Uncertainties in runoff projections in Southwest Western Australian and Central Chilean catchments

Pilar Andrea Barría Sandoval

School of Earth Sciences

The University of Melbourne

Victoria, Australia

August, 2017

This thesis is submitted in fulfilment of the requirements of the degree of Doctor of Philosophy

Abstract

Important runoff reductions have been reported in mid-latitude, Mediterranean-like climate catchments of the Southern Hemisphere (SH), in particular in the southwest of Western Australia (SWA) and in central Chile (CC). These changes have been driven by decreases in rainfall since the mid-1970s. Despite regional rainfall and runoff projections from Global Climate Models (GCMs) indicating that the observed trends are expected to continue during the 21st century, the projections are affected by large uncertainties that limit their utility to decision makers. The main source of uncertainty in runoff projections are the GCMs used to produce future climate projections. However, uncertainties arise from the observations of the climate variables, the statistical methodology to downscale the GCM simulations to the catchment scale and the hydrological model used to simulate runoff. In particular, the short length (<50 years) and poor spatially distributed observed climatological variables in mountainous catchments, characterized by steep topography, hampers a deep analysis of runoff trends and runoff variability, such as the case of CC mountainous catchments.

The impact of the GCM uncertainty on runoff projections has mainly been assessed through comparison of multi-model runs of future climate with little exploration of uncertainties inside the models due to different parameterisations. This thesis seeks to investigate the uncertainty response of projected runoff due to both: perturbed physics parameter variations within a GCM using a novel 2500-member ensemble from the HadCM3L model, the climaprediction.net data (CPDN), termed the within-GCM uncertainties, and from a multi-model ensemble of different GCMs collated by the CMIP5 project, termed the between-GCM uncertainties.

The impact of GCM uncertainties on runoff modelling for pluvial regimes in southwest Western Australian and Central Chilean catchments was assessed. Both regions share similar trends and climatic features, with major decreases in winter precipitation and runoff since the mid-70s that have been related to a displacement of the Southern Hemisphere storm track. Nonetheless one important difference between SWA and CC catchments, is the presence of nivo- pluvial regimes located at the foothills of the Andes in CC, whose hydrology is poorly understood mainly due to the lack of well distributed and long gauge records that represent its variability.

The results presented in this thesis show that the impact of within-GCM uncertainties on runoff projections in SWA catchments is very large; larger than previous estimates of within- GCM uncertainties impact on runoff. The perturbed physics approach indicates that current water management assessments underestimate uncertainties in runoff projections. Regarding the comparison of the impact of between-GCM and within-GCM uncertainties on runoff projections in SWA catchments quantified as the difference between the 5th and the 95th percentile of simulations, the impact of within-GCM uncertainties on runoff projections range between 39% and 65%. Whereas the impact of between-GCM uncertainties on runoff projections range between 44% and 83% for the Representative concentration pathway 4.5 (RCP4.5) scenario and about 38% and 72% under the RCP8.5 for the period between 2050-2080 compared to 1970-2000. Regarding CC catchments, between-GCM uncertainties of about 55% and 51% in runoff projections using the RCP4.5 and the RCP8.5 scenarios were found. The results here reported indicate that the impact of within-GCM and between-GCM uncertainties in SWA catchments runoff projections is very similar. The results also indicate that because some GCMs in the CMIP5 ensemble have multiple runs, using different initial conditions, CMIP5 gives some insight into within-GCM uncertainty as well. For these reasons and because CMIP5 provides runs that represent all regions of the world, it is recommended for use in hydrological assessments of climate change impact and the uncertainties around the projections.

Finally, and aiming to improve the understanding on runoff variability in mountainous catchments of the temperate region of CC, this thesis includes the first high elevation runoff reconstruction in Chile using 300 years of tree ring chronologies of Araucaria araucana and Astroceudrus chilensis. The upper part of Biobío river melting season runoff (October-March) and pluvial season runoff (April-September) were reconstructed and analysed to investigate the influence of large scale climatic drivers on runoff generation, current drought trends and to improve the understanding of long term hydroclimate variability in this region. Important differences in the runoff variability of the upper and the lower elevation catchments were identified, which are in part influenced by the large scale climatic features that drive runoff generation in both regions.

Declaration

This is to certify that:

I. The thesis comprises only my original work towards the PhD except where indicated in the Preface,

II. Due acknowledgement has been made in the text to all other material used,

III. The thesis is fewer than 100 000 words in length, exclusive of tables, maps, bibliographies and appendices.

Pilar Andrea Barría Sandoval

Preface

This thesis comprises material that has been published in a peer-reviewed journal:

- Chapter 4 is based on Barria et al. (2015) paper, of which the content is original and the co-authors contributed 20 percent of the work: Barria, P., Walsh, K.J., Peel, M.C. and Karoly, D., 2015. Uncertainties in runoff projections in southwestern Australian catchments using a global climate model with perturbed physics. Journal of Hydrology, 529, pp.184-199. DOI: 10.1016/j.jhydrol.2015.07.040.

- Chapter 5 in this thesis is based on a manuscript currently accepted to be published in Journal of Southern Hemisphere Earth Systems Science, Barria et al (2017), of which the content is original and the co-authors contributed 15 percent of the work: Barria, P., Peel, M.C., Walsh, K.J., and Garreaud, R., 2017. Analysis of within and between-GCM uncertainties of runoff projections in Mediterranean-like catchments. Journal of Southern Hemisphere Earth Systems Science.

- Chapter 6 in this thesis is based on a paper currently published in the International Journal of Climatology, Barria et al. (2018), of which the content is original and the co- authors contributed 10 percent of the work: Barria, P., Peel, M.C., Walsh, K.J. and Muñoz, A., 2018. The first 300‐year streamflow reconstruction of a high‐elevation river in Chile using tree rings. International Journal of Climatology, 38, pp.436-451. DOI: 10.1002/joc.5186.

The research was funded, in part, by the Australian Research Council (ARC) Centre of Excellence for Climate System Science (grant CE110001028). Additional funds were provided by the scholarship CONICYT Becas Chile, CONICYT PAI/INDUSTRIA 79090016. Murray Peel is the recipient of an Australian Research Council Future Fellowship (FT120100130).

To my best friend, my confidante,

the person who has given me incessant and unconditional love and support since I was 2 years old. To my wonderful sister, Noe. Acknowledgments

Firstly I would like to thank my supervisors Dr. Kevin Walsh and Dr. Murray Peel for their support, understanding and wise advice during all the years of my PhD, especially for their patience during all the difficult moments I had throughout my PhD years.

Especial thanks to Conicyt that provided the Becas Chile scholarship 79090016 which allowed me to conduct my studies during more than 4 years.

My sincere thanks to all those people and institutions that provide data to pursue the aims of my research. In particular, I would like to thanks Katherine Sadler for providing me the AWAP data, Francois Delage and Ben Henley for helping me with the CMIP5 data, David Karoly for helping me to get the climateprediction.net data from the Met Office in the UK, Ariel Muñoz for providing the tree ring chronologies data and Ricardo Gonzalez who helped me with the observed runoff data of high elevation catchments in Central Chile.

Thanks to the co-authors of the three papers I wrote during my PhD, all of them published in different journals, thanks David Karoly, Ariel Muñoz and René Garreaud for your time, revisions, comments and discussion.

I would also like to thank my beloved friends, who walked with me during all the PhD years, thanks Katherine Lizama, Estephany Marillo, Raul Lugo, Toni Cox, Christopher Chambers, Andreas Nedegard, Daniel Pazmiño, Joshua Soderholm, Annie, Jan Tympel, Andrea Dittus, Javiera Jofré, May-Lin Tay, Mauricio Quezada, Fernando Chong, Francisco Sabat and Sandra Perez. Everything would be much more difficult without your friendship.

Finalmente quiero agradecer a mi familia, muchas gracias a mis queridos padres, amados hermanos Noe, Nacho y Rodrigo y a Emilia. Gracias por apoyarme en todos los proyectos que he emprendido en la vida, por quererme tal cual soy y por ayudarme en todos los momentos, buenos y malos.

Contents

Chapter 1. Introduction ...... 1

1.1 Hydroclimatic trends in Mediterranean-like climate catchments ...... 1

1.2 Objectives ...... 4

1.3 Thesis structure ...... 5

Chapter 2. Literature Review ...... 7

2.1 Observed trends and projections of hydroclimatic variables in SWA catchments …………………………………………………………………………………8

2.2 Observed trends and projections of hydroclimatic variables in Central Chile 12

2.3 Uncertainties in runoff projections ...... 17

2.4 Summary ...... 20

Chapter 3. Hydrometeorological DATA and catchment characteristics ...... 23

3.1 Observed meteorological and runoff data ...... 23

3.1.1 Southwest of Western Australia ...... 23

3.1.2 Central Chile ...... 29

3.2 GCM data ...... 38

3.3 Tree ring data ...... 44

3.4 Climatic features reconstruction time series ...... 47

Chapter 4. Uncertainties in runoff projections in southwestern Australian Catchments using a global climate model with perturbed physics ...... 49

4.1 Introduction ...... 49

4.2 Methodology ...... 50

4.2.1 Evaluation of CPDN output...... 50

4.2.2 Bias Correction Methodology ...... 53

4.2.3 Precipitation Evaporation Runoff Model Description ...... 54

4.3 Results ...... 57

I

4.3.1 Evaluation of CPDN data ...... 57

4.3.2 Runoff Modelling ...... 62

4.3.3 Comparison of within GCM uncertainties from stochastic generation of data …………………………………………………………………………….70

4.3.4 Comparison of within-GCM uncertainties from GCM perturbed physics . 71

4.4 Summary ...... 73

Chapter 5. Exploring uncertainties on runoff projections in Mediterranean like catchments ……………………………………………………………………………76

5.1 Introduction ...... 76

5.2 Methodology ...... 78

5.3 Results ...... 79

5.3.1 CMIP5 and CPDN evaluation in SWA and CC catchments ...... 79

5.3.2 PERM model calibration and evaluation in SWA and CC catchments ...... 82

5.3.3 Comparison of within and between-GCM uncertainties in runoff projections in SWA catchments ...... 84

5.3.4 Between-GCM uncertainty of runoff projections in Central Chilean Catchments …………………………………………………………………………….88

5.3.5 Analysis of interplay between ozone recovery and GHG in CC and SWA catchments ……………………………………………………………………………92

5.3.6 Discussion ...... 95

5.4 Summary ...... 96

Chapter 6. 300 years of Streamflow reconstruction of a high elevation catchment in central Chile using tree rings ...... 99

6.1 Introduction ...... 99

6.2 Methodology ...... 100

6.3 Results ...... 103

6.3.1 Reconstruction ...... 104

6.3.2 Links between runoff reconstructions and large scale climatic features .. 107

6.3.3 Spectral analysis of reconstructions ...... 111

II

6.3.4 Drought analysis of the reconstructions ...... 114

6.3.5 Comparison with the lower Biobío catchment reconstruction ...... 115

6.4 Summary ...... 117

Chapter 7. Conclusions and future work ...... 120

7.1 Within-GCM uncertainties in runoff projections in SWA catchments...... 121

7.2 Comparison of the between-GCM and within-GCM uncertainties on projected runoff in southwest Western Australia...... 123

7.3 Analyses of the between-GCM uncertainties on runoff projections in CC catchments. ………………………………………………………………………………124

7.4 Study of the runoff variability at lower frequencies in a high elevation catchment of CC using a multi-century reconstruction of runoff...... 125

7.5 Future Work ...... 127

Chapter 8. References ...... 130

III

List of Figures

FIGURE 1.1 GIORGI REGIONS, BASED ON GIORGI ET AL. (2001) AND EXTRACTED FROM ROWLANDS ET AL. (2012) ...... 3

FIGURE 1.2 THESIS STRUCTURE ...... 6

FIGURE 3.1 LOCATION OF SWA CATCHMENTS ...... 24

FIGURE 3.2 CLIMOGRAMS OF SWA CATCHMENTS ...... 26

FIGURE 3.3 SEASONAL VARIATION OF RUNOFF IN SWA CATCHMENTS ...... 26

FIGURE 3.4 ANNUAL GAUGED RUNOFF AND PRECIPITATION DATA. THE DOTTED LINE IS THE FITTED LINEAR CURVE TO THE

ANNUAL RUNOFF IN A) DONNELLY AT STRICKLAND, B) HELENA AT NGANGAGURINGURING AND C) DENMARK AT

KOMPUP ...... 28

FIGURE 3.5 LAND COVER IN SWA CATCHMENTS. EXTRACTED FROM FIG. 2, SILBERSTEIN ET AL. (2012) ...... 29

FIGURE 3.6 LOCATION OF CENTRAL CHILEAN CATCHMENTS ...... 32

FIGURE 3.7 CLIMOGRAMS FOR LOW ELEVATION CENTRAL CHILEAN CATCHMENTS ...... 33

FIGURE 3.8 SEASONAL VARIATION OF CENTRAL CHILEAN CATCHMENTS ...... 33

FIGURE 3.9 (A) SEASONAL VARIATION OF THE UPPER PART OF BIOBÍO RIVER RUNOFF (MM) OVER THE COMPLETE

HYDROLOGICAL YEAR (APRIL-MARCH). THE BARPLOT PRESENTS THE CLIMATOGRAM OF THE AREA WEIGHTED

PRECIPITATION (MM) GAUGED IN ABANICO AND POLCURA STATIONS (1966-2000) OVER THE HYDROLOGICAL YEAR.

2.(B) COMPARISON OF THE SEASONAL VARIATION OF OBSERVED RUNOFF (MM) IN THE UPPER AND THE LOWER PART OF

THE BIOBÍO RIVER (1960-2002) ...... 34

FIGURE 3.10 ANNUAL GAUGED RUNOFF AND PRECIPITATION DATA.THE DOTTED LINE IS THE FITTED LINEAR CURVE TO THE

ANNUAL RUNOFF AT A) AT EL ARRAYÁN, B) CATO AT PUENTE CATO AND C) LUMACO AT LUMACO ...... 36

FIGURE 3.11 OBSERVED HIGH ELEVATION BIOBÍO RIVER SEASONAL RUNOFF AND TRENDS (1960-2015). THE FITTED LINEAR

TRENDS INDICATE REDUCTIONS OF 0.65 M3S-1 FOR ANNUAL RUNOFF, AN INCREASE OF ABOUT 0.07 M3S-1 FOR

PLUVIAL SEASON RUNOFF AND A REDUCTION OF ABOUT 1.4 M3S-1 FOR MELTING SEASON RUNOFF...... 37

FIGURE 3.12 LAND COVER IN THE CC REGION. DATA EXTRACTED FROM ZHAO ET AL. (2016) ...... 38

FIGURE 3.13 MAP SHOWING THE LOCATION OF THE UPPER PART OF THE BIOBÍO RIVER, THE RAINFALL AND TEMPERATURE

STATIONS (LIGHT BLUE CIRCLES), THE RUNOFF STATIONS (DARK BLUE CIRCLES), THE PRECIPITATION ISOHYETS (MM) AND

THE 12 TREE RING CHRONOLOGIES USED AS PREDICTORS ...... 45

FIGURE 4.1 BIAS CORRECTION METHODOLOGY FOR PRECIPITATION IN DONNELLY RIVER AT STRICKLAND DURING DECEMBER 54

FIGURE 4.2 PERM MODEL SCHEME ...... 56

FIGURE 4.3 COMPARISON OF RAW CPDN ANNUAL PRECIPITATION AND SCALED AWAP ANNUAL PRECIPITATION FOR SWA.

GREY LINES REPRESENT THE 2500 SIMULATIONS OF PRECIPITATION FROM CPDN. 95TH, 5TH PERCENTILES AND MEDIAN

OF THE SIMULATIONS ARE PRESENTED AS BLUE LINES AND AWAP ANNUAL PRECIPITATION IS PLOTTED WITH A RED LINE ...... 59

FIGURE 4.4 HISTOGRAM OF MEDIAN OF SIMULATED (CPDN) ANNUAL PRECIPITATION (MM) FOR THE PERIOD BETWEEN 1940

AND 2000. COMPARISON OF MEDIAN OF MODELLED BASED ON OBSERVED DATA (AWAP), AS INDICATED WITH THE

OPEN STAR, AND THE MEDIAN OF ALL THE SIMULATIONS DURING THE SAME PERIOD, INDICATED WITH AN ASTERISK. ... 59

IV

FIGURE 4.5 COMPARISON OF RAW CPDN ANNUAL TEMPERATURE AND SCALED AWAP ANNUAL TEMPERATURE FOR SWA.

GREY LINES REPRESENT THE 2500 SIMULATIONS OF TEMPERATURE FROM CPDN. 95TH, 5TH PERCENTILES AND MEDIAN

OF THE SIMULATIONS ARE PRESENTED AS BLUE LINES AND AWAP ANNUAL TEMPERATURE IS PLOTTED WITH A RED LINE ...... 60

FIGURE 4.6 HISTOGRAM OF THE MEDIAN OF SIMULATED (CPDN) ANNUAL TEMPERATURE (°C) FOR THE PERIOD BETWEEN

1940 AND 2000. COMPARISON OF MEDIAN OF MODELLED BASED ON OBSERVED DATA (AWAP) AS INDICATED WITH

THE OPEN STAR AND THE MEDIAN OF ALL THE SIMULATIONS DURING THE SAME PERIOD, INDICATED BY AN ASTERISK ... 60

FIGURE 4.7 5TH AND 95TH PERCENTILES OF SEASONAL EMPIRICAL CUMULATIVE DISTRIBUTION FUNCTIONS FOR PRECIPITATION

SIMULATED USING CPDN. BLUE CURVES REPRESENT THE OBSERVED PERIOD AND RED CURVES THE PROJECTION FOR THE

PERIOD BETWEEN 2020-2080...... 63

FIGURE 4.8 5TH AND 95TH PERCENTILES OF SEASONAL EMPIRICAL CUMULATIVE DISTRIBUTION FUNCTIONS FOR TEMPERATURE

SIMULATED USING CPDN. BLUE CURVES REPRESENT THE OBSERVED PERIOD AND RED CURVES THE PROJECTION FOR THE

PERIOD BETWEEN 2020-2080 ...... 64

FIGURE 4.9 BIAS CORRECTED ANNUAL PRECIPITATION OVER DONNELLY RIVER AT STRICKLAND. GREY LINES REPRESENT THE

2500 SIMULATIONS OF BIAS CORRECTED PRECIPITATION FROM CPDN PROJECT. 95TH, 5TH PERCENTILES AND MEDIAN

OF THE SIMULATIONS ARE PRESENTED IN BLUE LINES AND AWAP ANNUAL PRECIPITATION IS PLOTTED IN RED LINE ..... 66

FIGURE 4.10 BIAS CORRECTED ANNUAL TEMPERATURE OVER DONNELLY RIVER AT STRICKLAND. GREY LINES REPRESENT THE

2500 SIMULATIONS OF BIAS CORRECTED TEMPERATURE FROM CPDN PROJECT. 95TH, 5TH PERCENTILES AND MEDIAN

OF THE SIMULATIONS ARE PRESENTED IN BLUE LINES AND AWAP ANNUAL TEMPERATURE IS PLOTTED IN RED LINE ...... 66

FIGURE 4.11 SIMULATED ANNUAL RUNOFF OVER DONNELLY RIVER AT STRICKLAND. GREY LINES REPRESENT THE 2500

SIMULATIONS OF RUNOFF USING PERM MODEL RUN WITH BIAS CORRECTED PRECIPITATION AND TEMPERATURE FROM

CPDN. 95TH, 5TH PERCENTILES AND MEDIAN OF THE SIMULATIONS ARE PRESENTED IN BLUE LINES AND OBSERVED

RUNOFF IS PLOTTED IN RED LINE ...... 68

FIGURE 4.12 HISTOGRAMS OF SEASONAL CHANGES IN PRECIPITATION AND RUNOFF OVER DONNELLY RIVER AT STRICKLAND 68

FIGURE 4.13 HISTOGRAMS OF ANNUAL CHANGES IN PRECIPITATION AND RUNOFF IN ALL OF THE CATCHMENTS ...... 69

FIGURE 4.14 BOXPLOT OF UNCERTAINTIES IN PRECIPITATION AND RUNOFF USING CPDN DATA AND STOCHASTIC GENERATION,

IN THE DONNELLY RIVER AT STRICKLAND FOR THE PERIOD 2035–2064 ...... 71

FIGURE 4.15 HISTOGRAMS OF ANNUAL CHANGES IN RUNOFF CONSIDERING ALL THE SIMULATIONS OF CPDN AND THE GROUPS

OF SIMULATIONS WITH DIFFERENT PERTURBATIONS OF THE PARAMETER RHCRIT. RED LINE REPRESENTS THE MEDIAN OF

THE WHOLE ENSEMBLE AND DOTTED RED LINE THE MEDIAN OF THE SIMULATIONS FOR A PARTICULAR PERTURBATION OF

RHCRIT...... 73

FIGURE 5.1 METHODOLOGY OF RUNOFF PROJECTIONS ...... 78

FIGURE 5.2 A) COMPARISON OF MEAN ANNUAL PRECIPITATION SIMULATED BY CMIP5 IN THE HISTORICAL PERIOD AND

OBSERVED MEAN ANNUAL PRECIPITATION. B) COMPARISON OF STANDARD DEVIATION OF ANNUAL PRECIPITATION

SIMULATED BY CMIP5 IN THE HISTORICAL PERIOD AND STANDARD DEVIATION OF OBSERVED ANNUAL PRECIPITATION. 80

FIGURE 5.3 A) COMPARISON OF MEAN ANNUAL TEMPERATURE SIMULATED BY CMIP5 IN THE HISTORICAL PERIOD AND

OBSERVED MEAN ANNUAL TEMPERATURE. B) COMPARISON OF STA