The Columbia River estuary and plume: Natural variability, anthropogenic change and physical habitat for salmon PhD Candidate: Michela Burla Research Advisor:

The Columbia River estuary and plume:

Natural variability, anthropogenic change and physical habitat for salmon

PhD Candidate: Michela Burla

Research Advisor: Antonio M. BaptistaCenter for Coastal Margin Observation and Prediction, OHSU

Committee:

Edmundo Casillas, NOAA Fisheries

Daniel L. Bottom, NOAA Fisheries

Tawnya Peterson, CMOP, OHSU

The Columbia River

3

Climate variabilityand change

1800s

1930s-70s

Late1800s -

3,200-10,500 m3s-1

2001: 1,800 m3s-1

1996: 24,500 m3s-1

Salmon in the ecology, economy and culture of the Pacific NW

4

85% of Oregonians want salmon to be saved:

35% part of NW heritage36% measure of region’s environmental health15% commodity value

(The Oregonian, Dec 1997)

Columbia River Basin Salmon

5

Habitat degradation from mining, logging,

irrigation

Damdevelopment

(Lichatowich, 1999)

Salmon catch in the Columbia River, 1866-1994

Salmon recovery strategies in the CR

6

Technological fixes and hatchery production

Continuum of marine, estuarine, and riverine habitats critical to

preserve the diversity of salmon life histories

Production view Population view

Paradigm shift

(Lichatowich, 1999; Bottom et al, 2005, 2008; Fresh et al, 2005; NPPC 1997, 1998, 2009)

CORIE/SATURN: A coastal-margin observatory for the CR estuary-plume-shelf

Goal: to deliver quantifiably reliable environmental information, at the right time and in the right form to the right users.

Can complex models that simulate the physical environment provide credible and useful answers to the decision makers ?

Opportunity and challenge: can high-resolution numerical models address the time scales relevant to investigate the impact of anthropogenic activities in the context of natural variability and climate change?

7

Observation network

ELCIRC SELFE

Modeling system Information management

Research Objectives

8

Q1a:To what extent is the CORIE/SATURN modeling system capable to reproduce known dynamics of the CR plume?

Q1b:Can multi-year simulation databases of circulation further our understanding of the seasonal and inter-annual variability of the plume in its response to river, ocean and atmospheric forcings?

Introduction

Research Objectives

I. Seasonal and interannual variability of the CR plume

II. The CR plume and salmon survival

III. Salmon habitat opportunity in the CR estuary

IV. Future work: residence times in the CR estuary

V. Conclusions

Research Objectives

9

Q2:Does the CR plume play a role in the survival of juvenile salmon migrating from the Columbia River to the ocean?

Through what mechanisms?

Do inter-annual variability and climate and ocean regimes modulate that role?

Introduction

Research Objectives





V. Conclusions

Research Objectives

10

Q3:Can we use the high-resolution modeling capabilities of CORIE/SATURN to investigate the impact of natural variability and anthropogenic change on physical habitat opportunity for salmon in the CR estuary?

Introduction

Research Objectives





V. Conclusions

Research Objectives

11

Q4:How does variability in river, ocean and atmospheric forcings modify migration paths and residence times in the CR estuary and plume, potentially affecting survival success for outmigrating juvenile salmon?

Introduction

Research Objectives





V. Conclusions

Outline

12

Introduction

Research Objectives





V. Conclusions

Part I

13

Courtesy NOAA

Introduction

Research Objectives





V. Conclusions

Known patterns of variability of the CR plume

14

Classical view

Summer Winter

Barnes et al, 1972

Two winter plume patterns in response to wind (Hickey et al, 1998)

- Thicker, northward, coastally attached

- Thin, west to northwestward

Rapid changes in plume orientation and shape resulting from wind reversals (Fiedler and Laurs, 1990; Hickey et al, 1998)

Frequent summer bi-directional plume (Garcia Berdeal et al, 2002; Hickey et al, 2005)

Interannual variability associated with variability in river discharge and wind forcing (Thomas and Weatherbee, 2006)

A numerical exploration of CR plume variability

15

Q1a: To what extent is the CORIE/SATURN modeling system capable to reproduce known dynamics of the CR plume?

Q1b: Can multi-year simulation databases of circulation further our understanding of the seasonal and inter-annual variability of the plume in its response to river, ocean and

atmospheric forcings?

Analysis of plume variability | Evaluation of model skills

Seasonal and monthly climatologiesand anomalies of surface S

Integrative plume metrics

EOF analysis

1999-2006 simulation database(SELFE)

Model-obs and inter-model comparisons

Ability to represent known dynamics

Suite of skill scores

Conditional distributions of modeled salinity

Plume variability: River forcing

16

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

Riv

er

dis

cha

rge

(m3s-

1)

1999

2000

2001

2002

2003

2004

2005

2006

Seasonal-Sustained peaks during the spring snowmelt freshet-More episodic peaks generated by winter storms-Flows decreasing through the summer into the fall

Interannual-Intensity of winter storms and timing and intensity of the freshet (though reduced by flow regulation)-Highest flows of winter and spring 1999, followed by 2000-2001 drought

Plume variability: wind forcing

17Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

1999

2000

2001

2002

2003

2004

2005

2006

Win

d (m

s-1)

Seasonal-Winter downwelling -favorable winds to the north-Summer upwelling- favorable winds to the south-Stronger wind stress during winter storms

Interannual-Intensity of winter storms and timing of spring transition-E.g. strongly enhanced downwelling of Feb 1999-Weak northward winds and reversals of Feb 2003-Upwelling winds of Feb 2005 and 2006-Late spring transition of 2000 and 2005

Plume variability:seasonal climatologies

18

Climatologies of the surface S, generated from our 8-year simulations, are consistent with the known prevailing seasonal patterns

Winter Summer

DB14

Plume variability:monthly climatologies and anomalies

19DB14

Plume variability: plume metrics

20

Multi-year simulations of the 3D salinity field

Integrating over space

Salinity cutoff = 28 psu

Area of the surface plume

Plumevolume

Plumelocation

(centroid)

model output@ 15 min intervals

Plume area

Plumeaverage depth

1999

Plume variability: volume

21

Delayed response to increases in CR discharge

Largest volumes formed following the freshet season of 1999 and 2000, with seasonally larger volumes characterizing, in most years, the stormy winter season and the spring.

30-psu plume varied, in average-flow years, within a range comparable to the 20-110 km3 estimated in Hickey et al (1998)

Plu

me

vo

lum

e (

km3)

1999

2000

2001

2002

2003

2004

2005

2006

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov DecR

ive

r d

isch

arg

e (m

3 s

-1)

DB14

Plume variability: average depth

22

06

1218

2002 Plume average depth (m)

Wind (ms-1)

The ratio of plume volume to its surface area (average depth) in the simulations captures the prompt response of the plume to wind reversals

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

DB14

Plume variability: depth and orientation

23

Time series of plume depth at the northward inner shelf location consistent with the two basic winter structures observed in (Hickey et al, 1998)

Agreement with observed shallow summertime plumes and deeper wintertime plumes

Summer plume consistently present at ogi01 in 1999 (high discharge and consistent southward winds)

Occasional appearances at ogi01 if low flows and frequent wind reversals

Bi-directionality observed by Hickey et al. (2005) for the CR summer plume may apply at times to the winter plume as well.

Episodic winter plume reversals confirmed in CORIE/SATURN observations

Plume variability: EOF analysis

24

Plume variability: EOF analysis - winter

25

Our analysis:Winter months: Nov-March of all years 1999-2006

Hickey et al, 1998:EOF analysis of 1 m salinity survey data, October 25-November 28, 1990

EOF1: 57%CR plume when separated from the coast and oriented northward of the mouth

EOF2: 18%CR plume to the north but hugging the coast.

Plume variability: EOF analysis - winter

26

Plume variability: EOF analysis - summer

27

Evaluation of model skills: methods

Duplicative realizations of circulation database:o DB14 – SELFE (upwind )o DB11 – ELCIRC (ELM)o DB13 – SELFE (ELM)

Skill scoreso RMSEo Brier skill score=

1-MSE/MSEref

o Correlation skill score, ρMO

o (Unconditional) model bias:

MB=(E(M)-E(O))/ σO

o Normalized standard deviation for the model predictions, σM/σO

Distributions of modeled salinity conditional on the value of the observed salinity

28

Evaluation of model skills: scores

• RMSE is in most cases substantially reduced in DB14, except at deeper stations (at 5 and 20 m depth at the three RISE buoys)

• MB is consistently negative for DB11, and markedly larger, in absolute value, than the bias in DB14 (except at deeper stations) excessive freshness in ELCIRC simulations

• Larger biases at depth in DB14 are due to the use of terrain-following coordinates

• Despite the clear overall superiority of SELFE in DB14, ρMO reveals instances where DB14 simulations perform worse than DB11 in reproducing variability in observed salinity

• DB11: variability in modeled salinity is generally distinctively higher than the variability in observed salinity (σM>σO);DB14: σO>σM

• Consistently higher Brier skill scores for DB14 than for DB13: improvement in adopting an upwind method in place of ELM to solve the transport equation

29

Evaluation of model skills:conditional distributions

30

OGI011999

OGI012001

RISEN-1 m2004

RISES-1 m2004

OGI011999

OGI012001

RISES-1 m2004

RISEN-1 m2004

DB14 (SELFE)

DB11 (ELCIRC)

10th and 90th

25th and 75th

50th (median)

Percentiles:

Part I: Summary of findings

31

Introduction

Research Objectives





V. Conclusions

Correctly reproduced known patterns of variability.

Interannual variability around climatological seasonal conditions in agreement with the results of T&W (2006).

Integrative metrics proved valuable in capturing the evolution of the CR plume in its response to variability in river and wind forcing.

Differential influence of the CR plume on the Washington shelf across the years with potential implications on productivity.

8-year EOF analysis confirmed the two basic winter structures observed in 1990-91 (Hickey et al, 1998), indicating generality of the result.

First two EOF modes clearly related to the two key forcing mechanisms of seasonal and inter-annual variability of the CR plume

Part I: Summary of findings

32

Introduction

Research Objectives





V. Conclusions

Prevalent bi-directionality of summer plume regardless of interannual variability.

Short-term bi-directional plumes, previously observed or modeled only in summer, can occasionally develop also in winter as a result of episodically strong upwelling-favorable winds.

Confirmed overall superiority of SELFE in the multi-year DB14 simulations (small RMSE and bias) and excessive freshness of DB11 simulations (ELCIRC) .

DB14, to an extent, achieved better performance in terms of RMSE –even when exhibiting weaker correlation with the observations– by producing results that are conservatively less variable than the corresponding observations

No one score is adequate by itself to fully evaluate the skill of a model

33

Photo courtesy E. Keeley

Q2:Does the CR plume play a role in the survival of juvenile salmon migrating from the Columbia River to the ocean?

Through what mechanisms?

Do inter-annual variability and climate and ocean regimes modulate that role?

High quality of CORIE/SATURN simulations provides a rationale for using integrative metrics of CR plume structure to investigate the ecological implications of plume dynamics

Part II

34

Photo courtesy E. Keeley

Introduction

Research Objectives





V. Conclusions

Ocean environments in Pacific salmon survival

Both freshwater and ocean environments contribute substantially to egg-to-adult salmon mortality

For the ocean phase of salmon life history, most mortality occurs within the first few weeks or months of ocean residence

Effort in the last decade into understanding the relationship between Pacific salmon production and climate variability patterns, such as ENSO and PDO

Local marine environments (ocean-shelf upwelling, river plumes) may play as large a role in the early marine survival of salmon as the regime shifts operating at broader, regional scales

35

A role for the Columbia River plume?

Does the CR plume influence salmon survival?What we know

Higher abundance of juvenile salmon in the coastal region off the CR associated with the low-salinity plume waters and frontal zones compared to the surrounding ocean waters (De Robertis et al, 2005)

Juvenile salmon do not seem to take advantage of increased zooplankton biomasses at plume fronts, possibly due to their transience or small scale (Morgan et al, 2005)

Local conditions in the environments that connect the river migration corridor to the ocean more likely determine rapid change in survival during a migration season than conditions farther away (ocean feeding areas of the gulf of Alaska or Bering Sea) (Scheuerell et al, 2009)

Survival of outmigrating juvenile salmon varies at time scales consistent with changes in the CR plume

36

How does intraseasonal variability in salmon survival relate to variability in the physical plume

environment simulated by CORIE/SATURN?

Smolt-to-adult return rates (SARs)

37

PIT taggingBargingMigration

Through the estuary

Upstream migration to spawn

2-4 Years at sea

Adult detection

SAR = # Adults

# Juveniles

The correlation analysis

38

May 1999

The CR plume:a fast-changing hydrodynamic feature

Correlation analysis between daily values of SARs and plume metrics.

Since we could only roughly estimate time of ocean entry, we explored the cross-correlations at different time lags.

Analysis performed using anomalies from the 4-year climatologies

Non-parametric method to account for autocorrelation in testing significance of cross-correlations

DB14

Steelhead

39

1999 2001 2002 2003

Favorable large-scale ocean conditionsPoor large-scale ocean conditions

1999

lag (days)

Cro

ss-c

orre

latio

n co

effic

ien

t

DB14

Monthly PDO index

1900 1920 1940 1960 1980 2000

Chinook

40

1999 2001 2002 2003

Favorable large-scale ocean conditionsPoor large-scale ocean conditions

DB14

Strengths and uncertainties

Our results were robust to the high inter-annual variability in local ocean (plume) conditions, till the regime shift in the large-scale ocean conditions occurred.

41

Alternative interpretations (e.g. local upwelling, which may affect salmon survival through bottom-up forcing of the marine food web) do not explain the differential response of the two species.

SARs are a metric that encompasses several stages in the life history of the fish and multiple years: conditions that steelhead encounter in the plume at the time of ocean entry can explain only part of their overall survival (16-40% of its variability) .

Small numbers of returning adults upon which the SARs were based made their estimate fairly imprecise, but we believe that the trends of within-season variability are correctly captured .

Part II: Summary of findings

42

Introduction

Research Objectives





V. Conclusions

Lagged cross-correlations suggested that steelhead benefited from the plume environment at a narrow window of time around their ocean entry.

Contribution of plume conditions to the overall variability in steelhead survival became modest when large-scale ocean conditions turned unfavorable.

Daily variability of the plume did not affect survival of Chinook salmon.

Differential response between the two species is consistent with observed and previously reported behavioral characteristics

H: Steelhead mainly use the plume to move quickly away from coastal predation and for a more direct migration to ocean habitats.

43

Courtesy J. Burke

Q3:Can we use the high-resolution modeling capabilities of CORIE/SATURN to investigate the impact of natural variability and anthropogenic change on physical habitat opportunity for salmon in the CR estuary?

Succeeded in using the high-quality CORIE/SATURN simulations of plume dynamics to develop a biological hypothesis

Part III

44

Courtesy J. Burke

Introduction

Research Objectives





V. Conclusions

Physical Habitat Opportunity

45

Water depth:10cm d 2m

Water velocity:v 30cm/s

Salinity:0 s 5 psu

Temperature:0 T 19 oC

Estuarine PHO metrics

Sub-yearling ocean-type

salmon

Habitat Opportunity =

Hours of PHO per week @ each grid point

availability of habitat that, based upon physical factors, physiological constraints, and ecological interactions, salmon can access and which salmon can benefit from occupying (Bottom et al, 2005).

Long-term simulation

databases of circulation , u, S, T

LongLong--term term simulation simulation

databases of databases of circulationcirculation , u, S, T, u, S, T





Physical Habitat Opportunity

46

PHO accumulated per week over a specified region (hours*m2)

Averaged PHO within the inundated area (hours/week)

Water depth:10cm d 2m

Water velocity:v 30cm/s

Salinity:0 s 5 psu

Temperature:0 T 19 oC

Estuarine PHO metrics

Sub-yearling ocean-type

salmon









River flow (m3 s-1)

Habit

at

opport

unit

y

Habitat Opportunity = availability of habitat that, based upon physical factors, physiological constraints, and ecological interactions, salmon can access and which salmon can benefit from occupying (Bottom et al, 2005).

Estuarine regions

47

Mouth Middle estuary Tidal freshwater

Peripheral bays

Baker

Youngs

Grays

Cathlamet

Interannual variability and anthropogenic change

48

Scenario 1: Predevelopment (1880) bathymetry and flow

Scenario 2: Modern dikes in predevelopment scenario

Scenario 3: Predevelopment flow over modern bathymetry

Scenario 4: Modern (2004) flow over predevelopment bathymetry

Scenario 5: Modern flow over modern bathymetry

Anthropogenic change

Interannual variability

1999-2006 simulation database

Time series of weekly PHO climatologies and anomalies

Catalogue of anomaly maps

Water depth in the modern lower estuary

49

1999-2006climatology(109 h*m2)

1999

2000

2001

2002

2003

2004

An

om

aly

(re

lati

ve to

19

99

-20

06 c

lima

tolo

gy)

2005

2006

Week 42 – neap tide

Week 43 – spring tide

Influence of tides dominates variability in shallow water (and low-velocity) habitats in the modern CR lower estuary

Differential response to neap and spring tides across lower estuary

DB14

Water depth in the tidal freshwater region

50


1999

2000

2001

2002

2003

2004

An

om

aly

(re

lati

ve to

19

99

-20

06 c

lima

tolo

gy)

2005

2006

Only more extreme flows have an appreciable, but still modest, impacton PHO in the modern bathymetry

Strong historical freshets brought considerable gain in shallow water habitats through access to the floodplain in the predevelopment bathymetry

DB14

DB17

Velocity in the tidal freshwater region

51


1999

2000

2001

2002

2003

2004

An

om

aly

(re

lati

ve to

19

99

-20

06 c

lima

tolo

gy)

2005

2006

In the modern bathymetry, the moderate gain in shallow water habitat, as Q, tends to be canceled out by PHO loss due to velocity constraints

In the predevelopment bathymetry, loss in PHO due to increasing velocities stopped for flows higher than 15,000 m3s-1 (inundated floodplain)

DB14

DB17

Influence of temperature on PHO

52

Continuous improvements in the quality of the CORIE/SATURN simulations: DB17 skill in representing temperature changes in the middle estuary has been transformative.

Simulations confirmed that, by mid-July (and through September), habitat is scarcely available for salmon to rear in the middle estuary because of excessively warm temperatures.

Mod

el b

ias

Hou

rs

week

Influence of salinity intrusion on PHOin the middle estuary

53

Salt may penetrate deeper in the modern CR system, at times limiting habitat opportunity in Cathlamet Bay and off Grays Bay also at higher flows

Modest estimated loss in PHO due to deeper salt intrusion into the modern middle estuary

Order of magnitude not dissimilar from the loss determined by extreme low flows within the natural variability of the modern system

DB17

Part III: Summary of findings

54

Introduction

Research Objectives





V. Conclusions

Strategies aimed at re-establishing some connectivity between the river and its floodplain through modification of both flow and bathymetry are necessary to restore access to shallow and low-velocity rearing habitats in the upper estuary

Modest estimated loss due to deeper salt intrusion in the modern middle estuary

How salinity intrusion is changing relative to historical conditions needs to be a focus of further investigation

Confirmed rearing habitat scarcely available in the middle estuary because of excessively warm T by mid-July through September

Spatial connectivity among pockets of habitat opportunity needs to be investigated

Future work

55

Introduction

Research Objectives





V. Conclusions

Q4:How does variability in river, ocean and atmospheric forcings modify migration paths and residence times in the CR estuary and plume, potentially affecting survival success for outmigrating juvenile salmon?

Differential response of estuarine regions

56

Preliminary results from ELCIRC simulations (DB11)

Shallow environments and well-connected channels exhibit a differential response to changes in river discharge, both seasonally and interannually

Shallow regions are areas of longer retention

DB11

RTs in the estuary-plume continuum

57

While RTs in the estuary are clearly influenced by river flow regimes, dominant processes affecting RTs in the domain extending over the plume region are wind-driven, and not necessarily linked to the presence of the plume

1999

1999

2001

2001

DB11

Contributions

58

Introduction

Research Objectives





V. Conclusions

Demonstrated the quality of CORIE/SATURN simulations in reproducing known dynamics of the CR plume

Improved our understanding of CR plume variability, in particular by showing:

o that results of a bimodal winter plume and prevalence of a bidirectional plume in summer can be generalized regardless of interannual variability

o episodic winter plume bidirectionality

Evaluated skill of the simulations providing feedback to the model developers

Contributions

59

Introduction

Research Objectives





V. Conclusions

Demonstrated how high-resolution numerical models like SELFE and ELCIRC, in the context of CMOs, can be successfully used to:

o Formulate hypotheses for the mechanisms that link performance of biological species to their physical environment

o Address the temporal scales that are relevant to investigate natural variability and anthropogenic change

o Inform salmon recovery strategies in the CR basin

- Combination of flow and bathymetry modifications are necessary to restore access to shallow and low-velocity rearing habitats in the upper estuary

Acknowledgments

My committee Joseph Zhang Charles Seaton, Paul Turner,

Ethan VanMatre, Michael Wilkin John Williams, Charles ‘Si’

Simenstand, Doug Marsh Sergey Frolov Barbara Hickey, Ed Dever, Jen

Burke, Mark Scheuerell Sandra Oster Nate Hyde, Aaron Racicot OGI staff: Amy, Nancy, Alison…

60

The PDX Aliens Peter My family Bonnie Gibbs…

Funding support for this research: NOAA Fisheries National Science Foundation

The end

61

Back-up slides

62

Plume variability: location (centroid)

63

N-S relative to the CR mouth Distance from shore

Coastal Upwelling Index Columbia River discharge

DB14

Accounting for the autocorrelation in the data

64

The shape of the ACF and PACF suggested that simple AR1 models were not adequate to describe the plume metrics time series in our study.

We could not assume that frequencies lower than the daily sampling frequency (removed by removing autocorrelation) were unimportant.

Size of SAR dataset and non-stationarity of plume series: potential shortcomings of adjusting hypothesis testing procedure

Non-parametric test

DFTPhase-

randomized IFT2000

surrogatesPlume time

series

Empirical distribution for rCRIT generated ‘resampling’ in the frequency domain

Surrogates preserve the original autocorrelation structure

Documents

The Columbia River estuary and plume: Natural variability, anthropogenic change and physical habitat for salmon PhD Candidate: Michela Burla Research Advisor: