Coupling 3D models and earth observation to
develop algae forecasting services
Miguel Dionisio Pires (Deltares)
&
Yi Hong, Lucas Jardim Porto,
D. Fürstenau Plec, B. J. Lemaire, B. Vinçon-Leite
(LEESU, Ecole des Ponts ParisTech),
Lilith Kramer, Tineke Troost (Deltares)
User days
1/11/2017
CyMonS and EOMORES
• CyMonS
• Funded by ESA
• Coupling EO, IS and model for
cyanobacteria scum forecasting
• https://business.esa.int/project
s/cymons-fs
• https://www.youtube.com/watc
h?v=RmVFPL4k5x0
• EOMORES
• H2020
• Coupling EO, IS and model for
ecological status assessment
• http://eomores-h2020.eu/
2
This project is co-funded
by the European Union
Cyanobacteria Monitoring
Services (CyMonS)
Miguel Dionisio Pires
(miguel.dionisio@deltares.nl)
Wind speed
Wind direction
Cloud cover
Relative humidity
Air temperature
Delft 3D Flow
Transport
Water Temp
Delwaq-Bloom-
EcoFuzz
Delwaq
Species biomass
Buoyancy rate
Process parameters
Vertical transport
Horizontal transport
Scum potential
(Low, med, high, v. high))
Scum potential
& location
Validation
(Cyano biomass,
presence/absence)
EcoFuzz
Wind speed
Solar radiation
Time of day
Expert rules
Wind speed
Wind direction
Cloud cover
Relative humidity
Air temperature
Delft 3D Flow
Transport
Water Temp
Delwaq-Bloom-
EcoFuzz
Delwaq
Species biomass
Buoyancy rate
Process parameters
Vertical transport
Horizontal transport
Scum potential
(Low, med, high, v. high))
Scum potential
& location
Validation
(Cyano biomass,
presence/absence)
EcoFuzz
Wind speed
Solar radiation
Time of day
Expert rules
Algae forecasting
model
Meteorological
data
Sentinel-2
WISP-3/EcoSpot &
EcoWatch
System and Service Architecture
Earth Observation-based services for
Monitoring and Reporting of Ecological Status
EOMORES
EOMORES - project overview
Details
 EOMORES is a H2020 (EC) research project
 Project time: 3 year, starting 1 December, kick off
9 & 10 January
 There are 9 partners from 6 EU countries
 Almost all (8) partners have one or several users
in their country
 13 users
6
EOMORES overview
7
User relevant
• Tailor products and services to users
• Generate higher-level products that suit their
needs
• Integrate products into users’ systems
8
1. Earth observation example
Suspended matter maps for monitoring
turbidity
EOMORES
2. In situ component
Optical in situ instruments
 Quick-scans (direct result)
 Continue, automatic monitoring
 Also used for validation and
calibration of atmospheric correction
EOMORES
Above: WISP-3
Below: fixed position instrument
2. Validation optical in-situ
observations
EOMORES
12
Models in EOMORES
EWACSAlgae
P
N
C
N
NH4-N
NO3-N
P
PO4-P
Detritus
P
N
C
settlingsettling
respiration
photosynthesis
Nutrient
mineralisation
mineralisation
metabolism
mortality
DO
production
consumption
reaeration
Detritus in Sediment
C N P Si
Si
Si
N2 denitrification
mineralisation & nitrification
autolysis
Si
consumption
nitrification
Grazers
grazing
grazing
oxygen
consumption
biodeposition
AIP adsorption
Microphytobenthos
C N P Si
AIP in
sediment
settling
mortality
photosynthesis
BLOOM
Delft3D-WAQ
Algae
Radar
Monitoring and forecasting the cyanobacteria blooms in lakes
13
14
Lake Champs-sur-Marne - France
(Geoportail, IGN, 2017)
(Photos D. Plec, 2016)
Surface area 0.12 km2
Average depth 2.3 m
Maximum depth 4 m
Cyanobacteria
in Lake Champs-sur-Marne
15
Since 2006 Survey according to the French bathing regulation
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Monthly Bi-weekly Weekly Bi-weekly|Monthly
2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017
Phytoplankton counting and Chlorophyll Relative abundance
16
Main phytoplankton species
Anabaena Aphanizomenon Microcystis Ceratium Peridinium
Nostocales
DinoflagellatesCyanobacteria
(Peiffer, 2016)
Chroococcales
High frequency monitoring
2015 2016 2017
OSS-Cyano
17
0.5 m
1.5 m
2.5 m
Temperature
Dissolved Oxygen
Chlorophyll-a
Conductivity
A
B
C
Since 2015
5-minute time step
18
A
B
C
2015 2016 2017
OSS-Cyano
Since 2015
Vertical profile measurements
Surface
Bottom
Total Chl-a
Blue Green
Green
Diatoms
Cryptophyta
Temperature
Dissolved Oxygen
Conductivity
pH
BBE Fluoroprobe
Seabird CTD
Licor photometer
19
Phytoplankton concentration and temperature
at point A in May – Dec. 2015
Simulation period - July 2015
20
Model configuration
• Mesh
– Measured bathymetry
– 10 x 10 x 0.33 m cells, 10 layers
• Meteorological data
– Meteo-France Orly airport station
• Parameters
– Hydrodynamic model: drag coef. 0.0013;
Dalton coef. 0.0015; Stanton coeff. 0.00145;
horizontal viscosity and diffusivity 0.0025
m²/s;
– Calibration parameters: wind reduction
factor; albedo
– Biological model: Settling velocities; default
parameters
• Biological variables
– 4 phytoplankton groups
• Initial conditions
– Hydrodynamic model
– Biological model
21
Initial condition for phytoplankton groups
22
4 phytoplankton variables:
• Green algae; Cyanobacteria; Diatoms and
Dinoflagellates.
The initial concentrations of the 4
phytoplankton groups:
• Total chlorophyll measured continuously
• Spectrofluorometer vertical profile;
2015 5 mg/l
Hydrodynamics results
R2 = 0.63 - 0.89
MAE= 0.30 - 0.62°C
MRE < 3 %
23
The performance indicators
were calculated using the
hourly mean of the
measured data
data
model
• n=342
• MAE=5.8 mg/L
• R2=0.46
• RE= 42%
24
Total phytoplankton biomass at 1.5 m depth
13 - 27 July 2015
Phytoplankton groups
25
Two components:
1. Hydrodynamics Model: Delft3D-FLOW
2a. Water quality/Ecology Model:
Delft3D-WAQ (DELWAQ engine)
2b. Fuzzy logic model coupled to WQ
model
Model output:
• Map of cyanobacteria biomass in surface
waters for each model time-step
• Used to generate weekly scum bulletin
Flow Model
Transport
Water Temp.
Water Quality model
Vertical transport
Horizontal transport
Scum potential
(app/disappearance)
Scum presence
& location
Fuzzy Logic model
Set up EWACS model
HHNK 30 maart 2016
Delft3D - EWACS Model
27
EWACS Early Warning Against sCumS
Logical inference used to predict scum appearance and
disappearance in EcoFuzz (taken from Burger et al., 2008)
Delft3D - EWACS Model
28
Delft3D - EWACS Model
29
Delft3D - EWACS Model
30
Delft3D-BLOOM - EWACS Model
31
Next steps
32
• Validation on other periods – September 2017
Conclusion and Perspectives
33
• High-frequency monitoring system is a promissing approach to
improve model performance
• Firstly applied Ewacs – BLOOM coupling on small urban lake;
• The only scum forecasting model in the world
• Next steps …
• Collection of nutrient data
• Implementation and continue to improve over other lakes;
• Validate the outcomes of the model with satellite images;
• Real Time forecasting system
Thank you!!!

DSD-INT 2017 Coupling 3D models and earth observation to develop algae forecasting services - Dionisio Pires

  • 1.
    Coupling 3D modelsand earth observation to develop algae forecasting services Miguel Dionisio Pires (Deltares) & Yi Hong, Lucas Jardim Porto, D. Fürstenau Plec, B. J. Lemaire, B. Vinçon-Leite (LEESU, Ecole des Ponts ParisTech), Lilith Kramer, Tineke Troost (Deltares) User days 1/11/2017
  • 2.
    CyMonS and EOMORES •CyMonS • Funded by ESA • Coupling EO, IS and model for cyanobacteria scum forecasting • https://business.esa.int/project s/cymons-fs • https://www.youtube.com/watc h?v=RmVFPL4k5x0 • EOMORES • H2020 • Coupling EO, IS and model for ecological status assessment • http://eomores-h2020.eu/ 2 This project is co-funded by the European Union
  • 3.
    Cyanobacteria Monitoring Services (CyMonS) MiguelDionisio Pires (miguel.dionisio@deltares.nl)
  • 4.
    Wind speed Wind direction Cloudcover Relative humidity Air temperature Delft 3D Flow Transport Water Temp Delwaq-Bloom- EcoFuzz Delwaq Species biomass Buoyancy rate Process parameters Vertical transport Horizontal transport Scum potential (Low, med, high, v. high)) Scum potential & location Validation (Cyano biomass, presence/absence) EcoFuzz Wind speed Solar radiation Time of day Expert rules Wind speed Wind direction Cloud cover Relative humidity Air temperature Delft 3D Flow Transport Water Temp Delwaq-Bloom- EcoFuzz Delwaq Species biomass Buoyancy rate Process parameters Vertical transport Horizontal transport Scum potential (Low, med, high, v. high)) Scum potential & location Validation (Cyano biomass, presence/absence) EcoFuzz Wind speed Solar radiation Time of day Expert rules Algae forecasting model Meteorological data Sentinel-2 WISP-3/EcoSpot & EcoWatch System and Service Architecture
  • 5.
    Earth Observation-based servicesfor Monitoring and Reporting of Ecological Status EOMORES EOMORES - project overview
  • 6.
    Details  EOMORES isa H2020 (EC) research project  Project time: 3 year, starting 1 December, kick off 9 & 10 January  There are 9 partners from 6 EU countries  Almost all (8) partners have one or several users in their country  13 users 6
  • 7.
  • 8.
    User relevant • Tailorproducts and services to users • Generate higher-level products that suit their needs • Integrate products into users’ systems 8
  • 9.
    1. Earth observationexample Suspended matter maps for monitoring turbidity EOMORES
  • 10.
    2. In situcomponent Optical in situ instruments  Quick-scans (direct result)  Continue, automatic monitoring  Also used for validation and calibration of atmospheric correction EOMORES Above: WISP-3 Below: fixed position instrument
  • 11.
    2. Validation opticalin-situ observations EOMORES
  • 12.
    12 Models in EOMORES EWACSAlgae P N C N NH4-N NO3-N P PO4-P Detritus P N C settlingsettling respiration photosynthesis Nutrient mineralisation mineralisation metabolism mortality DO production consumption reaeration Detritusin Sediment C N P Si Si Si N2 denitrification mineralisation & nitrification autolysis Si consumption nitrification Grazers grazing grazing oxygen consumption biodeposition AIP adsorption Microphytobenthos C N P Si AIP in sediment settling mortality photosynthesis BLOOM Delft3D-WAQ Algae Radar
  • 13.
    Monitoring and forecastingthe cyanobacteria blooms in lakes 13
  • 14.
    14 Lake Champs-sur-Marne -France (Geoportail, IGN, 2017) (Photos D. Plec, 2016) Surface area 0.12 km2 Average depth 2.3 m Maximum depth 4 m
  • 15.
    Cyanobacteria in Lake Champs-sur-Marne 15 Since2006 Survey according to the French bathing regulation Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Monthly Bi-weekly Weekly Bi-weekly|Monthly 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 Phytoplankton counting and Chlorophyll Relative abundance
  • 16.
    16 Main phytoplankton species AnabaenaAphanizomenon Microcystis Ceratium Peridinium Nostocales DinoflagellatesCyanobacteria (Peiffer, 2016) Chroococcales
  • 17.
    High frequency monitoring 20152016 2017 OSS-Cyano 17 0.5 m 1.5 m 2.5 m Temperature Dissolved Oxygen Chlorophyll-a Conductivity A B C Since 2015 5-minute time step
  • 18.
    18 A B C 2015 2016 2017 OSS-Cyano Since2015 Vertical profile measurements Surface Bottom Total Chl-a Blue Green Green Diatoms Cryptophyta Temperature Dissolved Oxygen Conductivity pH BBE Fluoroprobe Seabird CTD Licor photometer
  • 19.
    19 Phytoplankton concentration andtemperature at point A in May – Dec. 2015
  • 20.
    Simulation period -July 2015 20
  • 21.
    Model configuration • Mesh –Measured bathymetry – 10 x 10 x 0.33 m cells, 10 layers • Meteorological data – Meteo-France Orly airport station • Parameters – Hydrodynamic model: drag coef. 0.0013; Dalton coef. 0.0015; Stanton coeff. 0.00145; horizontal viscosity and diffusivity 0.0025 m²/s; – Calibration parameters: wind reduction factor; albedo – Biological model: Settling velocities; default parameters • Biological variables – 4 phytoplankton groups • Initial conditions – Hydrodynamic model – Biological model 21
  • 22.
    Initial condition forphytoplankton groups 22 4 phytoplankton variables: • Green algae; Cyanobacteria; Diatoms and Dinoflagellates. The initial concentrations of the 4 phytoplankton groups: • Total chlorophyll measured continuously • Spectrofluorometer vertical profile; 2015 5 mg/l
  • 23.
    Hydrodynamics results R2 =0.63 - 0.89 MAE= 0.30 - 0.62°C MRE < 3 % 23 The performance indicators were calculated using the hourly mean of the measured data data model
  • 24.
    • n=342 • MAE=5.8mg/L • R2=0.46 • RE= 42% 24 Total phytoplankton biomass at 1.5 m depth 13 - 27 July 2015
  • 25.
  • 26.
    Two components: 1. HydrodynamicsModel: Delft3D-FLOW 2a. Water quality/Ecology Model: Delft3D-WAQ (DELWAQ engine) 2b. Fuzzy logic model coupled to WQ model Model output: • Map of cyanobacteria biomass in surface waters for each model time-step • Used to generate weekly scum bulletin Flow Model Transport Water Temp. Water Quality model Vertical transport Horizontal transport Scum potential (app/disappearance) Scum presence & location Fuzzy Logic model Set up EWACS model HHNK 30 maart 2016
  • 27.
    Delft3D - EWACSModel 27 EWACS Early Warning Against sCumS Logical inference used to predict scum appearance and disappearance in EcoFuzz (taken from Burger et al., 2008)
  • 28.
  • 29.
  • 30.
  • 31.
  • 32.
    Next steps 32 • Validationon other periods – September 2017
  • 33.
    Conclusion and Perspectives 33 •High-frequency monitoring system is a promissing approach to improve model performance • Firstly applied Ewacs – BLOOM coupling on small urban lake; • The only scum forecasting model in the world • Next steps … • Collection of nutrient data • Implementation and continue to improve over other lakes; • Validate the outcomes of the model with satellite images; • Real Time forecasting system
  • 34.