Predicting algal growth
under climate change in
the upper Thames
Mike Hutchins, CEH Wallingford
(plus Richard Williams, Christel
Prudhomme, Sue Crooks)
Changes in the Thames by 2080
• Brought about by economic and social
change but in particular climate change...
– Slower flowing
– Warmer
– Have more sunlight hours
– Have higher nutrient concentrations if only
due to less in stream dilution
• These are better environmental conditions
for phytoplankton blooms; and will favour
potentially-toxic Cyanobacteria species
... Defra policy interest
• How and when will climate change have a
discernible and significant impact on water
quality?
• Commissioning of a case study
demonstrating modelling tools and
datasets for assessing these changes:
 three Lake District lakes
 Yorkshire Ouse (focus on River Ure)
 Upper River Thames
Model chain of three main components
• Climate data from Hadley Centre’s 11-member
ensemble projection (from regional climate
modelling (RCM)) used as part of the UKCP09
scenarios. The 11 members represent a range of
model parameterisations reflecting uncertainty. All
use the SRES A1B emission scenario.
• Future Flows Hydrology (FFH) dataset. Derived via
rainfall-runoff modelling under an EA project to
provide a UK-wide consistent set of future daily
river flows.
• Water quality predictions using QUESTOR, a semiempirical, process-based model of river networks
QUESTOR river quality model (Thames)
Model inputs: (1) Flow
and quality data in (a)
tributaries (b) effluents
from sewage works,
(2) Solar radiation
Upstream QUESTOR boundary
CEH weekly water quality (2009 - )
Major urban areas outside London
Eynsham
Wallingford
LONDON
Tidal limit
Represents biochemical interactions in
the river channel environment; and
energy balance for water temperature
Blooms likely in long slow-flowing rivers
...with sufficient light, nutrients and temperature to thrive. All
these variables used in hydrological modelling at daily resolution
of chlorophyll-a, and dissolved oxygen (DO) impacts.
upper quartile chl-a (µg/L)
40
River Thames (2009-10)
35
CEH Thames Initiative data
QUESTOR model
30
Wallingford
(92 km downstream)
25
20
15
10
5
0
0
25
50
75
distance downstream (km)
Effect of increasing residence time
Chlorophyll-a content of
different types of
phytoplankton is known,
making it a useful
100
surrogate for biomass
QUESTOR calibrated in 2009-10 (e.g. Eynsham)
0
700
25
20
15
10
5
0
-5
-10
-15
-20
-25
8
600
7
100
Temp
80
60
Simulated
Observed
40
20
500
Flow
N
6
5
400
4
300
200
3
P
2
100
0
Jan-2009 Apr-2009
1
Jul-2009
Oct-2009 Jan-2010 Apr-2010
Jul-2010
0
Oct-2010 Jan-2011
Nitrogen: mg NO3-N/L
Phosphorus: µg SRP/L
Flow (m3s-1)
120
Water temp (oC)
Is model simulating physical/chemical parameters well?
For algae, good summer flow/temp simulation is critical
Model performance at Abingdon in 2009
Limitation due to light
Nutrients are in excess
Phytoplankton
biomass (mg chl-a/L)
High flows wash phytoplankton out of system
An unexplained mid- to late- summer suppression of phytoplankton is apparent
Comparing 2009 & 2010: simulated blooms similar
1. However, large variations
observed between years. Far
more phytoplankton in 2009.
So a 2009-10 model is a
compromise
2. Best fitting year-specific
models perform much
better. They are identical,
apart from having different
grazing rates
Bar charts of
upper quartile
chl-a at Wallingford
Invasive zebra mussels are abundant in the Thames. We
assume that there are good and bad years for grazers but
we don’t know why? Over-winter flow/temperature
regimes. Interactions higher up food chain
Model evaluation and future priorities
• Environmental variables well represented. Can
identify suitable temperature-controlled growth rates
for a mixed phytoplankton population. All optimised
models have doubling rates of 48 h (+/- ~ 6 h)
• By altering year-specific death rates, model can
represent magnitudes of blooms year-on-year.
• Remaining gaps in understanding:
–
–
–
–
–
controls on over-winter survival of phytoplankton grazers
reasons for late-summer phytoplankton suppression
water quality response to extreme events
how will nutrient concentrations change in the future?
what will be the impact of population growth, and changes
to management/treatment of water resources and waste?
What are impacts of flooding on water quality?
Wallingford – Dec 2012
July 2007 floods resulted in low DO (Oxford – Reading)
Many potential sources of uncertainty
Had-RM3 Perturbed Physics Ensemble Climate Model
5
PE
Rainfall
Air temp
Solar radiation
Key sources
to isolate
Bias correction
Downscaling
4
Rainfall-runoff Model
(CLASSIC, CERF)
3
Donate and scale
flows to unmodelled
tribuaries
Flow
1
Regression
Attenuation by
trees and in
water column
2
Water temp
Photosynthetically
active radiation
Pollutant
loads from
tributaries
(and STWs)
Water Quality Model (QUESTOR)
DO BOD nutrients Phytoplankton biomass (chl-a)
Uncertainty due to hydrological modelling
Only 5 of the 11 gauged tributaries were
modelled under the FFH project - so, 3 runs:
I. A baseline QUESTOR model, set up using
all available flow data (2009-10)
II. Re-run QUESTOR replacing observed flows
in un-modelled tributaries with observed
flows donated and scaled from the modelled
tributaries.
III. Re-run again, also replacing observations
with modelled flows (where possible)
Errors due to donating (II) & modelling (III) flows
1
1
Run I
Run II
Run III
0.9
0.8
0.9
0.8
0.7
0.7
0.6
0.6
0.5
0.5
0.4
0.4
0.3
0.3
0.2
0.2
0.1
0.1
0
0
Flow
DO
Eynsham
Run I
Run II
Run III
Temp
Flow
DO
Temp
Wallingford
Nash-Sutcliffe goodness-of-fit values (y-axis): impacts only small
How well is extreme water quality modelled using climate drivers?
• For RCM, 1961-90 is taken as a standard period indicative of
present day. RCM (and FFH) do not reproduce “real weather”.
• Days per year when undesirable thresholds exceeded (WFDrelevant conditions: DO < 6 mg L-1, BOD > 4 mg L-1, Temp >
25 ºC, Chl-a > 0.03 mg L-1):
Eynsham
Wallingford
DO
BOD
Temp
Chl-a
Run I (2009-10)
0
0
0
27
30 year RCM/FFH
7.1-21.7
0.2-4.3
3.7-17.5
23.2-44.7
Run I (2009-10)
0
33
1.5
99.5
30 year RCM/FFH
1.7-9.2
13.0-36.9 1.7-17.4
75.9-103.0
• When using climate model drivers the frequency of incidence
of extreme conditions is probably overestimated. Why?
Water quality is most vulnerable at low flows in summer
Flow Q95 (m3s-1)
Eynsham
Days Lock
1961-90 observed
1.17
3.36
2009-10 observed
1.44
4.02
Run I (2009-10)
2.12
4.11
Run II (2009-10)
0.68
3.06
Run III (2009-10)
1.33
3.47
30 year RCM/FFH
0.10-0.78
1.18-2.51
• Lowest flows are underestimated when using RCM/FFH
• Analysis of RCM outputs and climate records suggest
the highest air temperatures simulated by the models are
unfeasibly extreme.
• Climate model drivers suggest even in present day
conditions the Thames above Oxford is vulnerable to
drying out. This is not realistic.
Summary results
• Changes in drivers by the 2040-69 period (Wallingford):
 + 3-5 ºC 90th percentile (i.e. summer) air temperature
 + 4-10% solar radiation (70th percentile)
 - 25% Q95 flow i.e. summer low flow (range: +7.3 to -41.3)
50
Eynsham
50
45
Threshold values:
DO = 6 mg/L
BOD = 4 mg/L
Temp = 25 C
chl-a = 0.03 mg/L
45
35
30
25
20
15
35
30
20
15
10
5
5
0
0
-5
BOD
Temp
2. The bar represents
the mean of changes
seen from the 11
applications of the
model chain
25
10
DO
1. The increase
represents the future
2040-69 situation
relative to present day.
40
Increase in days per year
Increase in days per year
40
Wallingford
chl-a
DO
-5
BOD
Temp
chl-a
3. Error bars represent
the maximum and
minimum change.
Conclusions
• Whilst simulations derived from RCM applications
appear reliable across the inter-quartile range (and to a
large degree to 5th and 95th percentile levels), the most
extreme conditions are not simulated reliably.
• The future projections should not be presented as
absolute indicators of water quality, rather as a change
relative to present day conditions.
• Accelerated phytoplankton growth in future will lead to
more limitation (including self-shading) and greater risk
of blooms crashing, leading to possible DO sags.
• Uncertainty in model chain:
Climate
modelling
Water quality
modelling
Hydrological
modelling