Climatic variability and trends in thesurface waters of coastal BC

Patrick Cummins and Diane MassonInstitute of Ocean Sciences, DFO

Acknowledgments: Peter Chandler & Mike Foreman

Introduction• Data collected on SST and SSS at lighthouse stations along the coast

of British Columbia represent some of the longest records available from the coastal waters of Canada.

• We use these records to characterize the climatic variability and trends over the region.

• Previous examination of these data given in, for example, Freeland (1990, 2013), Freeland et al. (1997), Emery & Hamilton (1985), McKinnell & Crawford (2007), Moore et al. (2008; only Race Rocks ).

• Special attention is paid to the representation of the variability using low-order (AR-1) climate models.

• Such models have previously been considered to have limited applicability to the coastal ocean (e.g., Hall & Manabe, 1997).

Lighthouse stations (monthly means)

1. Amphitrite Pt – 78 yrs2. Kains Is.– 78 yrs3. Entrance Is. – 76 yrs4. Pine Is. – 76 yrs5. Langara Is. – 73 yrs 6. Race Rocks – 71 yrs7. Bonilla Is. – 53 yrs8. Chrome Is. – 50 yrs

Met buoys+ Bakun upwelling index

Wind observations

Freshwater fluxes

Gold River (proxy for coastal freshwater flux)Fraser River

Estevan Pt (rain)

Data processing

1. Monthly anomalies are formed by removing climatological

mean for 1981-2010.

2. Except where noted, no smoothing of data is done.

3. Detrended anomalies are formed by removing least-

square fitted linear trends.

4. EOFs of lighthouse temperature (T) and salinity (S) used

to characterize coast-wide, regional variability, and related

to climate indices.

5. Cross-correlation analysis used to relate SST and SSS to

forcing (wind & freshwater flux).

6. Long and shorter trends are examined.

Leading principal component (smoothed) from EOF analysis

• Coastal BC SSTs co-vary with the PDO and large scale, NE Pacific SSTs.• SSS is not linked to large-scale climate indices.• PC1SST and PC1SSS only weakly related.

SST SSS

PDO r = 0.72

Power spectrum of leading principal components

4.6

2 2 2( )

4

CS f

f

Dotted line gives best-fit red-noise power spectrum:

De-correlation time scale:

SST

SSS

months

months

Relation to freshwater discharge and alongshore wind stress

• Model equation: with

- the SST or SSS anomalies - a damping time scale - forcing under consideration (e.g., freshwater discharge, alongshore wind stress) - additional, uncorrelated noise forcing (e.g., eddy noise)

• The ‘null hypothesis’ for climate studies: low frequency variability in ocean variable due to integration of noisy weather fluctuations. Two time-scale assumption.

• Response has a red noise spectrum, given white noise forcing

• In discrete form we have: , a first-order, auto-regressive process (AR-1),

with (for example) . . This is a measure of the ‘memory’ of the process.

X

XX F G

t

1X

( )F t

( )X t

( )G t

11 nn n nX X F G

1 (2 ) ( 2 )X Xt t

( )X t

Stochastic climate model (Hasselmann, 1976)

AR-1 model: lagged cross-correlation

• Defined as : with - discrete time lag, positive for forcing leading the response - standard deviations of and • For white noise forcing we have

where and

• Cross-correlation has a highly asymmetrical form• Inclusion of additional white noise forcing (G) reduces the amplitude, but does not affect the form of• In the following, is determined from the lag-1 autocorrelation of the SST or SSS anomalies and we compare with the

form of .

( )XF X Fn nX F

,X F F

X

1 , 0,1,2,

0, 1, 2,XF A

2 2 2 211 1 G FA 1 1XX

XF 1

( )XF

Forcing autocorrelations

5 10

1, 0

0, 0FF

For white noise

•Gold River is fed by a rainfall dominated (pluvial) watershed, and anomalies are well correlated with rainfall at Estevan Pt. (r=0.58, p<<0.01).•WCVI – a rainy coast. Gold Rv. is taken as a proxy for freshwater runoff along the WCVI.•Autocorrelation shows that discharge anomalies are well represented as a white noise process

•Fraser Rv. drains a snowmelt-driven (nival- glacial) watershed; discharge anomalies have significant autocorrelation (not white noise).

•Autocorrelation structure of alongshore wind stress is well approximated by white noise.

Influence of coastal freshwater on SSS

West Coast of Van. Is. Strait of Georgia

SSS – Gold Rv.AR-1

SSS – Fraser Rv.

• On the WCVI, SSS variability is consistent with integration of noisy freshwater discharge, as in AR-1.

• In the Strait of Georgia, SSS is highly correlated to Fraser Rv. discharge. Because the river is snow-melt dominated, the response differs from the cross-correlation of a white noise driven AR-1 process.

Integrate AR-1 model to hindcast SSS time series: 11

GR

n n nSSS SSS Q

• Captures low-frequency variability reasonably well

Relation with alongshore wind stress

• Relation to lighthouse SST anomalies is similar for the (sign-reversed) Bakun index and wind stress from Buoy 205.

• SST variability at Langara Is. is

accounted for, in part, as an integration of noisy alongshore wind stress, according to an AR-1 process.

Langara Is. SSTvs. –Bakun Index

Langara Is. SST vs. Buoy 205 winds

AR-1

AR-1

1 0.67

1 0.67

1% level

Relation with alongshore wind stress (cont’d)• AR-1 model is not unreasonable

for SST and SSS at Amphitrite Pt. and Kains Is. on the outer coast.

• Bakun index is correlated with Gold

Rv. discharge (r=-0.29).

• To isolate the influence of wind stress, a modified Bakun Index was constructed in which the freshwater-related component is removed.

• The SST-wind stress relation is essentially unaffected, but the SSS-wind stress cross-correlation is no longer significant.

AR-1

AR-1

Amphitrite Pt. SSTvs. –Bakun Index &

vs. modified B.I.

Amphitrite Pt. SSS vs. –Bakun Index &

vs. modified B.I.

AR-1

AR-1

1 0.67

1 0.45

Seasonal correlations: Alongshore wind stress and SSS on WCVI

  Winter (JFM) Spring (AMJ) Summer (JAS) Fall (OND)

-0.24 -0.46 -0.11 -0.14

-0.29 -0.54 -0.32 -0.20(*)

-0.09 -0.28 0.04 0.10

-0.16 -0.40 -0.25 0.04

l̂

APˆ,SSSr l

KIˆ,SSSr l

MOD APˆ ,SSSr l

MOD KIˆ ,SSSr l

Removing influence of fresh water forcing

• Spring is the only season with a meaningful relation between alongshore winds and SSS, likely reflecting variations in timing of onset of upwelling season in Spring.• Removing freshwater influence weakens this relation, especially at Amphitrite Pt. station.• Nearshore signal associated with upwelling appears generally weak.

Summary • Leading principal component (PC1) of SST variability represents an index of variability

for coastal BC waters.• This PC1 is very well correlated with the PDO index, which is the leading PC of SST

variability over the entire extra-tropical N. Pacific.• SSS anomalies have smaller spatial scales than SST and appear to be locally driven,

displaying a clear relation with run-off anomalies. • Along the WCVI, salinity anomalies are consistent with integration of forcing by white

noise freshwater flux anomalies as an AR-1 process.• Along the outer coast, SST anomalies also appear to integrate noisy atmospheric forcing

represented by longshore winds. However, influence of upwelling on nearshore SSS is relatively weak.

• BC coastal waters are warming (0.89 oC/century) and generally freshening.• On time scales of concern to the management of marine resources, natural variability

can easily overwhelm secular trends associated with global warming. This variability has a white spectrum at low frequencies.

Seasonal correlations: SST and alongshore wind stress

  Winter (JFM) Spring (AMJ) Summer (JAS) Fall (OND)

0.60 0.06 0.01 0.45

0.52 -0.02 0.21 (*) 0.47

0.51 -0.04 0.37 0.46

0.48 -0.18 -0.19 0.38

l̂

APˆ,SSTr l

KIˆ,SSTr l

EIˆ,SSTr l

LIˆ,SSTr l

Bold: significant at the 1% (*5%) level

• Seasonal relation between SST and wind stress is strongest in Winter & Fall, weak to non-existent in Spring and Summer.• Poleward winds (+’ve wind stress) in Winter drives warm water poleward. This also are associated with warm air masses and enhanced air/sea heat fluxes.• In Spring and Summer these effects tend to cancel as poleward winds (cyclonic air flow) are associated with cold air.• The similar relation seen at Entrance Island in the Strait of Georgia suggests that alongshore advection is not the only process involved.

n the Strait of Georgia, SSS follows Fraser River discharge anomalies throughout the year.

• Positive SST anomalies in Strait during Summer associated with reduced Fraser Rv. streamflow.• Here the strongest relation is between SSTs during spring and streamflow anomalies in the following Summer: r(SSTAMJ, QJAS) = -0.50.

Mean annual climatology

Station-pair correlations

SST correlations generally larger than SSS, indicating larger spatial scales of variability for temperature.

SST

SSS

All entries significant at the 1% level, except (*) – 5% level, and (ns) entry.

 Langara

Is.

Bonilla

Is.

Pine

Is.

Kains

Is.

Amphitrite

Pt.

Race

Rocks

Entrance

Is.

Chrome

Is.

Langara Is. 1 0.79 0.71 0.75 0.62 0.52 0.42 0.29

Bonilla Is. __ 1 0.71 0.73 0.63 0.52 0.35 0.26

Pine Is. __ 0.18 (*) 1 0.81 0.75 0.68 0.48 0.37

Kains Is. __ 0.20 0.36 1 0.75 0.62 0.50 0.38

Amphitrite Pt. __ 0.16 (*) 0.33 0.59 1 0.64 0.42 0.39

Race Rocks __ 0.08 (ns) 0.30 0.27 0.25 1 0.54 0.43

Entrance Is. __ 0.26 0.36 0.36 0.32 0.42 1 0.76

Chrome Is. __ 0.14 (*) 0.24 0.37 0.27 0.55 0.63 1

Long-term trends

StationStart year,

duration (yrs)

SST trend

(ºC century –1)

SSS trend

(psu century –1) σT (°C) σS (psu)

Amphitrite Pt. 1935, 78 0.71 ± 0.53 -0.82 ± 0.57 0.77 0.89

Kains Island 1935, 78 0.43 ± 0.74 -0.47 ± 0.54 0.80 0.82

Entrance Island 1937, 76 1.32 ± 0.96 0.70 ± 1.38 0.79 1.23

Pine Island 1937, 76 0.74 ± 0.81 -0.66 ± 0.42 0.61 0.32

Langara Island 1940, 73 0.75 ± 0.86 __ 0.78 __

Race Rocks 1942, 71 1.39 ± 0.68 -0.38 ± 0.53 0.52 0.38

Bonilla Island 1960, 53 0.72 ± 0.95 0.14 ± 0.76 0.73 0.44

Chrome Island 1963, 50 3.38 ± 0.91 0.43 ± 1.56 0.72 0.91

TS

Overall warming and offshore freshening consistent with Freeland (1990, 2013) and Freeland et al. (1997). Average SST trend for the 6 longest records is 0.89±0.62 oC/century.

Bold: significant at the 5% level

20-year running trends in lighthouse SST data

Based on SST anomalies with trend retained

Average histograms for SST trends

Probability of trend ≤ 0over 20 years: 39%over 30 years: 34%over 40 years: 17%

Similar behaviour seen in results from climate models, albeit with weaker variability (Easterling& Wehner, 2009).