This is the presentation I gave for my PhD defence

1. Trends in surface temperature from new long-term
homogenised thermal data by applying remote
sensing techniques and its validation using in-situ
data of five southern European lakes
Sajid Pareeth
Department of Biology, Chemistry and Pharmacy, Freie Universität Berlin
Centre for Research and Innovation, Fondazione Edmund Mach
Supervisors: Dr. Markus Neteler, Dr. Nico Salmaso,
Prof. Dr. Rita Adrian, Prof. Dr. Klement Tockner
Disputatio – 12 July 2016

2. Presentation outline
Rationale & Research questions
Objectives & Background
Study area & Data
Milestones
M1: Resolve the geometrical issues of earlier satellite data
M2: New Climate Data Record (CDR) – thirty years of daily lake
surface water temperature (1986 – 2015)
M3: Results – validation of new LSWT and trend analysis
Discussion & Future work
Thesis outcome – Publications
Acknowledgments
Introduction Outline 2 / 44

3. Rationale: Climate change perspective
Lake Surface Water Temperature (LSWT) - is a good indicator in
understanding the changes in lake characteristics
Long-term changes in LSWT could give us a cue towards
changing climate and lake’s biological properties
Globally the summer mean LSWT are reported to be significantly
increasing
Lakes’ response to long-term warming varies regionally
O’Reilly, C. M. et al. Rapid and highly variable warming of lake surface waters around the globe. Geophysical Research Letters
42, 10,773–10,781 (2015)
Adrian, R. et al. Lakes as sentinels of climate change. Limnology and oceanography 54, 2283–2297 (2009)
Introduction Motivation 3 / 44

4. Rationale: Data perspective
Need high spatio-temporal resolution data to understand the
long-term dynamics of thermal variations over lake’s surface
The in-situ data from the study lakes are spatially and temporally
coarse
Satellite data acquired at Thermal Infra-red Region(TIR) are
considered as a good alternative
Availability of daily satellite observations since 1980’s at 1 km
spatial resolution
Surface temperature is one of the accurate measurements using
remote sensing
Introduction Motivation 4 / 44

5. Research questions
How can we develop a reliable time series of daily LSWT from the
historical discrete satellite data?
What level of accuracy can be achieved with satellite derived
surface temperature over large lakes in Northern Italy?
Are the large lakes in the north of Italy warming due to climate
change and at what rates?
Introduction Research Questions 5 / 44

6. Objectives
To reconstruct thirty years (1986 – 2015) of homogenised time
series of daily LSWT for five large Italian lakes by combining
thermal data from multiple satellites
To assess the quality of the satellite derived LSWT using
long-term in-situ data
To report the summer and annual trends in LSWT using statistical
tests for last thirty years (1986 – 2015)
Introduction Objectives 6 / 44

7. Background (1/3)
Satellite sensors record the reflected energy in the optical range,
while they record the emitted energy in the thermal region
All objects with a temperature above absolute zero
(0 K or -273.15 ◦C) emit electromagnetic radiation
Introduction Background 7 / 44

8. Background (1/3)
Satellite sensors record the reflected energy in the optical range,
while they record the emitted energy in the thermal region
All objects with a temperature above absolute zero
(0 K or -273.15 ◦C) emit electromagnetic radiation
Introduction Background 7 / 44

9. Background (2/3)
Introduction Background 8 / 44

10. Background (3/3)
At-sensor radiance to Brightness Temperature(BT) using the
inverse Planck’s equation.
BT to surface temperature is calculated using Split-Window (SW)
equation.
LSWT = Ti + c1(Ti − Tj) + c2(Ti − Tj)2
+ c0 (1)
c0 - c2: Satellite specific split-window coefficients
Ti: Thermal data (BT) acquired from satellites at 10.5 - 11.5 µm
Tj: Thermal data (BT) acquired from satellites at 11.5 - 12.5 µm
Jimenez-Munoz, J.-C. & Sobrino, J. Split-Window Coefficients for Land Surface Temperature Retrieval From Low Resolution
Thermal Infrared Sensors. IEEE GREL 5, 806–809 (2008).
Introduction Background 9 / 44

11. Study lakes
Red rectangle: bounding box of processed satellite data
Introduction Study area 10 / 44

12. Thirty years of satellite data
NOAA9
NOAA11
NOAA12
ERS1
NOAA14
ERS2
NOAA16
NOAA17
Envisat
NOAA18
Terra
Aqua
NOAA19
1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012 2014
Year
Satellite
Sensor
A(A)TSR
ATSR1
ATSR2
AVHRR
MODIS
Introduction Data 11 / 44

13. Data
Thermal data from 13 satellites (five sensors)
Daily dual thermal channels
Ti [10.5 - 11.5 µm]
Tj [11.5 - 12.5 µm]
Long-term in-situ data from study lakes obtained from the
collaborators
Lake Garda – Fondazione Edmund Mach
Lake Iseo – Uni-Milano Bicocca (Dr. Barbara Leoni)
Lake Como – ARPA (Dr. Fabio Buzzi)
Lake Maggiore – ISE-CNR (Dr. Giuseppe Morabito)
Lake Trasimeno – Uni-Perugia (Dr. Alessandro Ludovisi)
Introduction Data 12 / 44

14. Daily acquisition time of satellite data
1000
1100
1200
1300
1400
1500
1600
1700
1986 1993 1997 2000 2003 2005 2006 2008 2010 2012 2015
Year
Time
Satellites
NOAA9
NOAA11
NOAA12
NOAA14
NOAA16
NOAA17
NOAA18
NOAA19
ERS1
ERS2
Envisat
Terra
Aqua
Introduction Data 13 / 44

15. Major milestones
M1: Resolve the geometrical issues of earlier AVHRR data
M2: New homogenisation method to develop daily LSWT time
series for thirty years (1986 - 2015)
M3: Validation of new LSWT and trend analysis
Miestone 1 Resolving AVHRR issues 14 / 44

16. Major milestones
M1: Resolve the geometrical issues of earlier AVHRR data
M2: New homogenisation method to develop daily LSWT time
series for thirty years (1986 - 2015)
M3: Validation of new LSWT and trend analysis
Miestone 1 Resolving AVHRR issues 15 / 44

17. Geometrical issues with AVHRR sensor data
Navigational discrepancies due to on board clock errors and
satellite orbital angular errors
Orbital drifts of the NOAA satellites
Miestone 1 Resolving AVHRR issues 16 / 44

18. Geometrical issues with AVHRR sensor data
Navigational discrepancies due to on board clock errors and
satellite orbital angular errors
Orbital drifts of the NOAA satellites
Miestone 1 Resolving AVHRR issues 16 / 44

19. Automated workflow for accurate AVHRR data
processing
New python readers: read raw data and calibrate, contributed to
the public repository
Feature matching technique deployed to extract homologous
points with respect to a reference image
Geo-rectification using the automatically extracted matching points
Miestone 1 Resolving AVHRR issues 17 / 44

20. Feature matching based geometric correction
Input – Date of acquisition - 09 Aug 1997
Red and blue points: homologous point pairs
Miestone 1 Resolving AVHRR issues 18 / 44

21. Feature matching based geometric correction
Zoomed into study lakes
Miestone 1 Resolving AVHRR issues 19 / 44

22. Feature matching based geometric correction
Output – Date of acquisition - 09 Aug 1997
Corrected image with aligned boundaries
Miestone 1 Resolving AVHRR issues 20 / 44

23. Feature matching based geometric correction
Zoomed into study lakes
Miestone 1 Resolving AVHRR issues 21 / 44

24. Validation: Geometric correction of AVHRR
sensor data
Using 12,000 random pixels from 2000 images
Sub-pixel accuracy with an overall RMSE of 755.63 m (Nominal
pixel size is 1 km)
Miestone 1 Resolving AVHRR issues 22 / 44

25. Major milestones
M1: Resolve the geometrical issues of earlier AVHRR data
M2: New homogenisation method to develop daily LSWT time
series for thirty years (1986 - 2015)
M3: Validation of new LSWT and trend analysis
Milestone 2 Homogenisation of LSWT 23 / 44

26. Work flow to homogenise LSWT time series
Milestone 2 Homogenisation of LSWT 24 / 44

27. Daily acquisition time of satellite data
1000
1100
1200
1300
1400
1500
1600
1700
1986 1993 1997 2000 2003 2005 2006 2008 2010 2012 2015
Year
Time
Satellites
NOAA9
NOAA11
NOAA12
NOAA14
NOAA16
NOAA17
NOAA18
NOAA19
ERS1
ERS2
Envisat
Terra
Aqua
Milestone 2 Homogenisation of LSWT 25 / 44

28. Diurnal Temperature Cycle Model (DTC)
Homogenisation using the monthly diurnal cycles derived from
LSWT using DTC model
Correction factor (cf) is estimated for each LSWT observation
using the respective monthly diurnal cycle
cf = abs(Ts(t) − Ts(12)) (2)
Original LSWT observation is then adjusted using the correction
factor to standardise it to 12:00 UTC
Time correction of an image taken in June at 14:00 UTC:
Milestone 2 Homogenisation of LSWT 26 / 44

29. Gap filling using harmonic analysis
Harmonic ANalysis of Time Series (HANTS) is applied to filter
outliers and fill the gaps due to cloud cover
HANTS output for the year 2003 as an example for Lake Garda
Milestone 2 Homogenisation of LSWT 27 / 44

30. Seasonal and annual climatologies
Summer mean LSWT (June/July/August) for the year 2003
Seasonal and annual mean LSWT were developed for all lakes
Milestone 2 Homogenisation of LSWT 28 / 44

31. Summer climatology of Lake Garda
Milestone 2 Homogenisation of LSWT 29 / 44

32. Major milestones
M1: Resolve the geometrical issues of earlier AVHRR data
M2: New homogenisation method to develop daily LSWT time
series for thirty years (1986 - 2015)
M3: Validation of new LSWT time series and trend analysis
Milestone 3 Validation and trends 30 / 44

33. Validation of new LSWT time series
Cross-platform LSWT’s between same day observations from a
satellite pair reported an average RMSE of 0.88 ◦C
Final homogenised and gap-filled LSWT against in-situ data
reported an average RMSE of 1.2 ◦C
Lake wise results using respective in-situ data are given below:
Name RMSE (◦C) MAE (◦C) R2 N Sampling
Lake Garda 1.06 0.83 0.98 217 Monthly
Lake Iseo 1.08 0.95 0.97 129 Monthly
Lake Como 1.14 0.96 0.96 83 Monthly
Lake Maggiore 1.13 0.97 0.97 207 Monthly
Lake Trasimeno 1.38 1.13 0.98 4392 Daily
Milestone 3 Validation and trends 31 / 44

34. Summer mean LSWT variation: 1986 – 2016
Milestone 3 Validation and trends 32 / 44

35. Summer and annual trends from 1986 to 2016
Trends estimated using Sen-slope and Mann Kendal test for
significance
Significant trends obtained for annual and summer means
Average summer and annual trends at the rate of 0.032 ◦C yr-1
and 0.017 ◦C yr-1 respectively
Lake Summer Annual
Lake Garda 0.036*** 0.02*
Lake Iseo 0.017 0.019*
Lake Como 0.032* 0.012
Lake Maggiore 0.033* 0.017
Lake Trasimeno 0.044*** 0.017
***(P < 0.001), **(P < 0.01), *(P < 0.05)
Milestone 3 Validation and trends 33 / 44

36. Temporal coherence between summer mean
LSWT
Lake Garda
21.0 22.0 23.0
0.76
*** 0.83
***
21.5 22.5 23.5 24.5
0.76
***
22.523.524.5
0.84
***
21.022.023.0
Lake Iseo
0.87
*** 0.73
*** 0.64
***
Lake Como
0.86
***
20212223
0.62
***
21.523.024.5
Lake Maggiore 0.64
***
22.5 23.5 24.5 20 21 22 23 24.5 25.5 26.5
24.525.526.5
Lake Trasimeno
Scatter plot matrix showing temporal coherence between summer
mean LSWT of all the lakes
Milestone 3 Validation and trends 34 / 44

37. Applications of new LSWT time series
Inter-lake comparisons using data from same source
High temporal frequency of new data helps to detect local
minima/maxima of LSWT during a season/month.
To study the ecological consequences due to warming
It could help in understanding the variation in timing of thermal
stratification
To understand the disappearance of large lakes and its impact to
surrounding landscapes
Climate Data Record (CDR) to various modelling frameworks
Discussion Applications 35 / 44

38. Discussion
The method is extensible and reproducible to other geographical
locations
The new LSWT time series provides an opportunity to fill the gap
due to lack of high frequent in-situ data
Good alternative for lakes with difficult accessibility
The satellites measure temperature over skinlayer (thin micro
layer between lake surface and atmosphere), while in-situ data
represents epilimnion layer
The skin layer explains the higher RMSE obtained in this study
Discussion Applications 36 / 44

39. Future work
To study the influence of larger climatic indices like North Atlantic
Oscillation (NAO) and Eastern Atlantic (EA) oscillations on the
derived seasonal means from new LSWT series
To understand the ecological impacts of the reported warming on
the study lakes.
Discussion Future work 37 / 44

40. Conclusion
A total of 62,799 images (6 TB) were processed in high
performance computer using GRASS GIS and Python
New method for developing homogenised time series of LSWT
from multiple satellites is developed
The study lakes are reported to be warming at an average rate of
0.032 ◦C yr-1 during summer and 0.017 ◦C yr-1 annually
High coherence was reported between summer mean LSWT of
study lakes
Highest coherence was reported between Lake Trasimeno and
Lake Garda during summer, depicting the higher influence of
Mediterranean climate over Lake Garda.
Discussion Conclusion 38 / 44

41. Thesis manuscripts
M1: Pareeth, S., Delucchi, L., Metz, M., Rocchini, D., Devasthale, A., Raspaud, M.,
Adrian, R., Salmaso, N., Neteler, M., 2016. New Automated method to develop
geometrically corrected time series of brightness temperatures from historical AVHRR LAC
data. Remote Sensing 8, 169.
M2: Pareeth, S., Salmaso, N., Adrian, R., Neteler, M. [In review]. Homogenized daily lake
surface water temperature data generated from multiple satellite sensors: A long-term
case study of a large sub-Alpine lake. Nature Scientific Reports.
M3: Pareeth, S., Bresciani, M., Buzzi, F., Leoni, B., Lepori, F., Ludovisi, A., Morabito, G.,
Adrian, R., Neteler M., Salmaso N. [In review], Warming trends of perialpine lakes from
homogenised time series of historical satellite and in-situ data, Science of the Total
Environment.
Discussion Publications 39 / 44

42. Conference contributions
Pareeth, S., Salmaso, N., Adrian, R., Neteler, M. New homogenized daily lake surface
water temperature data of three decades from multiple sensors confirm warming of large
sub-alpine lake Garda (Poster). In: EGU 2016, Vienna, Austria, 17-22 April 2016
Pareeth, S., Delucchi, L., Metz, M., Salmaso, N., Neteler, M. An open source framework
for processing daily satellite images (AVHRR) over last28 years. In: FOSS4G-Europe,
Como, Milan, Italy, 14-17 July 2015
Pareeth, S., Delucchi, L., Metz, M., Buzzi, Fabio., Leoni, B., Ludovisi, A., Morabito, G.,
Salmaso, N., Neteler, M.. Inter-sensor comparison of lake surface temperatures derived
from MODIS, AVHRR and AATSR thermal bands. In: 35th EARSeL symposium 2015,
Stockholm, Sweden, 15-19 June 2015
Pareeth, S., Metz, M., Rocchini,D., Salmaso,N., Adrian,R., Neteler, M. (2014). Lake
surface temperature as a proxy to climate change: Satellite observations versus multi
probe data. Poster at Climate Symposium, Darmstadt, Germany, 13-17 October, 2014
Pareeth, S., Metz, M., Neteler, M., Bresciani, M., Buzzi, F., Leoni, B., Ludovisi, A.,
Morabito, G., Salmaso, N. (2014). Monitoring and retrieving historical daily surface
temperature of sub-alpine Lakes from space. In: 15th World Lake Conference (WLC15),
ISBN: 978-88-96504-05-5.
Pareeth, S., Metz,M., Rocchini,D., Salmaso,N., Neteler,M. (2013). Warm Lakes: retrieval
of lake surface water temperature (LSWT) for large sub-alpine lakes from multiple sensor
satellite imageries. In: XXI Congresso dell’AIOL, Lignano Sabbiadoro (Ud), 23-26
September 2013: 9.
Pareeth, S., Metz,M., Rocchini,D., Salmaso,N., Neteler,M. (2013). Lake surface water
temperature (LSWT) for large sub alpine lakes from satellite sensor derived surface
temperature. In: EULAKES, Brescia, Italy, 30 May 2013
Discussion Publications 40 / 44

43. Acknowledgements
Discussion Acknowledgements 41 / 44

44. Acknowledgements
GIS and RS group members at FEM
Hydrobiology group members at FEM
Collaborators - Mariano Bresciani (IREA-CNR); Barbara Leoni
(Uni Milano Bicocca); Alessandro Ludovisi (Uni-Perugia); Fabio
Lenti (SUPSI); Fabio Buzzi (ARPA); Giuseppe Morabito
(ISE-CNR)
Flavia Zanon, Elisabetta Perini, Alessandro Gretter
CRI PhD students and colleagues
Data providers - NASA, ESA, NOAA
Grant providers - FIRST-FEM, IRSAE, European Commission,
AIOL
Open source projects - GRASS GIS, Pytroll and their wonderful
communities
Discussion Acknowledgements 42 / 44

45. THANK YOU !!!
Discussion Acknowledgements 43 / 44

46. References
Adrian, R. et al. Lakes as sentinels of climate change. Limnology and oceanography 54, 2283–2297 (2009).
O’Reilly, C. M. et al. Rapid and highly variable warming of lake surface waters around the globe. Geophysical Research
Letters 42, 10,773–10,781 (2015).
Jimenez-Munoz, J.-C. & Sobrino, J. Split-Window Coefficients for Land Surface Temperature Retrieval From Low-
Resolution Thermal Infrared Sensors. IEEE Geoscience and Remote Sensing Letters 5, 806–809 (2008).
Riffler, M., Lieberherr, G. & Wunderle, S. Lake surface water temperatures of European Alpine lakes (1989–2013) based
on the Advanced Very High Resolution Radiometer (AVHRR) 1 km data set. Earth System Science Data 7, 1–17 (2015)
Dokulil, M. T. et al. Twenty years of spatially coherent deepwater warming in lakes across Europe related to the North
Atlantic Oscillation. Limnology and Oceanography 51, 2787–2793 (2006).
Pareeth, S. et al. New Automated Method to Develop Geometrically Corrected Time Series of Brightness Temperatures
from Historical AVHRR LAC Data. Remote Sensing 8, 169 (2016).
Jin, M. & Treadon, R. E. Correcting the orbit drift effect on AVHRR land surface skin temperature measurements.
International Journal of Remote Sensing 24, 4543–4558 (2003
Discussion Acknowledgements 44 / 44