KATHOLIEKE UNIVERSITEIT LEUVEN FACULTEIT INGENIEURSWETENSCHAPPEN DEPARTEMENT BURGERLIJKE BOUWKUNDE LABORATORIUM VOOR HYDRAULICA Kasteelpark Arenberg 40, B-3001 Heverlee

Rainfall variability and rainfall-runoff dynamics in the Paute River Basin – Southern Ecuadorian Andes

Promoters: Prof. Dr. ir. Jean Berlamont Prof. Dr. ir. Jan Feyen

Proefschrift voorgedragen tot het behalen van het doctoraat in de ingenieurswetenschappen door Rolando CELLERI ALVEAR

June 2007

KATHOLIEKE UNIVERSITEIT LEUVEN FACULTEIT INGENIEURSWETENSCHAPPEN DEPARTEMENT BURGERLIJKE BOUWKUNDE LABORATORIUM VOOR HYDRAULICA Kasteelpark Arenberg 40, B-3001 Heverlee

Rainfall variability and rainfall-runoff dynamics in the Paute River Basin – Southern Ecuadorian Andes

Jury: Prof. Ann Haegemans, voorzitter Prof. Jean Berlamont, promotor Prof. Jan Feyen, co-promotor Prof. Patrick Willems Prof. Jaak Monbaliu Prof. Guido Wyseure Prof. Niko Verhoest (Universiteit Gent, Belgium) Prof. Felipe Cisneros (Universidad de Cuenca, Ecuador)

UDC 556.161:556.53:91(866) June 2007

Proefschrift voorgedragen tot het behalen van het doctoraat in de ingenieurswetenschappen door Rolando CELLERI ALVEAR

©Katholieke Universiteit Leuven – Faculteit Ingenieurswetenschappen Kasteelpark Arenberg 40, B-3001 Heverlee, Belgium

Alle rechten voorbehouden. Niets uit deze uitgave mag worden vermenigvuldigd en/of openbaar gemaakt worden door middel van druk, fotocopie, microfilm, elektronisch of op welke andere wijze ook zonder voorafgaande schriftelijke toestemming van de uitgever.

All rights reserved. No part of the publication may be reproduced in any form by print, photoprint, microfilm or any other means without written permission from the publisher.

Wettelijk Depot D/2007/7515/81 ISBN 978-90-5682-847-9

All streams flow to the sea, yet the sea is never full. To the place where streams come from, there they return again. Ecclesiastes 1:7

Challenges facing mountains and mountain communities are as big as mountains themselves. Florence Chenoweth

Acknowledgements

People who have had the opportunity to travel through the Ecuadorian Sierra (Andean region) can surely have perceived the variability of climates and the diversity of landscapes one can find in short distances. And given that agricultural production is the main economic activity of this region, and that the largest Ecuadorian hydropower plants are located in the Andes, it is no wonder that weather, rainfall, and river discharge are very important topics of every day life. However, little research has been carried out to understand the processes behind our observations for providing sound knowledge to water managers, decision makers and farmers. With this issue in mind, and after working for some years in water engineering projects at Universidad de Cuenca, I embraced the challenge of researching the hydrology of our mountains in the framework of a doctoral project. Moving to Belgium with my lovely wife in her 6th month of pregnancy was the first big challenge. Thus, it is not surprise that she is the first person that comes to my mind at the moment of expressing my recognition to people who contributed to this venture. Her understanding and sacrifice throughout these years in Leuven are things I deeply appreciate. Mi amor, thank you for your constant support, encouragement and love. Before starting my endeavour I knew that the keys to success were having both a good project proposal and an excellent supervisory team. I was lucky to have them both!. Firstly, I had the aid of an extraordinary Professor. Prof. Jan Feyen who was my master’s thesis supervisor shaped my ideas and ambitions and

i

provided me a solid research framework that paved the road ahead. Later, during the execution of the research he was a source of ideas, guidance and support. Therefore I would like to express my gratitude to Prof. Feyen. It has been a great privilege, as well as pleasure, to have had the opportunity to work under his wise direction. The first task he had was of convincing Prof. Jean Berlamont that a young researcher from South America had the potential to carry out a doctoral program. He succeeded!, and Prof. Berlamont became my promoter and opened the doors of the Hydraulics Laboratory where I had a comfortable office space and all the required facilities for conducting my research. Since then his advice and sharp and thought-provoking questions were a foundation for my research. For all of this I would like to express my sincere appreciation to him. I would like to specially acknowledge Prof. Patrick Willems. He was my day-today advisor and certainly he shaped my research project. He was a permanent source of ideas and suggestions, and always managed to find free time for our discussions. His unreserved support pushed the project forward. If I have been able to finish the doctoral programme in due time is thanks to his help. I also received support from Prof. Jack Monbaliu and Prof. Guido Wyseure. They were always present in my doctoral seminars and were ready to offer me their advice. In addition, I have learned many more things beyond engineering from them. The Katholieke Universiteit Leuven Research Council Fund provided the financial support for my studies in the framework of the Interuniversity Agreement between the K.U.Leuven and universities in Latin America. To this organisation and to Universidad de Cuenca, my home university I am especially thankful. I am also very grateful to the people from the K.U.Leuven International Office who administered my scholarship, especially to Edmund Guzman for his friendship and efficient assistance.

ii

I would like to thank the members of my examination board for their willingness to evaluate this thesis. Their comments and suggestions were highly valuable to improve its clarity. I would like to thank the Programa para el Manejo del Agua y del Suelo (PROMAS) from Universidad de Cuenca for providing the data I used in my research. I want to mention and thank my friend Paul Torres from PROMAS as he was ready to help me every time I needed and made me feel a bit closer to Ecuador. In the same way, I want to thank Anita Vermunicht and Imelda Uyttebroeck from the Hydraulics Laboratory who helped me with the logistics for international travels and lately with the many details for my doctoral defence. I would like to express my deepest gratitude to Felipe Cisneros, Bert De Bièvre and Wouter Buytaert. The encouragement to start the Ph.D. came from Felipe and Bert. I worked with them since 1996 when I was given the opportunity to join PROMAS in a junior position. Their advice, support and confidence shaped my career. In 2001 I met Wouter when he started researching the hydrology of the páramo ecosystem and soon I was actively working in this field as well. Since then he became a friend with whom I have learned from field hydrology to preparing the best canelazos. His contribution to my Ph.D. is highly appreciated. In addition, I would like to thank all my friends in Leuven and my fellow doctoral students at the Hydraulics Laboratory who made of this an enjoyable and unforgettable stay in Belgium. Last but not least, I would like to thank my wonderful family: Rolo Grande, Ruth and Carlitos, for their support and spiritual assistance. Your letters, emails and those unexpected phone calls kept my spirits up! I am very proud to be part of you.

iii

iv

Summary

At least half of the global fresh water resources derive from mountains; they play a critical role in the global water cycle and are a main source for drinking water supply, hydropower and irrigation. However, despite its importance, mountain hydrology is still poorly understood and therefore the sustainable development of water resources for mountain communities cannot be properly addressed. Therefore, the purpose of this research was to contribute to a better understanding of rainfall patterns, variability of rainfall in space and time, and hydrological processes in a typical mountain basin of the tropical Andes. The Paute River Basin (5070 km2) located in Southern Ecuador was selected as case study given its relatively dense network of rainfall and discharge stations. The precipitation analysis showed large rainfall variability in the relatively small study area. Rainfall delivery was highly different between the interAndean valleys and the headwaters (i.e. the páramo ecosystem) with the former showing substantially longer and more frequent dry spells. On the other hand, rainfall in the páramo is almost uniformly spread throughout the year, which in addition to the remarkable water retention capacity of its soils gives the páramo its remarkable characteristic of authentic water reservoir for the Andean communities. This study identified four main rainfall regimes and their subsequent rainfall sub-regions which constitute an important element for identifying the sources of runoff production and therefore understanding the basin’s hydrology.

v

For the modelling of the rainfall-runoff process the VHM approach (a databased approach for the identification of the structure and parameters of a lumped hydrological model) was tested in six basins ranging in size from 145 to 1260 km2 and with contrasting physical characteristics. Results showed that simpler models can outperform complex ones. The identified models captured the hydrological behaviour of the test basins and depending on the quality of the input data series (especially rainfall) good simulations were obtained. These results are highly encouraging considering the scale of application, the limitation of rainfall data and the variability of physical properties of the basins. In addition it was shown that inaccurate rainfall estimations lead to highly uncertain model parameters, and that rain gauges in the highlands are especially important for good hydrological modelling. Furthermore, by applying the VHM approach at different time scales it was possible to illustrate why the approach leads to robust model parameters. Moreover, analysing this information allowed learning about basin-scale processes such as the evolution in soil moisture state. Finally, it was demonstrated that transferring model parameters between neighbouring basins, although still a complex task, can lead to acceptable modelled discharges in ungauged basins for which good rainfall estimates are available. These simulations were superior than the normal practice of scaling down discharges from the basin to the sub-basin according to area ratios.

vi

Samenvatting

Mondiaal is minstens de helft van de beschikbare oppervlaktewaterbronnen afkomstig van berggebieden. Zij spelen dus een essentiële rol in de globale watercyclus en zijn de voornaamste bron van drinkwatervoorziening, electriciteitsproductie via waterkracht en irrigatie. Ondanks dit belang, is de hydrologie van berggebieden slechts in beperkte mate gekend; ontwikkeling van duurzaam waterbeheer voor de gemeenschappen in deze gebieden verloopt bijgevolg zeer moeilijk. Het doel van voorliggend onderzoek bestond erin om een beter inzicht te bekomen in de neerslagpatronen, in de neerslagvariatie in zowel ruimte als tijd en in de hydrologische processen van een typisch bekken in het tropische Andesgebergte. Het Paute-rivierbekken (5070 km2) is gelokaliseerd in Zuid-Ecuador en werd geselecteerd als gevalstudie omwille van de relatief hoge dichtheid aan neerslag- en debietmeetstations. De analyse toonde grote neerslagvariabiliteit over het relatief beperkte studiegebied. Neerslaghoeveelheden verschillen sterk tussen de valleigebieden in het Andesgebergte en de brongebieden (het zogenaamde “páramo”ecosysteem), waarbij de valleigebieden beduidend langere en meer frequente droge perioden kennen. Anderzijds is de neerslag in het “páramo”-gebied nagenoeg gelijkmatig verdeeld over het jaar. Dit geeft het gebied een unieke waterbergingscapaciteit; het vervult de rol van autentiek waterreservoir voor de Andesgemeenschappen. De voorliggende studie heeft vier typen neerslagregimes en bijhorende deelgebieden geïdentificeerd. Zij hebben een

vii

beter inzicht gegeven in de bronnen van runoffproductie en de bekkenhydrologie van het gebied in het algemeen. Voor de modellering van het neerslagafstromingsproces werd de VHMmodelleringsmethode toegepast en getest. Het is een data-gebaseerde methode waarbij stapsgewijs bijkomende en complementaire informatie wordt afgeleid uit de gegevensset van neerslag- en potentiële evapotranspiratieinvoer en rivierdebietmetingen afwaarts van deelbekkens, en gebruikt voor identificatie van de meest optimale spaarse conceptuele deelgebiedsgemiddelde modelstructuur. De methode werd getest voor zes deelbekkens met oppervlakten variërend van 145 tot 1260 km2 en met contrasterende fysische gebiedseigenschappen. Resultaten hebben aangetoond dat de eenvoudigere modelstructuren nauwkeurigere resultaten opleveren in vergelijking met de meer complexe structuren. De geïdentificeerde modellen geven een nauwkeurige beschrijving van de macroscopisch hydrologisch invoeruitvoerverbanden. De nauwkeurigheid is vooral afhankelijk van de neerslaginvoer die de belangrijkste bron van onzekerheid bleek in de modellering. Gegeven de ruimtelijke schaal van het bekken, de beperkte tijdsen ruimteresolutie van de neerslaggegevens, en de sterke variabiliteit in de fysische gebiedseigenschappen, modelresultaten zijn zeer beloftevol. De studie heeft verder aangetoond dat onnauwkeurige neerslaginvoergegevens tot sterk onzekere parametercalibraties kunnen leiden, en dat overeenkomstig de installatie van bijkomende pluviografen in het hooggebergte sterk aanbevolen moeten worden. Door middel van modelstructuuridentificaties en -calibraties bij verschillende tijdschalen heeft de studie aangetoond waarom de VHM-methode tot robuuste modelparameters leidt. Analyse van deze resultaten heeft verder inzicht gegeven in de hydrologische processen op rivierbekkenschaal zoals de tijdsvariatie in bodemvochtgehaltee. Ten slotte is aangetoond dat transfer van modelparameters tussen naburige parameters toelaat om voldoende betrouwbare modelresultaten te bekomen voor deelbekkens met voldoende neerslaggegevens. Deze benadering bleek superieur in vergelijking met de gangbare aanpak waarbij debieten vanuit rivierbekkens wordt neergeschaald naar de kleinere deelbekkens op basis van oppervlakteverhoudingen.

viii

Table of Contents

Acknowledgements

i

Abstract

v

Samenvatting

vii

Table of Contents

ix

List of Symbols

xiii

1 Introduction 1.1. Importance of mountain hydrology 1.2. Objectives of the doctoral research 1.3. Overview of the study area: The Paute River Basin 1.4. Overview of data availability 1.5. Outline of the doctoral thesis

1 1 3 6 9 10

2 Description of the hydrology of meso-scale basins 2.1. Introduction 2.2. Analysis of rainfall series 2.2.2. Completion of missing gaps 2.2.3. Interpolation of precipitation 2.3. Hydrology of meso-scale basins

13 13 14 16 18 21

ix

2.3.1. Methods 2.3.2. Results 2.4. Conclusions

22 22 25

3 Analysis of space-time rainfall variability 3.1. Introduction 3.2. Data and study area 3.3. Methods 3.3.1. Relation between elevation and rainfall 3.3.2. Rainfall regimes 3.3.3. Seasonality 3.3.4. Trends 3.4. Results 3.4.1. Relation between elevation and rainfall 3.4.2. Rainfall regimes 3.4.3. Seasonality 3.4.4. Trends 3.5. Conclusions 4 Modelling the hydrology of meso-scale Andean basins using a data-mining approach 4.1. Introduction 4.2. Materials and Methods 4.2.1. The data-mining approach and lumped-conceptual model 4.2.2. Model calibration, validation and performance evaluation 4.2.3. Step-wise methodology 4.2.4. Sensitivity analysis 4.2.5. Study basins and data available 4.3. Results and discussion 4.3.1. Application of modelling approach 4.3.2. Sensitivity analysis for model parameters 4.3.3. Sensitivity analyses for the base flow separation process and the criteria to select independent events 4.4. Conclusions

x

29 29 32 33 33 33 34 35 35 35 36 40 44 48

51 51 56 56 59 61 64 66 67 67 74 78 83

5 Impact of limited hydro-meteorological input data on the VHM approach and hydrological model parameters 5.1. Introduction 5.2. Materials and Methods 5.2.1. Study basins and data available 5.2.2. Methodology 5.2.2.1. Impact of sub-daily input data 5.2.2.2. Impact of rain gauge density 5.3. Results and discussion 5.3.1. Impact of daily versus sub-daily discharge data 5.3.2. Impact of rain gauge density 5.4. Conclusions

87 87 89 89 91 91 92 93 93 98 102

6 Transference of model parameters between basins 6.1. Introduction 6.2. Materials and Methods 6.2.1. Study basins and data available 6.2.2. Methodology 6.3. Results and discussion 6.3.1. Calibration of model parameters from observations at different time-scales of aggregation 6.3.2. Transference of model parameters between similar basins 6.4. Conclusions

105 105 107 107 108 109 109 113 117

7 Conclusions 7.1. Recapitulation 7.2. Further research

121 121 124

References

125

Curriculum Vitae

141

xi

Appendix A. Map of study area Appendix B. The VHM approach Appendix C. The WETSPRO model Appendix D. Summary of modelling results

xii

List of symbols

Abbreviations BF BM DBM DEM DJF EF ENSO ET GORE IAHS IDS IDW IE IF IFM IFMIE IFMSE INAMHI INECEL ITCZ

Base flow Bi-modal Data-based mechanistic model digital elevation model December-January-February Nash-Sutcliffe coefficient of efficiency El Niño - Southern Oscillation Evapotranspiration Goodness of Rainfall Estimation International Association of Hydrological Sciences Inverse Distance Squared Inverse Distance Weight Infiltration excess Interflow Interflow module Interflow module, infiltration excess mechanism Interflow module, saturation excess mechanism Instituto Nacional de Meteorologia e Hidrologia (National Institute for Meteorology and Hydrology), Ecuador Instituto Nacional de Electrificacion (National Institute for Electricity), Ecuador Intertropical Convergence Zone

xiii

JJA MAM ORSTOM

POT PROMAS PUB QF QFM QFMIE QFMSE RI SE SI SON SSM SWAT UM VHM

WETSPRO WMO

June-July-August March-April-May Institut français de recherche scientifique pour le développement en cooperation (now L'Institut de recherche pour le développement – IRD) Peak-Over-Threshold Programa para el Manejo del Agua y del Suelo, Universidad de Cuenca, Ecuador Prediction in Ungauged Basins Quick flow Quick flow module Quick flow module, infiltration excess mechanism Quick flow module, saturation excess mechanism Replicability Index Saturation excess Seasonality Index September-October-November Soil Storage Module Soil water assessment tool Uni-modal Generalized lumped conceptual and parsimonious model structure-identification and calibration, according to the Dutch abbreviation Water Engineering Time Series PROcessing tool World Meteorological Organization

Symbols/Parameters λ aOF,1, aOF,2 aOF,3 aIF,1, aIF,2 aIF,3 au,1, au,2 au,3 BC()

Parameter of Box-Cox tranformation Soil storage module parameters Soil storage module parameters Soil storage module parameters Box-Cox transformed variable

xiv

D dα0 ea ep kBF kIF fIF(t) fOF(t) fQF(t) fSF(t) fu(t) fTF(t) Mij n p Pi PiE SI SIa SImean Qobs Qmod r Ri u uevap uinit umax WBerr wBF wIF

Soil depth (mm) Distance between locations α and 0 Actual evapotranspiration Potential evapotranspiration Recession constant for base flow Recession constant for interflow Rainfall fraction that contributes to interflow Rainfall fraction that contributes to overland flow Rainfall fraction that contributes to quick flow Rainfall fraction that contributes to slow flow Rainfall fraction that contributes to soil storage Rainfall fraction that contributes to total flow Monthly rainfall for year i, month j Number of time steps of the period Porosity (mm/mm) Best estimate of precipitation input to the basin on day i Estimate of precipitation input to the basin on day i computed with a subset of the rain gauge network Seasonality index Seasonality index calculated from average monthly values Mean seasonality index Observed discharge Modelled discharge Exponent for rainfall estimation in IDW method Annual rainfall for year i Actual soil water content Soil water content at maximum evapotranspiration Initial water content Maximum soil water content Parameter for base flow separation. Proportion of total hydrograph volume that remains after base flow separation Parameter for interflow separation. Proportion of hydrograph volume that remains after interflow separation

xv

e e x SF , xue , xQF

Estimated values of x for different submodels

m m x SF , xum , xQF

Modelled values of x for different submodels

xIF(t) xOF(t) xQF(t) xSF(t) x(t) xTF(t) xu(t) yQF(t) ySF(t) yOF(t) yIF(t) z(u0) z(uα)

Rainfall portion that contributes to interflow Rainfall portion that contributes to overland flow Rainfall portion that contributes to quick runoff flow Rainfall portion that contributes to slow runoff flow Rainfall input Rainfall portion that contributes to total runoff Rainfall portion stored as soil moisture storage Modelled quick flow discharge Modelled base flow discharge Modelled overland flow discharge Modelled interflow discharge Rainfall estimate at location u0 Rainfall observation at location uα

xvi