2D Height 02 0.206 0.7745 0.3786 0.8174 03 0.311 1.6844 0.4394 0.8852 04 0.3218 1.0537 0.2156 0.5423 05 0.2192 0.1635 0.4876 0.6304 06 0.1771 0.1413 0.4654 0.3347 Averag e 0.24702 0.7634 0.3973 0.642 LEX Corrected has high standard deviation compared to LEX New algorithm kept the improved 2D positioning accuracy by 0.15m But the average elevation is better in LEX due to 40% of corrected data showed high error.

Software Tools Developed LEX Massage decoder & data extractor. LEX to SP3 converter. LEX data Corrector. LEX Massage Generator.

Conclusion New Algorithm developed by fusing LEX, IGS data improved the over all LEX PPP positioning accuracy by 37%. The New Algorithm has improved the LEX based positioning The new algorithm was able to effectively identify and remove outliers in the LEX data. Analysis showed IGS Ultra Rapid data provides the best possible real time positioning solution for PPP given that GSM coverage is available.

Multi-GNSS Joint Experiment QZSS based real time PPP observation for slow moving flood detection

Flood Modeling Flood Models : HEC-RAS, MIKE11, and TELEMAC-2D Flood Modeling Accuracy Accuracy of the DEM Number of river gaging stations Changing ground situation http://www.halcrow.com/isis/training_isis2d.asp

Research Question How to improve the flood modeling accuracy using GNSS? Accurate DEM. Changes with sand bags. Flood height data. Real ground situation

Objective First stage Identify measurable minimum vertical movement using real time PPP(Precise Point Positioning) QZSS-LEX. Develop algorithm to accurately calculate the flood water increasing rate using QZSS-LEX based PPP. Final Target Develop real time flood monitoring sensor network and new flood model that can use the data for real time dynamic flood modeling.

Why use PPP PPP No need for base station Not limited to base line distance No ground infrastructure necessary No cost for country level implementation DGPS-RTK Need Base station (can be flooded) Limited to 100km of base line Need ground infrastructure High cost is necessary for country level network Why QZSS LEX based PPP? Does not depend on Internet connection

Sensor GNSS receiver(qzss LEX) Height calculation GSM Module Transmit data Digital Rain Gage Rain fall data Gyroscope Detect movement Digital Camera Visual verification Micro computer To process the data and Identify accurate elevation changes

System Frame Work GPS Satellites QZSS Satellites QZSS Emergency massage Flood Modeling L1,LE X GSM Network Central Monitoring Station

Advantages Highly accurate flood monitoring. Any location Chang data gathering point according to the situation. Accurate mitigation Monitor how ground situation is deviation from the model and response. Sand bag walls will change the water flow. Centralized system. Relatively low implementation cost.

Data Collection

Initial observation data Average Water increasing rate 0.028 cm/sec Linear fit Gradient 0.025 cm/sec

Initial observation data Average Water increasing rate 0.026 cm/sec Linear fit Gradient 0.025 cm/sec

Initial observation data decreasing water Average Water increasing rate 0.008 cm/sec Linear fit Gradient 0.03cm/sec

Initial observation data Average Water increasing rate 0.008 cm/sec

Initial observation data Cont..

Next step Analyzing the GNSS Observation data. Test with low cost IMU. Develop an algorithm that will accurately identify the flooding rate.