Hedi Rasulkth.diva-portal.org/smash/get/diva2:1201373/FULLTEXT01.pdfTRITA-LWR DLT-188 ISBN...

TRITA-LWR DLT-188

ISBN 978-91-7729-755-0

WATER IN ROADS: FLOW PATHS AND

POLLUTANT SPREAD

Hedi Rasul

June 2018

Hedi Rasul TRITA-ABE-DLT-188

ii

© Hedi Rasul 2018 PhD thesis Division of Land and Water Resources Engineering (LWR) Department of Sustainable Development, Environmental Science and Engineering (SEED) School of Architecture and Built Environment (ABE) Royal Institute of Technology (KTH) SE-100 44 STOCKHOLM, Sweden Reference to this publication should be written as: Rasul, H. (2018) “Water in roads: Flow paths and pollutant spread” TRITA-ABE-DLT-188, 52p.

Water in roads: Flow paths and pollutant spread

iii

DEDICATION

To, My mom, family and friends. Şingal, Kobanî and Efrîn. All victims of the wars in Kurdistan and around the world.


iv


v

FOREWORD

“The greatest threat to our planet is the belief, that someone else will save it” Robert Swan


vi


vii

SUMMARY IN SWEDISH

För att kunna bygga långsiktigt hållbara vägar och minimera underhåll och skador på miljö och grundvatten är det viktigt att kunna övervaka och modellera hydrologiska effekter i och på vägen samt risken för påverkan på grundvatten. Vatteninnehållet i de obundna vägskikten förändras över tid och vattenflödet uppträder ofta längs preferentiella flödesvägar vilket kan medföra att föroreningar snabbt kan spridas till grundvattnet. I dag utförs vägbyggnadsplanering vanligen utan att specifikt överväga hydrologiska kriterier. För att förbättra förståelsen av kopplingarna mellan vattenflöde i vägar och grundvatten har denna avhandling utvecklat undersökningsmetoder och använt numeriska simuleringar för att beräkna säsongsvariationer, flödesvägar och föroreningspridning. Säsongsförändringar i vägvattenhalten i en väg under drift, infiltration i vägrenen och perkolation ner till grundvattnet övervakades icke-destruktivt genom användning av elektrisk resistivitetstomografi (ERT). Kloridkoncentrationsförändringar uppskattades utifrån inverterade ERT-data. En övervakningsmetod med resistivitetsmätningar utvärderades och dataanalys utfördes på ERT-data från olika delar och väglager, vilket analyserades statistiskt och korrelerades med nederbörd, temperatur och markfuktinnehåll. Informationen har insamlats från en unik vägforskningsstation vid en motorväg (E18) nordväst om Stockholm samt i spårämnesförsök på typiska vägar i södra och centrala Sverige. Tvådimensionella (2D) modeller av värme- och fuktförändringar har utvecklats och testats för en vägsektion med hänsyn till ångtryck och vattenhaltsförändringar som beskrivs av partiella differentialekvationer (PDE). Modellparametrarna optimerades baserat på markfuktighet och temperaturdata från vägforskningsstationen. En PDE-modell användes för beräkning av halter av vatten och is i vägstrukturen vid olika scenarier baserat på väggeometri och designändringar av vägens uppbyggnad. Såväl transportvägar för infiltrerande vatten samt transporttiden har utvärderats utifrån 2D och pseudo 3D-invers modellering av ERT-mätningarna. Fältstudierna visade tydliga preferentiella infiltrations- och flödesvägar för vatten och salt som varierade tydligt mellan olika årstider. Huvuddelen av infiltrationen uppkom i vägens stödremsa och perkolationshastigheten var större på moderna vägar med grovt material i stödremsan jämfört med äldre vägar som huvudsakligen var uppbyggda av omgivande lokalt material. Simuleringarna visade att säsongsmässiga förändringar samt angivna övre gränsförhållanden var viktiga faktorer som kontrollerade vattenhalten i olika väglager. Denna kunskap ger en ökad förståelse för vatten i vägar och kan därför utgöra ett steg mot mer hållbart och miljövänligt vägbyggande och vägunderhåll. Dessutom ger forskningsresultaten ökad kunskap som kan användas praktiskt både vad gäller övervakning och vägbyggande. För övervakning har en ny metod för analys av ERT-data testats. För konstruktion och underhåll föreslås miljöskyddande åtgärder, vilket bland annat omfattar tätning av stödremsan med ett mer finkornigt material eller täckning med vegetation.


viii


ix

SUMMARY IN KURDISH (SORANI)

ئاوى پٮسبوونى و هژينگ كچونىێت ۆب كارۆھ هتێبب ىەوهئ ێبهب گاوبانڕێ ىەوهچاككردن و بنياتنان باشتر ۆب تێبكر گاڕێ كانىهكھاتێپ هچين رهسهل ئاو ىێش كانىهريهكاريگ ووردى رىێچاود هوٮستێپ واهئ, وىەز رێژ ىۆھ هتێبەد و وىەز رێژ ئاوى وەرهب تێچەد هك (pollutants) ژڕێق و پيسى ىڵهجوو رىێچاود هڵگهل

ناديار ئاساىەوێش و گاڕێ و هاندايۆڕگ هل وامىەردهب هب گاڕێ كانىهكھاتێپ هچين ناوهل ئاو ىێش. پٮسبوونى ئاوى هب گاتەد ژڕێق و پيسى وڕێڕە مانهھ هب ەو, وىەز رێژ ئاوى وەرهب دايوەز یکانهنيچ ناوهب تڕێبەد ناوهل ئاو ىڵهجوو و ىێش ىەربارەد ويستێپ يشتنىهگێت. هژينگ پٮسبوونى كارىۆھ هتێبەدەو وىەز رێژ ێب گرڕێ هك درووست كىهيەوێشهب گاوبانڕێ دروستكردنى هل تێبەد خشهسوودب گاڕێ كانىهكھاتێپ هچين ەبوار مهب ۆوتهئ گرينگى گاوبانڕێ دروستكردنى ىهسۆپر هل هسات مهئ تا. وىەز رێژ ئاوى پٮسبوونى هل گاىڕێ دانىێپەرهپ هل هبريتي هيەوهژينێتو مهئ ئامانجى, ەبوار مهئ رهس هينهبخ تيشك ىەوهئ ۆب. ەدراوهن

هانكارييۆڕگ زانينى و مآلندنهخ ۆب (numerical simulation) يىەژمار سازىەوێھاوش و شيكارى رێژ ئاوى هب پيسى يشتنىهگ تىهنيۆچ و گاڕێ كانىهكھاتێپ هچين ناوهل ئاو ىڵهجوو و ىێش كانىهرزييەو .وىەز

ۆب ەنراوێكارھهب (ERT) بايىەكار رگرىهب بوى درووست ىهنێو گاىڕێ, داهيەوهژينێتو مهل تێبكر ستكارىەد هگاكڕێ ىەوهئێب راێخ كىهگايڕێ كانىهكھاتێپ هچين هل كانهانكارييۆڕگ ريكردنىێچاود بايىەكار رگرىهب ى بو درووست ىهنێو گاىڕێ مانهھ و. ۆھاتوچ ستانىەو ۆب كارۆھ هتێبب يان ەتواو ىێخو ىەژڕێ. گاوبانڕێ واريێل هل ئاو كردنى ەدز دواداچوونىهب و شانێكهنێو ۆب ەنراوێكارھهب ووىڕ رهسهل (frost) شکڵئا/ زقم نى بو درووست هل كردن گرىۆڕێب زستاندا هل كارھاتووهب ىێخو( ى بو درووست ىهنێو گاىڕێ مانهھ نانىێكارھهب هب ەنراوڵێمهخ گاڕێ كانىهكھاتێپ هچين ناوهل) گاكانڕێ شيكارى هگڕێ و (data) كانهزانياريي ىەوهكردنۆك ۆب ەنراوێكارھهب ێنو وازىێش. بايىەكار رگرىهب

. بايىەكار رگرىهب كانىەژمار (statistical analysis) يىەئامار شيكارى ۆب ەنراوێكارھهب تيشهتايب ندىەيوهھاوپ تا ەدراو نجامهئ ۆب يىەئامار شيكارى و ەكراو نێلۆپ هگاكڕێ جياى جيا شىهب ۆب كانەژمار

(correlation) ىێش ىەژڕێ و رماهگ ىهپل, بارين باران, واهھ و شهك کانىەمارۆت هڵگهل ەوهتێزرۆبد ۆب هل گاوبانڕێ ىەوهژينێتو ۆب ناياب كىهيهستگێو هل ەوهتهكراونۆك کانەمارۆت ىهربۆز. خاك ناو هل ئاو مآلندنىهخ ۆب (2D simulation model) دوورى دوو كىهسازيەوێھاوش. مۆڵكھۆست شارى زيكێن واريێل و گاڕێ كىێشهب ۆب ەكراوەئاماد گاڕێ كانىهكھاتێپ هچين ناو ئاوى ىێش ىەژڕێ و رمىهگ كانىهپل شکڵئا/ زقم كانىهانكارييۆڕگ و ئاو مىهڵھ ىۆستهپڵهپا ۆب ركارىێژم هكهسازيەوێھاوش هل. هگاكڕێ ۆب ەكراو ێجهبێج (PDE) جياكارى ىهشێھاوكهشهب نانىێكارھهب هب هركاريانێژم مهئ. ەدراو نجامهئ كارى دياريكردنى) parameters( كانىهلكۆھاوك .نجامهئ هب يشتنهگ و كانهشێھاوك كردنىڵههاڵگ

هب ستنهپشتب هب ەدراو نجامهئ ۆب) optimization( باشترينى ىەبژاردهڵھ, هكهسازيەوێھاوش ىەژڕێ و رمىهگ كانىهپل داتاى تىهتايبهب, گاوبانڕێ ىەوهژينێتو ىهستگێو هل ەكراوۆك) یاريزان(داتاى .ىێش

ىێخو و ئاو هك خاتەريدەد وونىهڕب ەوهژينێتو ىهستگێو (fieldwork results)كانىهييەكردار هنجامهئ رێژ ئاوى گاوڕێ كانىهكھاتێپ هچين ناو وەرهب ەوهگاكڕێ واريێل هل كاتەد ەدز رىۆز كىێشهب ەتواو. سالدا جياكانى ەرزەو هل ەجياواز و تۆڕێگەد (spread pattern) هكەوهوبوناڵب ئاساىەوێش. وىەز رگرىهب مارىۆت كانىهنجامهئ ىەوهاندنهڕگهڵھ هب ەكراو رىێچاود كانەوێش و هكەوهوبوناڵب رايىێخ

هب هگاكڕێ كانىهكھاتێپ هچين ناو هل هرايێخ رۆز هكەوهوبوناڵب خاتەريدەد هك دوورى ێس بايىەكار .سروشتى خاكى هڵگهل راووردهب

هب. ەستيارهھ رۆز) upper boundary condition( ەوەرهس سنوورى رجىهم هكهسازيەوێھاوش هل و شکڵئا/ زقم كانىهانكارييۆڕگ, ئاو ىێش ىەژڕێ ۆب مآلندنهخ توانرا هكهسازيەوێھاوش نانىێكارھهب, گاڕێ ىەندازهئ ينىۆڕگ هب. تێبكر ىەوەرێژ خاكى گاوڕێ كانىهكھاتێپ هچين رمىهگ كانىهپل

هچين ناوهل رمىهگ و ئاو ىڵهجو انكارىۆڕگ كانىەكارۆھ زانينى ۆب كراەئاماد جياواز سازىەوێھاوش هل بريتين كانهي انكارىۆڕگ كانىهكيەرهس ەكارۆھ هك وتهركەد هوانهژينێتو مهب ندهپاب. گاڕێ كانىهكھاتێپ و بنياتنان باشتر وەرهب مانباتەد كێنگاوهھ هزانياريان مهئ. كانداەجياواز ەرزەو هل واهھ و شهك انكارىۆڕگ

ەرهسەچار هل ێكهي. .وىەز رێژ ئاوى و هژينگ اگرتنىڕ پاك هڵگهل گاوبانڕێ ىەوهچاككردن گاڕێ واريێل شينىۆداپ هل هبريتي گاوبانڕێ وارىێل هل ئاو كردنى ەدز ىەوهمكردنهك ۆب كانهييەكردار

.وزايىهس و ۆڵخ هب


x


xi

SUMMARY IN KURDISH (KURMANJI)

Ji bo çêtirîn avakirine û çêtirîn parastina reyan, bêyî ku zirara xwêzayî û avê jêrzemîn bê kirin, pêdiviye çavdêrkirine bandorên hîdrolografîk li ser rêyan were bizanin û ku modelek bê çekirin. Bi tevî demsalan, şilahîye ku di navê herî û rê de heye tê guhertin. Herwiha giredeyî şilahîyê herî, av û qirejîyen di nav avê, jêr ve reyan guhertî digirin. Wexta ku rê ten avakirin û ten karkirin, zanabûna ev rewşa şilahîye herî, bandorê ku bi xwezayî bê kirin kehm dike. Heta îro zanastiye rê avakine pir giringahî nedaye ev beşa. Armanca ev lekolinê, bi arîkarîye karên zevî û sîmulasyonê hijmarî, çêtirin têgihiştandine tevdanê şilahîye herî û pêş de birina modelekî ye.

Belavbûna aven demsalî û qirejî bi bikaranîna tomarografîya dijberîya elektrîk (ERT) hatiye çavdêr kirin. Her wiha guhertinen xwêya di nave ave bi reya ERT hate pîvandin. Methodên nû yen çavdêriya û nirxandinên daneyên statîstîk ji qadan û ji pelen cuda yen herî hatiye pêş de xistin. Peyvendiya navberê ERT û daneyên hawahi û şilahî bi arikariye komkirine daneyên ku li seranserê ji rawestgehe bakurê rojavayê Stockholmê tê hate danîn. Herwiha bi bikaranîna metodê tracer karên zevî yen din ji pek anîn. Modela du-dimensional (2D) yên germê û şilahî hatê amadekirin. Ji bilî vê parameterên li ser şilahîye axê û daneyên germê ji hêla rawestgeh testê ya E18-ê hatiye pêşvebirin. Ji bo hesabkirina guhertinên avê helandî û naveroka qeşa, li ser bingeha senaryoyên bi giredayî geometrî û designen cûda, modelekî PDE hatiye bikar anîn. Herdu şop û demên rêwîtiyê bi bikaranîna metodên jor hatiye hesabkirin.

Wekî encam guhartinên şopên demsalalî û belavkirina qirêjî hatê dîtin. Piranîye ketina qirêjîya nave axê li dora milê rê hatê dîtîn. Guherînên demsalî û rewşa pelen jorînen axê bûne faktorên sereke ya diyarkirinê naverokê avê ku di nav pelen cuda yen rêyan. Encamên pêvajoya çavdêrî di danûstendinê daneyên û analîzê de rêbazek nû dide. Di dema avakirin û parastina rêyan de, cekirine milê rêyekî xurt û lêzêdekirina pelekî herîya bi çînandinê sinahîyêtête pêşniyar kirin.


xii


xiii

ACKNOWLEDGEMENTS

Many thanks to the Ministry of Higher Education and Scientific Research-Kurdistan Regional Government (KRG) for a Higher Capacity Development (HCDP) scholarship programme for initiating and supporting the majority of this PhD study, in spite of war, crises and political instability in the Kurdistan region. ÅForsk foundation supported this study from July 2016 to June 2017, The Lars-Erik Lundberg and J. Gust. Richert (SWECO) foundations supported the project until April 2018. Nordiskt vägforum (NVF) supported field work involving tracer tests. Thanks to Knut och Alice Wallenbergs stiftelse-2015 and ÅForsk Foundation-2018 for travel grants. First of all, I gratefully acknowledge my main supervisor Professor Bo Olofsson, for excellent guidance, help, support, excursions, positive energy and nice music. Many thanks also to Associate Professor Joanne Robison Fernlund, who was my main contact person at KTH and my first co-advisor, which she facilitated and coordinated between KRG and KTH. I am grateful to other my co-advisors, Professor Per-Erik Jansson, Dr. David Gustafsson and Dr. Mousong Wu. It was great to have such an amazing group of diverse knowledge and broad expertise. I am very grateful to Professor Vladimir Cvetkovic for the quality check and internal review of this thesis at KTH and his valuable suggestions. Thanks to Aira Saarelainen, Britt Chow, Britt Aguggiaro, Katrin Grünfeld, Kosta Wallin and Magnus Svensson for all the help and administrative works. Jerzy Buczak was the best IT support. Thanks to Dr. Mary McAfee for all language corrections. Thanks to Dr. Annika Lundmark (Water Unit, County administration board of Jämtland) and Dr. Klas Hansson from Trafikverket for collaborations and good discussions. Most of the data were collected at the environmental road test station on European highway E18, which belongs to the Swedish Traffic Authority (Trafikverket). The weather data downloaded from SMHI and COMSOL support were a great help in preparation of the model. I am very grateful to all my multi-cultural colleagues at KTH, I see you as my global family, it was a privilege to spend this time at KTH and have you all around, I learned from you, got inspired, participated in many scientific discussions, got help in my field work and enjoyed many social and sport activities and Fika-coffee breaks. Among these wonderful friends, special mention to Liangchao, Caroline, Imran, Robert, Veljko, Ezekiel, BoLi, Alireza, Lea, Sofie, Sofiia, Zahra, Sara, Juan, Rajabu, Liwen, Minyu, Xi, Yuanying, Wen, Kedar, Martin, Mårten, Marija, Emad, Benoît, Flavio, Mauricio, Enrico, Elias, Elena, Kajsa, Ricardo, Rajib, Seema, Simona, Hadi, David and Carlos. Thanks to all colleagues and members of the Kurdish-Swedish academic association (Kurdsvenska AF), especially Baran. I am very grateful to Department of Civil Engineering, Faculty of Engineering, Koya University, for permission to study abroad and all support, especially from Dr. Dilan Roshani, Mr. Mariwan Mirhaj, Dr. Brwa Sardar, Mr. Ladeh Sardar and Ms. Galawez Bapir. Many thanks to Katerina Kucerova and Associate Professor Irene Blanco from Universidad Politécnica de Madrid, for their indirect help in ensuring that I could participate in Stockholm Environment Institute´s WEAP workshop in Madrid and eventually visit KTH and plan for this PhD study. Thanks to Dr.


xiv

Asoos Rasool and Dr. Hogir Rasul, Mr. Sherwan (Dyako Group-CEO), Kaify Gaidy, Kamal, Jamal, Saman & Srood Tailor for all logistic help. Special thanks to the HCDP team for administrative help, especially Dr. Howri, Ranj, Shkar and Maqsood. Last but not least, thanks for the love, patience and support of all my family members, especially my wife, Dr. Jingjing Yang, and my daughter Kani.


xv

LIST OF APPENDED PAPERS

This thesis is based on the following papers, which are appended at the end of the thesis:

I. Rasul H., Zou L., Olofsson B. 2018. Monitoring of moisture and salinity content in an operational road structure by electrical resistivity tomography. Near Surface Geophysics: Accepted for publication on 25 January 2018. HR carried out the field work, developed and performed the analysis and discussion, wrote most parts and formulated the conclusions. BO developed and set up the resistivity lines and protocols, reviewed the paper and contributed to the discussion and conclusions. LZ gave valuable feedback and helped with data analyses with Matlab.

II. Rasul H., Earon R., Olofsson B. 2018. Detecting seasonal flow pathways in road structures using tracer tests and ERT. Submitted to Water, Air and Soil Pollution in March 2018. HR formulated the research idea, conducted the literature review, performed the field work, carried out the data analyses and wrote most of the paper. Other co-authors helped in field work and added valuable feedback and inputs to the discussion and conclusions.

III. Olofsson B., Rasul H., Lundmark A. 2017. Spread of water-borne pollutants at traffic accidents on roads. Water Air and Soil Pollution 228:323. HR conducted the 2D simulations using COMSOL Multiphysics® software for two different types of road, participated in the analysis, wrote parts of the paper and, together with the co-authors, formulated the discussion and conclusions. The co-authors formulated the research idea and carried out the field work. BO wrote a significant part of the paper.

IV. Rasul H., Wu M., Hansson K., Olofsson B. 2018. Two-dimensional modeling of heat and moisture dynamics in Nordic roads: Model set-up and parameter sensitivity. Submitted to Cold Regions Science and Technology in April 2018. Some of the data were presented in a poster at American Geosciences Union 2017. HR developed the main research idea, performed all the analyses and simulations, produced the conclusions and wrote most of the paper. MW contributed greatly in developing the model, formulated the partial differential equations and gave valuable comments at different stages of the paper. Other co-authors reviewed the paper and gave valuable inputs and suggestions.

The following conference papers are relevant to this topic and were produced during this PhD study, but are not appended in the thesis.

Rasul H., Earon R., Olofsson B. 2014. Environmental impacts on soil and water from a new highway section based on long-term resistivity results. Nordic Hydrology Conference. NHC Stockholm: XXVIII-KTH.


xvi

Rasul H., Karlsson C., Jamali I., Earon R., Olofsson B. 2015. Geophysical methods for road construction and maintenance. European Geoscience Union, EGU. Vienna. 17(egu2015):10603.


xvii

TABLE OF CONTENT

DEDICATION‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐III

FOREWORD‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐V

SUMMARY IN SWEDISH ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐VII

SUMMARY IN KURDISH (SORANI)‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐IX

SUMMARY IN KURDISH (KURMANJI)‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐XI

ACKNOWLEDGEMENTS‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐XIII

LIST OF APPENDED PAPERS‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐XV

TABLE OF CONTENT‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐XVII

ABSTRACT‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐1

1. INTRODUCTION‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐2

1.1. Ongoing problem ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ 4

1.2. Aim and objectives ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ 4

2. BACKGROUND‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐5

2.1. Electrical resistivity tomography environmental monitoring and uncertainties ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ 5

2.2. Modelling water and heat transport in roads ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ 8

3. MATERIAL AND METHODS‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐9

3.1. Study site and field work ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ 9

3.2. Modelling approach ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ 15

4. RESULTS AND DISCUSSION‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐20

4.1. Monitoring water and salinity in roads (Paper I) ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ 20

4.2. Monitoring flow speed and pathways using tracer test (Papers II & III) ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ 25

4.3. Spread of pollutant from traffic accidents (Paper III) ‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐ 30

4.4. Modelling heat and water content changes while considering phase changes (Paper IV) ‐‐‐‐‐‐‐‐‐‐‐ 34


xviii

5. GENERAL DISCUSSION‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐41

6. CONCLUSIONS‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐43

7. RECOMMENDATIONS FOR FUTURE STUDIES‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐44

REFERENCES‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐45


1

ABSTRACT

For better road construction and maintenance while minimising damage to the environment and groundwater, it is essential to monitor and model hydrological impacts on roads and consider pollution of groundwater. Water content in unbound material in road layers changes continuously and water flow usually occurs along pathways that are the main corridors for pollutant spread to groundwater. Good awareness of hydrological conditions and of water and solute transport in road layers down to the groundwater can be helpful in minimising environmental impacts during construction and operation. Today, road planning is usually carried out without specifically considering hydrological criteria. To improve understanding of the links between water in roads and groundwater, this thesis developed investigation methods and used numerical simulations for estimating seasonal variations, flow pathways and pollutant spread. Seasonal changes in road water content in an operational road, tracer tests pathways from the road shoulder and percolation down to groundwater were monitored non-destructively using electrical resistivity tomography (ERT). Chloride concentration changes were estimated based on ERT data inversion. New monitoring methodology was assessed and data analysis was performed on ERT data from different road zones and layers, which were analysed statistically and correlated to precipitation, temperature and ground moisture content. Data were collected at a unique road test station on a motorway north-west of Stockholm and in tracer experiments on typical roads in southern and central Sweden. Two-dimensional (2D) models of heat and moisture changes were prepared for a road section, considering vapour pressure and frozen water content changes using partial differential equations (PDE). Model parameters were optimised based on soil moisture and temperature data from the E18 road test station. A PDE model was used for calculating liquid water and ice content changes in different scenarios based on geometry and design changes. Both pathways and travel times were traced by 2D and pseudo 3D inverse modelling of the ERT measurements. The field data revealed clear preferential pathways of moisture and salt in the road shoulders that varied significantly during different seasons. Most infiltration occurred directly into the road shoulder, but entered the road embankment with higher percolation speed in modern roads than in old roads consisting of natural soils. The simulations showed that seasonal climate changes and the upper boundary condition were key factors determining water content in different road layers. These findings advance understanding of water in roads and represent a step towards more sustainable and environmental friendly road construction and maintenance. In addition the research results give lessons for practice both regarding monitoring and road construction. For monitoring it provides a new method in data collection and analysis. For construction and maintenance, mitigation measures are suggested, which comprise a tight road shoulder, by e.g. adding a fine grained layer on the shoulder or covering with vegetation.

Key words: De-icing salt; Road; Tracer test; ERT; Water content; Road pollutants; Flow pathways.


2

1. INTRODUCTION

Temperature and moisture changes in unbound road layers are the main consequence of road deterioration and failure. Understanding temperature and moisture changes in roads is highly important for analyses of the road structure and for developing methods for groundwater protection and improved road sustainability. Roads and transport of hazardous liquids are both considered significant sources of pollution for surrounding soils and groundwater systems (Harrison and Wilson 1985, Thunqvist 2004, Lindström 2006, Lundmark and Olofsson 2007, Yisa 2010, Earon et al. 2012, Aljazzar and Kocher 2016, Paper III). Both field investigations and modelling are required to achieve a better understanding of water and heat transport in road layers. Therefore use of non-destructive monitoring methodology in field studies and modelling of water, heat and de-icing salt movements in road layers are essential steps in understanding the impacts of roads on the surrounding environment. Water is a very important factor in road construction. During compaction of road material, it is essential to achieve maximum dry unit weight and minimum water content (Baldwin et al. 1997, Dawson 2009). Road pavement deterioration is usually caused by temperature and moisture changes, in cold regions especially by processes such as frost heave (Hermansson 2002, Hansson et al. 2005, Jansson et al. 2006, Ekblad and Isacsson 2007, Ghazavi and Roustaie 2010, Kroener et al. 2014, Salour 2015, Sarady and Sahlin 2016). Moisture considerations in pavement design in the Nordic countries, including Sweden, are influenced by frost action, while in warmer countries the focus is on controlling excess water after heavy rain events. In both cases, climate conditions are the main factor in road pavement cost and performance (Erlingsson et al. 2009). Due to climate change, by 2100 the cost of maintenance and road failures in Sweden can be as high as 20 billion Swedish Crowns, according to the Swedish Commission on Climate and Vulnerability (SCCV, 2007), even without any environmental cost considerations. Water in roads is also the main domain for pollutant transport from road to groundwater. In the Nordic countries and other cold regions, de-icing salt is frequently used during winter to improve driving conditions on roads. This frequent use of de-icing salt can be seen as a large-scale tracer test from road to surrounding environment (Lindström 2006, Paper III). Tracing de-icing salt also increases understanding of water pathways. It is important to understand the factors affecting pollutant spread and pollutant pathways when seeking to develop methods for better environmental monitoring and protection. Some previous research has been undertaken to optimise the use of de-icing salt (Blomqvist et al. 2011, Riehm 2012). Electrical resistivity tomography (ERT) is a useful non-destructive tool for environmental monitoring without interrupting pollutant transport systems. Changes in ERT can trace environmental changes in soil


3

and groundwater (Rhoades et al. 1976, Abu-Hassanein et al. 1996, Samouëlian et al. 2005, Besson et al. 2008, Bryson and Bathe 2009). This is because ERT changes are mainly caused by changes in moisture, temperature and solute concentrations (Telford et al. 1990, Shevnin et al. 2007, Chambers et al. 2014). Two-dimensional (2D) resistivity surveying can be an effective tool for environmental applications, especially hydrogeological mapping (Dahlin 1996). Thus, ERT has been widely used in different monitoring programmes, including groundwater pollution changes at industrial sites (Mao et al. 2015, Cuong et al. 2016), pollution from roads (Leroux and Dahlin 2006, Lundmark and Olofsson 2007, Olofsson and Lundmark 2009, Minas 2010, Earon et al. 2012, Paper I), seasonal changes in flow paths (French et al. 1999a, French et al. 1999b, Aaltonen 2001, Aaltonen and Olofsson 2002, French et al. 2002, French and Binley 2004) and other soil and groundwater pollution (Daily et al. 1992, Daily and Ramirez 1995, Barker and Moore 1998, Slater and Sandberg 2000, Jackson et al. 2002, Slater and Binley 2003, Olofsson et al. 2005, Swarzenski et al. 2006, Cassiani et al. 2006, Ogilvy et al. 2007, Looms et al. 2008, Wilkinson et al. 2010, Gunn et al. 2014). Hydraulic modelling is commonly used for analyses of pollutant spread and pathways from roads. Lindström (2005) concluded that it will take decades for the groundwater to reduce the increased chloride concentration due to de-icing salt on roads. For example, more than half the total chloride concentration in the Sagån river basin has been shown to originate from the use of de-icing salt (Thunqvist 2004). Moreover, as much as 50% of the chloride content in deep-drilled private wells within 500 m of major roads in Sweden derives from de-icing salt (Olofsson and Sandström 1998). Modelling of water and heat transport in porous media is well described mathematically (Beskow 1935, Edlefsen and Andersen 1943). Numerical simulations in one and two dimensions for heat and water changes in soil are also covered by numerous studies (Harlan 1973, Guymon and Luthin 1974, Jansson and Halldin 1980, Flerchinger and Hanson 1989, Lytton et al. 1993, Simonsen et al. 1997, Šimůnek et al. 1999, Jansson and Karlberg 2001, Sheshukov and Nieber 2011, Karra et al. 2014, Zhang et al. 2016, Nickman 2016). Hansson et al. (2004) used a modified Richards’s equation in developing a one-dimensional (1D) numerical model for coupled heat and different water flow in road layers and suggested a more realistic and comprehensive field dataset to be tested with similar models. In this thesis, data from a road test station were used and the model devised by Hansson et al. (2004) was modified to two dimensions to cover a road section using COMSOL Multiphysics® software. The model was combined with numerically calculated soil moisture and temperature changes determined by Zhang et al. (2016) in partially frozen soil while considering phase changes and vapour pressure in normal soil, as the intention was to solve similar problems in road layers.


4

1.1. Ongoing problem A good awareness of hydrological conditions and good knowledge of water and solute transport in road layers down to the groundwater can be helpful in minimising the environmental impacts during road construction and operation. Today, road planning is usually carried out without specific consideration of hydrological criteria, e.g. hydrological systems are designed from a technical perspective instead of a more environmentally friendly perspective. For efficient planning and maintenance of roads, more consideration must be given to hydrological criteria, since a road can be a source of pollution to groundwater from road materials, vehicles and its parts, flow transport of hazardous goods after road accidents, and winter maintenance using de-icing salts. To understand this and create a useful numerical model for future road developments, knowledge from the literature and data from a road test station in Sweden (Test site E18) and from intensive field investigations elsewhere in Sweden were used in this thesis to answer some of many unsolved problems and to determine the links between roads and groundwater.

1.2. Aim and objectives The main aim of this thesis was to advance understanding of water and pollutant movement through road layers to the groundwater, by developing investigation methods and numerical simulations for estimating seasonal variations, flow paths and pollutant spread. The research focused on cold region roads, particularly a road test station located at Sweden. Specific objectives were:

To understand the behaviour of moisture dynamics and salt intrusion in road structures through assessing changes in measured electrical resistivity based on long-term in situ electrical resistivity tomography measurements; and to statistically analyse the relationship between electrical resistivity data and local weather data.

To assess electrical resistivity tomography monitoring methodology and test different analytical methods for better estimation of water movement in the road structure and the retardation time of de-icing salt, with estimation of the quantity changes.

To investigate the flow pathways in the road material beneath a typical highway surface during different seasons using non-destructive electrical resistivity tomography methods.

To develop and test non-destructive methods for tracing pollutant infiltration and percolation in various roads in Sweden.

To analyse the spread of pollutants and its pathways from road accidents.

To compare pollutant spread estimated from analytical calculations, modelling and field tracer tests.

To develop and test a two-dimensional simulation model for water and heat balance and flow in the modern roads.


5

To analyse the sensitivity of the parameters in the two-dimensional simulations, based on field data from an operational road test station.

To assess the importance of phase changes and frozen ground for flow patterns and flow models.

2. BACKGROUND

Natural hydrological cycles have shaped the landscape over millennia. The hydrological cycle involves different hydrological processes such as precipitation, runoff, freezing/thawing, evapotranspiration and percolation, which cause land erosion and re-distribution of different materials within a watershed. This process is mainly gravity-driven and water tends to take the shortest route to downstream recipients or groundwater (Freden 1994). Adding roads to a landscape tends to change the natural water and material movement, in addition to causing pollution. The road often acts as a dam in the landscape, changing the natural hydrological pathways. The road sometimes also acts as a drainage pipe and surplus runoff water may infiltrate into the road bank through the shoulder at the asphalt fringe (Kalantari 2014, Nickman 2016). Over time, this will affect the stability of the road and also facilitate transport of pollutants from road materials and traffic to groundwater. The work in this thesis consisted of two main parts. The first involved non-destructive environmental monitoring of roads using electrical resistivity tomography (ERT) methodology. The second part concerned water and heat transport in porous media. The results from the field measurements using ERT and the heat and from water transport simulations using one- and two-dimensional models were then combined, in order to obtain a better understanding of water and solute transport in roads.

2.1. Electrical resistivity tomography environmental monitoring and uncertainties

Electrical resistivity tomography was the main geophysical investigation method used in field investigations in this thesis. In the ERT method, the electric potential in the road is measured and subsurface resistivity is calculated and estimated by modelling. Besides being a non-destructive monitoring method, ERT has the advantage that the changes it detects in road layers are mainly due to changes in water content, minerals and porosity, which are the main parameters in the road that need to be investigated. The basic physical law behind the ERT method is Ohm’s law for calculating the resistance in the ground. The vector form of this equation is (Loke 2016):

[1]

where is the density of the current, is the conductivity of the

medium, which is equal to the reciprocal of resistivity , and

is the intensity of the electrical field. Resistivity of the ground can be calculated simply as:


6

[2]

where is the resistivity in ohm-m, V is measured voltage in volts, I is the current in amperes, A is the cross-sectional area of a conductor in m2 and L is the length of the conductor in m. Different arrangements and arrays have been developed by varying the distances between the current electrodes and the potential electrodes. For a typical arrangement with four electrodes, Eq. [2] can be rewritten as:

[3]

where A and B are the distances between the current electrodes and M and N are the distances between the potential electrodes. In this thesis, a 2D Wenner array and 3D pole-dipole arrays were used in different field studies. The Wenner array (Figure 1) is one of the electrical resistivity methods developed by Frank Wenner (1912a,b) and became popular through the Griffiths research group at the University of Birmingham (Griffiths and Turnbull 1985). The apparent resistivity for the Wenner array is given as:

2 ∆ [4]

where ∆V is the potential difference and is the spacing between electrodes. The Wenner array is a good method for data collection with high background noise (Loke 2016) and for monitoring horizontal structures (Barker 1979). A pole-dipole array is more suitable for 3D measurements (Loke 2016). The apparent resistivity for a pole-dipole array is given as:

2 1 [5]

where n is the distance between potential electrodes and is the distance between the potential electrode and the closest current electrode. The outer fixed current electrode is positioned at more than 10 times the distance between the other electrodes (Figure 2).

Figure 1 Electrode set-up and signal contribution in Wenner array, where C is current electrode, P is potential electrode and a is the distance between electrodes (Reynolds 2011).


7

Figure 2 Electrode set-up and signal contribution in pole-dipole array, where C is current electrode, P is potential electrode, a is the distance between the potential electrodes and (n×a) is the distance between potential electrode and the closest current electrode (Reynolds 2011).

In general, the resistivity of fresh water ranges between 10 and 100 ohm-m, but at low chloride concentrations can be down to 1 ohm-m (Reynolds 2011). The resistivity values for soil and rock types are generally higher than for metals and bedrock with high metal content. Several uncertainties need to be taken into consideration with ERT measurements. These uncertainties can arise from the precision and resolution of the ERT equipment during data collection or the inversion process. Environmental factors, such as freezing and thawing of soil water, and technical factors, such as degree of road compaction, can also be a source of uncertainties in time-lapse ERT measurements. Moreover, there is some uncertainty due to systematic error (e.g. poor contact between electrodes and the ground or breaks in the cables) (Loke 2016). In this thesis, bad data points in the initial data arising from systematic errors were removed in three stages, in order to decrease the uncertainties. Initially, some data were lost during measurements due to high contact resistance between the electrodes and the ground. Some of these bad data points were due to negative resistivities, which originated from ‘anomaly inversion’ as described by Loke (2016) (for more details see Paper I). Such bad data were removed from the raw data. In the second stage, bad data points were removed manually during the pre-inversion process using ‘Exterminate bad datum points’ in RES2DINV. In the third stage, in the post-initial inversion process using the cut-off from root mean square error (RMS) in RES2DINV, all the data misfits of more than 50% to 100% were removed. An example of removal of bad data and the resulting improvement in the quality of the data is presented in Figure 3, where Figure 3a shows an apparent resistivity pseudosection, Figure 3b the bad data points and measured data and Figure 3c the RMS error histogram from the display section of the RES2DINV software.


8

Figure 3 Initial data quality check for the first dataset, from measurements on 29 November 2013. a) Pseudosection plot, where Y is depth and X is electrode station, b) bad data removal window from RES2DINV and c) RMS error distribution histogram.

2.2. Modelling water and heat transport in roads Most modern road materials consist of partially saturated porous media. After the construction process, the road layers are left with moisture content as close as possible to the optimum moisture content for maximum bearing capacity. This condition soon


9

changes with time due to weather and loading factors. Moisture and heat in road layers are continually in a dynamic process in different seasons, depending on precipitation, temperature and the condition of the road asphalt and shoulder. To model moisture and heat changes in road, it is necessary to address the boundary conditions and material properties carefully. Small changes in boundary conditions cause a change in the degree of saturation in different layers of the road (Erlingsson et al. 2009). The geometry of the road section and location of the road in the terrain cause great variation in the boundaries, for example variations in the water balance. A general water balance equation for a section of road is as follows:

∆ [6]

where is precipitation, is external inflow, is surface runoff, is evaporation, is percolation to the groundwater and ∆ is the storage change. Units for each term are the unit of flow over an area. Water flows in different layers of the road depend on the particle size distribution. Sieve analyses and hydrometer tests can be performed to obtain a particle size distribution for a sample of the road material. The recommendations for suitable particle size distribution for road design and construction in different countries depend on the local soils and the local climate conditions. The grain size distribution makes it possible to determine other soil parameters such as the uniformity coefficient, the effective size, coefficient of gradation and the relationship between matric suction and water content for a given soil, which is known as the soil water characteristics curve (SWCC) (Arya and Paris 1981). Other parameters which need to be considered in the water flow are permeability and saturated and residual water content for a soil type. Heat transport in porous media is very closely related to water transport. Heat transport in soil can occur as radiation, conduction, convection and diffusion (Hermansson et al. 2009).

3. MATERIAL AND METHODS

3.1. Study site and field work Most of the field work was performed and data were collected from a road test station (Test site E18), operated by the Swedish Traffic Agency, with coordinates (59º38'0.2"N, 16º51'14.0"E, WGS84). This is an unique road test station near Stockholm, located on the E18 motorway between the cities of Enköping and Västerås (Figure 4). Nine other site investigations along old and modern roads located around Stockholm and in Småland, southern Sweden, were carried out for Paper III. Test site E18 is on a motorway which is built up of four typical layers based on Swedish standards (Trafikverket 2005). The station is equipped with sensors for simultaneous measurement of ground temperature, soil moisture content, air temperature, humidity and other


10

meteorological and traffic data. It is also equipped with electrodes beneath the top layer of the road surface for electrical resistivity monitoring across both lanes and along the highway (Figure 5). Most of the data used in this thesis were collected during 2013, when annual precipitation was 441 mm, mean air temperature was 6.7 ºC, groundwater level was between 5 m and 6.25 m below the surface and average daily traffic load was 18,884 vehicles per day (with 11% heavy trucks). Average annual de-icing salt use on this road between 2012 and 2015 was around 11.7 ton salt/km (Trafikverket 2015). In addition, tracer test data collection was performed during 2015 (Paper II). Test site E18 is located on post-glacial and glacial clay sediment overlying a layer of glacial till and bedrock (Earon et al. 2012). Monthly ERT data were collected from the main line, crossing both traffic lanes, together with other simultaneous data for statistical analysis, and used for setting the model boundary and parameter estimations.

Figure 4 (Left and top right) Location of the road test site on the E18 highway in central Sweden and (bottom right) illustrated cross-section of the road structure.


11

Figure 5 Data collection equipment installed at Test site E18.

3.1.1. Two-dimensional ERT acquisition Along the main line, 15 ERT acquisitions were made between 29 November 2013 and 19 December 2014, using a Wenner array of 64 electrodes with 1 m spacing and acquiring 581 data points each time. In these ERT acquisitions, an ABEM Terameter SAS 1000 instrument was used with the acquisition parameters shown in Table 1 (Paper I). Table 1 Acquisition parameter settings Parameter Setting (units)

Max. output current 50 (mA) Acquisition delay 0.3 (sec) Acquisition time 0.5 (sec) Total cycle 3.8 (sec) Error limit 1.0 % Minimum stack number 1 Maximum stack number 2 The inversion process for the collected data was performed using RES2DINV software. Loke and Barker´s (1996) least square inversion method was used with time lapse analyses (Loke et al. 2014) and L1-norm inversion (Claerbout and Muir 1973). This setting provides good resolution of the boundary, since there is a high contrast in resistivity of the soil and road material (Zhou and Dahlin 2003, Abdul-Nafiu et al. 2013). The topography was used in the inversion process and was incorporated into the model by shifting subsurface nodes along the same vertical mesh line (Loke 2016)


12

The data analyses following ERT acquisitions were performed using a new method of sorting the raw data into different zones of interest, i.e. different parts of the road structure. In Paper I the domain was divided into four zones (Figure 6): the road top layers (Zone 1), side slopes, road shoulder and middle strip (Zone 2), the zone between the road top layer and the groundwater (Zone 3) and the groundwater zone (Zone 4).

3.1.2. Three-dimensional ERT acquisitions and tracer test Sodium chloride was used as a tracer and monitored by 3D ERT measurements. Two different settings for ERT acquisitions were used. In both cases, a 64-electrode system and a pole-dipole array were used. In the first case, an ABEM_LS instrument was used in four lines of 16 electrodes (4×16), with spacing 0.5 m and 0.4 m between lines and electrodes, respectively (Figure 7). The purpose in this case was to detect the flow pathways in different seasons. A tracer consisting of 50 L water and 30 g/L NaCl was applied uniformly along line 4, using a perforated pipe. Each acquisition consisted of 736 data points and each measurement series took about 30 min.

Figure 6 Road section zones used to divide resistivity data into different groups (data points shown without considering topography) (Paper I).

Figure 7 Electrode set-up (4×16) used on the E18 road shoulder for tracer test monitoring by electrical resistivity tomography (ERT) for detecting flow pathways during different seasons (Paper II).


13

In the second case, an ABEM SAS4000 instrument was used in eight lines of eight electrodes (8×8) with two different electrode spacings, 0.2 m and 0.4 m, as shown in Figure 8 (Paper III). The time lapse between different measurement series was between 30 and 120 min, and each resistivity measurement series took about 20 min. The acquired data comprised 245 acquisitions and finite difference modelling was carried out using RES3DINV (Loke 2007). Further calculations were performed using ResCalc (©Olofsson), Matlab-B, OriginLab (OriginLab 2015) and pseudo-3D plots made by Voxler. In the second case, 50 L of water and 1000 mg/L NaCl were used as a tracer and all 50 L were released instantly at the asphalt fringe to simulate a road accident. It was not possible to transform the resistivity values into mass transport of water and pollutant because the initial water content and porosity beneath the top layer were unknown. Therefore, the pollutant spread and pollutant pathways were estimated based on a threshold change in modelled resistivity. Accuracy decreases and uncertainty increases with time for penetration depth estimates. A threshold of 25% was selected in this case, based on previous field studies (Aaltonen and Olofsson 2002, Lundmark and Olofsson 2007, Olofsson and Lundmark 2009). Infiltration capacity was also measured at the test location, using a single ring infiltrometer (Ø 14.5 cm) for comparison. However, it was not possible to take accurate measurements of infiltration due to the coarse material on the shoulder. Sieve analyses and hydrometer tests were performed for the top layer soil samples.

3.1.3. Estimation of chloride concentration from ERT Estimation of chloride concentration in the road structure was carried out at the E18 test station. The first step in estimation of chloride concentration was to correct the resistivity data due to measured soil temperature. Correction of the data to 25 °C temperature was performed following Keller and Frischknecht (1966) and Brunet et al. (2010):

č 25 [7]

where is electrical resistivity corrected to 25°C, is electrical resistivity at temperature T and č is a coefficient obtained empirically and often equal to 0.025/°C (Brunet et al. 2010). The soil temperature was measured by a temperature probe consisting of a set of 41 PT100 sensors, extending 2 m vertically down from the road surface (Paper II). This system was manufactured and installed by the Swedish National Road and Transport Institute, VTI (Wilhelmsson 2017). The annual temperature variation in the road layers measured over two years by some of the 41 sensors is shown in Figure 9. Some of the 41 sensors, representing the ground temperature at different depths, were selected for further analysis.


14

Figure 8 Electrode set-up (8×8) in electrical resistivity tomography (ERT) tracer test monitoring for analysing the spread of pollutants after road accidents (Paper III). Chloride concentration changes were estimated for the resistivity changes based on Archie’s law, which is an empirical equation between resistivity and salinity obtained for sand samples with porosity range 10% to 40% (Archie 1942):

[8]

where is the resistivity of saturated sand, is resistivity of fluid in the pores and ( ) is a formation factor obtained from porosity and an empirical parameter related to pore connectivity (cementation exponent). For poorly cemented material, the parameter m is equal to 2, based on the literature (Keller and Frischknecht 1966, Singha and Gorlick 2005). In this thesis a value of 2 was used, due to low clay content in road material. Based on Meinzer (1923), the porosity range was assumed

Figure 9 Soil temperature measurements obtained from the Swedish National Road and Transport Research Institute’s PT100 sensors (all 41 sensors) at the test site on the E18 highway (Paper I).


15

to be between 21% and 31% by volume for a sandy and gravelly mix. Following Singha and Gorlick (2005, 2006), the change in apparent conductivity was used to estimate the chloride concentration:

∆ 3.19 ∆ [9]

where ∆ is the change in chloride concentration and it is estimated from electrical resistivity in mg/L and ∆ is the apparent conductivity change in mS/m. The calculation procedure was as follows: Conductivity data were corrected for measured soil temperature using equation [7]. Seasonal conductivity changes from the first measurements, in April 2015, were then calculated using equation [9]. Finally, for each measurement the sum of all changes in the chloride concentration was calculated.

3.1.4. Climate data and statistical analyses Most of the data on road and climate used in this thesis were obtained from the test station on the E18. Data on precipitation were downloaded from the website of the Swedish Meteorological and Hydrological Institute (SMHI 2017). Precipitation, air temperature and moisture content data for the base material at 19 cm below the asphalt layer were collected using seven time domain reflectometry (TDR) sensors (type TDR100; Campbell 2015) (Figure 10). These data were used both in statistical analyses with ERT data and in the model boundary and initial conditions. Some of the data (i.e. cumulative precipitation, local moisture content from TDR measurements, air temperature and ground temperature) were analysed statistically to detect any spatial and temporal changes in the pattern of electrical resistivity, to find any related correlation with changes in resistivity by dividing the resistivity data into different zones (Paper I). A non-parametric test was used in statistical analyses of the change in electrical resistivity and other climate data, using the Spearman´s rank correlation method. For more details about Spearman’s rank correlation, see Paper I.

3.2. Modelling approach

3.2.1. Water transport Water and vapour flow in unsaturated porous media in road layers was described by the mass conservation equation:

q_l q_v S [10]

where q_lis liquid water flux (m/s), q_v is vapour flux density (m/s), S is transpiration (L/s), t is time (s) and θ is total volumetric water content (m3/m3), and:

θ θ θ θ [11]

where θ is volumetric water content (m3/m3), ρ is specific gravity (kg/m3) and i, l, and v correspond to ice, liquid water, water and water vapour, respectively.


16

Figure 10 a) Geometry of the road section, with points indicating the location of soil moisture from time domain reflectometry (TDR) sensors and temperature measurement sensors at the surface of the road (T 0.0) and down to 2 m below the road surface (T-2.0). b) Precipitation, air temperature and moisture content readings from all sensors over two years.

The equation for liquid water flow in 2D porous media, based on Richard’s equation for water flow in unsaturated frozen soil (Harlan 1973), and for water vapour from sum of isothermal flux


17

and thermal flux components (Philip and de Vries 1957) is given as:

q qh qT K h y K T [12]

where qhand qT are isothermal and thermal water or vapour density (m/s), h is water pressure head (m), y is vertical coordinate (m), which is equal to zero for vapour flow, K is isothermal hydraulic or vapour hydraulic conductivity (m/s), K is thermal hydraulic or vapour hydraulic conductivity (m2/K/s) and T is soil temperature (K). After solving for water and vapour flow due to pressure head and temperature gradient from [11] and [12], the following equation was applied:

. K h y K T . K h

K T [13]

where K and K is isothermal hydraulic conductivity for thermal and vapour components (m/s), respectively, and K and K is thermal hydraulic conductivity for hydraulic and vapour components, respectively (m2/K/s). For solving this, a formula for water through porous media in COMSOL Multiphysics® software (COMSOL 2016) was used to solve the partial differential equation (PDE) coefficient. The main PDE equation for water content was:

∙ c Theta αTheta γ β ∙Theta aTheta f [14]

where , , is mass coefficient (s), is damping factor

or mass coefficient (1), c is diffusion coefficient (m2/s), a is absorption coefficient (1/s), f is a source term (1/s), α is conservative flux convection term (m/s), β is convection coefficient (m/s) and γ is conservative flux source (m2/s). From equation [11], the source term was:

f [15]

All the other coefficients were assigned a value of zero except for the damping factor for mass coefficient, which is equal to 1, and conservative flux source, which is γ (q_l+q_v for both coordinates. Soil hydraulic properties Hydraulic properties for unsaturated porous media were predicted based on the van Genuchten (1980) and Mualem (1976) equations:

h [16]

S | |

1 [17]


18

K K S 1 1 S

K [18]

where K is saturated hydraulic conductivity (m/s), K is isothermal hydraulic conductivity (m/s), S represents effective saturation and n,m and α (1/m) are the van Genuchten-Mualem parameters. Thermal hydraulic conductivity was calculated after Saito et al. (2006):

K K 0.1425 4.76 10 T 273.15 [19]

where G is a gain factor and r is surface tension at 25 ºC (71.89 gm/s2). Diffusion of water vapour in the model was calculated based on Fick’s law (Saito et al. 2006) (for more details, see Paper IV). Frozen water content was calculated based on the empirical relationship by Xu et al. (2010):

θ 0.01 a T 273.15 [20]

θθ

θ . [21]

θ θ θ [22]

where a and b are empirical coefficients and θ is volumetric liquid water content (m3/m3) in relation to the frozen temperature.

3.2.2. Heat transport The heat transport equation described by Zhang et al. (2016) was used to create a fully coupled model of heat and moisture by considering heat conduction and convection of water vapour and liquid, as well as evaporation and phase change in the freezing and thawing process. The main equation for energy conservation is:

C λ T q_l C q_v C T L ρ

L ρ L ρ q_v [23]

where C is volumetric heat capacity of moist soil (J/m3/K) and consists of the sum of volumetric heat capacities of each component of solid, liquid and ice multiplied by its fraction. L is

latent heat of ice fusion 3.34 10 , L is latent heat of

water vaporisation and depends on temperature change,

L 2.501 10 2369.2T , and λ is the moist soil thermal conductivity (W/m/K):

λ λ λ λ λ [24]

where λ , λ , λ and λ is the thermal conductivity of soil, ice, liquid and air (W/m/K), respectively. The main PDE equation in COMSOL for heat transport is:


19

∙ c T αT γ β ∙ T aT f [25]

where , , is mass coefficient (kg/(m.s.K)), is

damping factor or mass coefficient ((J/m3/K)), c is diffusion coefficient (W/m/K), a is absorption coefficient (W/m3/K), f is a source term (W/m3), α is conservative flux convection term (J/m2/K), β is convection coefficient (J/m2/K) and γ is conservative flux source (W/m2). In the present case, mass coefficient , diffusion coefficientc λ and convection coefficient was:

β q_l C q_v C [26] The source term was:

f L ρ L ρ L ρ q_v [27]

3.2.3. Boundary conditions The upper boundary temperature was taken as the measured temperature at the road surface at the test station on the E18. Precipitation data for this location were obtained from the Swedish Meteorological and Hydrological Institute (SMHI 2017). After running the initial model for 365 days, the initial values for water content and temperature were obtained for further simulations. The assumed initial water content was 0.3 m3/m3 and the temperature was assumed, based on the measured temperature probe data, to be -1 ºC at the surface and +8 ºC at the lower boundary of the model, after using extrapolation from available data at 2 m depth. Based on the literature (Piguet 2007), the infiltration through asphalt is as low as 1.8%. In this thesis, the hydraulic conductivity of the asphalt was assumed to be very low (1×10-10m/s). Hence, the asphalt was almost impermeable. Runoff is greater on the shoulder of the road due to the granular material and the slope. The coefficient of runoff is 0.3 to 0.9, depending on the type of shoulder (Piguet 2007). In this thesis a flux added to the shoulder represented the infiltrated precipitation. Due to high granular material in the shoulder, low surface runoff was expected and was surface runoff was anticipated only to occur during high-intensity rainfall. Based on the intensity of the rainfall, the flux varied. For low-intensity rainfall (<0.2 x 10-7 m/s), infiltration into the shoulder was assumed to be 100%, i.e. all precipitation was assumed to infiltrate, while for higher-intensity precipitation infiltration into the shoulder was assumed to be 35%.

3.2.4. Simulations and scenarios The geometry of the road model for simulations was similar to the actual design of the road. The basic model geometry shown in Figure 11 was used. Model dimensions were 10 m width and 3 m depth. Only a part of the road was modelled, because most infiltration of water occurred in the shoulder according to the results from the field investigations and monitoring at earlier stages (Paper I, Paper III).


20

Figure 11 Basic geometry of the road model, with layers of mesh.

Several different scenarios were tested to see the difference in the flow and frozen water content patterns. Adding a layer of clay on the shoulder, increasing and decreasing the thickness of the road layers and different side slopes were tested. The extremely fine physics-controlled mesh in COMSOL Multiphysics® software (COMSOL 2016) was used in all scenarios. The relative tolerance, which controls the internal time step of the solver, ranged from 0.01 to 0.00001.

4. RESULTS AND DISCUSSION

4.1. Monitoring water and salinity in roads (Paper I) The ERT field investigations were very useful in providing an initial assumption about water flow pathways and salinity changes in the road structure. Figure 12 shows inverted resistivity and percentage change in inverted resistivity for the period between November 2013 and December 2014 with root mean square error ranges between 6 and 13.2 %. Significant differences in resistivity values were observed between the road upper layers and other zones. The average resistivity values for the groundwater zone and below the road structure (Zone 3 and Zone 4 in Figure 6) were less than 100 ohm-m. The highest average resistivity (more than 900 ohm-m) was found in the road structure and shoulder layers (Zone 2 in Figure 6) and varied more than within the other zones. This can be due to the fact that the upper layers had more material heterogeneity and lower moisture content and were more influenced by weather changes. Visual interpretation of Figure 12 indicates that seasonal flow occurs mainly through the road shoulders. In order to understand and analyse the seasonal flow pattern, changes in resistivity for different zones of the road structure and layers (see Figure 6) were analysed and compared against climate data. Both increases and decreases in resistivity (Figure 12a) were analysed, in


21

order to detect the small changes in resistivity that are mainly due to changes in chloride concentrations or water content (Figure 13).

Figure 12 a) Inverted resistivity distribution in the road cross-section at different times and b) percentage change in resistivity relative to the first measurement (Paper I).

The average changes in the logarithmic resistivity are presented in Figure 13a, and all the negative and positive values are separated out in Figures 13b and 13c, respectively. The standard deviation of the decrease in resistivity is presented in Figure 13d. Since de-icing salt causes reductions in resistivity values, decreases in resistivity were of more interest in this thesis.


22

Figure 13 a) Average (spatial within each zone) percentage change in resistivity logarithm, b) percentage resistivity change (only the cells with positive changes considered) in comparison with the first measurement, c) percentage resistivity change (only the cells with negative changes considered) in comparison with the first measurement and d) standard deviation of resistivity changes (only decreasing cells considered) in different zones (1-4) of the road section (Paper I). The average resistivity of the top layers was rather similar at the beginning and end of the year. The changes in resistivity in deeper layers (Zone 4) started to decrease after April and reached their


23

lowest value (-2%) by the end of the year. The resistivity decreased in all zones when only the negative resistivity changes were considered (Figure 13c). The decrease in Zone 4 can be an effect of accumulation of traffic pollutants and de-icing salt from winter maintenance. The changes in resistivity in Zone 1 started to increase after spring and were -3% by the end of the year. Until October, Zone 2 and Zone 3 were fairly well correlated, and then the changes in Zone 2 continued to decrease to -9% by the end of the year. The downward transport of water and solutes accelerated in the base and sub-base within Zone 1 after spring, but in Zone 3 only after August, while Zone 2 acted as the main transport pathway. Zone 4 was the destination zone for vertical solute transport, especially de-icing salt, whereas the other zones only showed a small decrease in total resistivity changes by the end of the year. A quantitative description of the heterogeneity in resistivity changes is also presented, as standard deviation of the resistivity decrease (Figure 13d). Zone 4 showed the smallest variations in resistivity values and Zone 2, including the roadsides and all road shoulders, showed the largest variations. The changes in electrical resistivity data were analysed statistically and compared with climate data, i.e. precipitation, air temperature and ground temperature, in four different zones and 15 different layers. Local climate conditions contributed directly or indirectly to the resistivity changes. Each of the climate factors showed a negative or positive correlation with the change in electrical resistivity, but due to the combined effect of all factors the correlation coefficients were low. Looking at one of the correlations between the resistivity changes and the temperature changes in the layers (Figure 14a), it is clear that the layers down to 1.78 m were most directly affected by the temperature changes. The moisture data from the seven TDR sensors (see Figure 10) revealed that during periods when the temperature was below zero, the moisture content measured by sensors TDR-1, TDR-2 and TDR-3 for the first three months in 2013 was less than 10% and almost constant. The moisture content measured by the other sensors (TDR-4, TDR-5, TDR-6 and TDR-7) showed relatively larger fluctuations for the same period. This indicates that during freezing periods, flow of water and moisture still occurs through the road shoulder and side slopes. This might be caused by the use of de-icing salt, which decreases water freezing point, and hence will cause existing ice to melt. Analysing the mean resistivity values of the whole road section without dividing the road domain into different zones was not very informative and changes over time could scarcely be seen (Paper I). Therefore dividing the road domain into four different zones was essential in order to present and analyse the electrical resistivity distributions and percentage change, as well as to study their correlations with climate and environmental conditions. Correlations between moisture readings from TDR sensors and resistivity change at 25 cm depth in Zone 1 are presented in Figure 14b. Because the TDR sensors were located only within the uppermost layer, this layer was selected for


24

further statistical analysis. The results indicated a negative correlation between resistivity changes and moisture data from TDR sensors in most cases. The only positive correlation found was for TDR-5, which is located directly beneath the road shoulder. This is mainly because the road shoulder is highly permeable and enables relatively fast water and temperature movements, as observed from the resistivity change. From the TDR measurements, it can be seen that the TDR sensors in the road shoulder reacted differently regarding moisture content measurements than the TDR sensors under the pavement and at the side slopes. This could be due to water movement in the vertical direction, which dominates in the road shoulder. Overall, inverse correlations were observed between electrical resistivity changes and moisture content. The road structure beneath the asphalt layer acted as storage for moisture and ions (e.g. chloride), which were partly spread into deeper road layers. By the end of the winter, most of the de-icing salt applied had been washed down to groundwater.

Figure 14 a) Spearman correlation between temperature and the change in resistivity in different layers, and b) correlation between moisture content


25

from time domain reflectometry (TDR) and the resistivity decreases in Zone 1 and layer depth 0.25 m (Paper I).

4.2. Monitoring flow speed and pathways using tracer test (Papers II & III)

Tracer tests with infiltration mainly into the road shoulders were performed in Papers II and III. On analysing ERT measurements, different patterns in resistivity changes were observed during different seasons. Figure 15 shows an 1D presentation of the resistivity distribution with depth for each line and during different seasons. The reduction in resistivity was found to be greater in lines 3 and 4, as the tracer was added beside line 4 (close to the paved surface, while line 1 was located 2 m from the paved surface, close to the inner ditch; see Figure 7). In April, small uniform changes in the patterns of the resistivity changes were observed, indicating slow and shallow infiltration of the tracer. Significant negative changes were observed during August and gave a more scattered pattern compared with in December, indicating unpredictable and highly non-uniform spatial behaviour. In August, the tracer infiltrated faster than in December and infiltrated mostly vertically, without large spread on the surface. More clustered negative resistivity changes were observed for measurements during December, showing a very strong decrease in resistivity for all measurement lines except the first time lapse for line 1 (at longest distance from the asphalt). This could indicate the presence of preferential flow paths, with measurements still unaffected by the tracer (Figure 15).

(a)


26

(b)

(c)


27

(d)

Figure 15 Percentage logarithmic change in resistivity in a) line 1, b) line 2, c) line 3 and d) line 4, with time lapses of 0.5, 1.0, 1.5 and 2.5 hours. *The last time lapse was not fixed and was 22 hours in the April tracer test.

Plotting the average resistivity for all measurements as a function of depth (Figure 16) showed similar results as in the 1D plots. The tracer test in April affected the road layers down to a depth of about 0.4 m, which indicated that the tracer remained at shallow depths forming a plume within the top layer. This in turn indicated that the tracer flow was retarded, probably due to the high degree of water saturation in the soil. In August, the spread of tracer in the top layer of the shoulder was small. However, the percolation downwards occurred through many pathways. During December the spread on the surface was larger and the tracer infiltrated and percolated downwards, clearly indicating preferential flow with fewer pathways. Figure 17 shows combined 2D resistivity measurements for the second time lapse (=1 h). The resistivity changes were significantly smaller in April than in August and December. In April, the resistivity was reduced only at the top layer close to line 4, adjacent to the road. However, the measurements taken during August (Figure 17b) and December (Figure 17c), using a similar volume of tracer (50 L of water with 30 g/L NaCl), caused a reduction in resistivity within the entire depth (2 m). This indicates fast percolation downwards. However, in August the spread of resistivity changes on the surface was less than in December. The analytical outputs from Res2DInv were also used to create pseudo 3D plots. The inversion error for the 2D resistivity lines was less than 5% in all four lines. Figure 18


28

presents the results for the first time lapse of the tracer test made in April, August and December with 22% of the resistivity reduction. As the diagram clearly indicates, the tracer caused the resistivity changes and the pattern of the spread to vary between different seasons. In April the changes were only found near the surface, whereas in August the changes extended downwards along the whole tracer test site, reaching at least a depth of 2 m. In December, clear preferential flow paths could be detected. Similar results were found in all time lapse analyses. Thus similar conclusions could be drawn from the results in the 1D and 2D analyses.

Figure 16 Percentage average resistivity change with depth in the road section in different seasons.

Figure 17 Percentage change in resistivity for all lines in combined 2D presentations after the second time lapse (1 h) in the tracer tests in a) April, b) August and c) December.


29

Figure 18 Resistivity changes in the road section plotted in pseudo 3D for the first time lapse with 22% resistivity reductions of the tracer in a) April and b) August and (c) December.

The results from the infiltration measurements are presented in Figure 19. In April the infiltration was about 0.02 mm/s, which was much lower than in August and December. In August, infiltration was faster after the first set of measurements, since the initially dry road material impeded the infiltration process. In December the initial infiltration velocity was significantly higher than in April and August but, due to some frozen pores and ground, the flow did not reach stable conditions. Lateral flow occurred within the top few centimetres of the soil and was observed during the test. In general, the initial infiltration was high in all seasons and it varied at different locations depending on varying grain sizes in the shoulder. Within less than 10 minutes, the average infiltration at most locations was between 0.02 and 0.04 mm/s, which is low compared with the time needed for 50 L of tracer to infiltrate over an area of approximately 4 m2.

Figure 19 Infiltration values obtained using the double-ring infiltration test for the E18 road shoulder at the test station site (all measurements).


30

During the tracer tests, the time needed for the entire volume of tracer to infiltrate was less than 3 minutes. Based on the tracer test, the average infiltration was estimated to be 0.07 mm/s, which was higher than determined in the double-ring infiltrometer test. In order to estimate the changes in chloride concentration based on the resistivity changes from the tracer tests, a sum of all changes in the chloride concentrations was calculated and is presented in Figure 20. Overall, chloride concentrations in the soil road layers increased during the tracer experiment period. The tracer test in April showed very small changes in chloride concentration during all measurement series, probably due to retardation of the tracer in the upper layers. In the background measurements before each tracer test, the changes in chloride concentration showed a very small difference from the original conditions, probably due to de-icing salt from previous tests being crystallised and stored in the road material. After adding water with more tracer, the measurements showed elevated chloride concentration, probably due to stored crystallised salts being dissolved. The spread of values generally increased during the tracer tests, showing the heterogeneous character of tracer spread. After the third measurement series (usually 2.5 h) the average chloride content decreased, probably due to the number of cells affected by the tracer decreasing when the tracer percolated downward to the groundwater. The last measurement in December (Dec16-05) showed a drastic reduction in the chloride concentration, either due to rapid transport downwards along distinct flow paths or due to freezing of the infiltrated water. Figure 20b shows the corresponding changes in average resistivity Even though there was preferential flow during the frozen period, the spread of values indicated some retardation of chloride in the road layers.

4.3. Spread of pollutant from traffic accidents (Paper III) An estimate of the transport time of pollutants within the road structure after traffic accidents was made based on a number of tracer tests using 3D ERT measurements along existing roads in southern and central Sweden. Analyses were compared to analytical and numerical calculations using analytical solutions, 1D simulations with CoupModel and 2D simulations with COMSOL. (Paper III). Based on ERT data, the estimated percolation velocity, penetration depth and volume of affected soil at 1.2 m depth for all nine locations studied are shown in Table 2. The resistivity measurements indicated fast downward flow, ranging from 0.5 to 3.1 m/h. Two types of road were studied, modern roads built according to Swedish standards (Trafikverket 2005) and old roads built from local soil material. The fastest infiltration rate (1 m/20 min) was observed in modern roads (roads 897 and 261). The highest amount of soil volume affected by the tracer was found at these sites.


31

Table 2 Calculated flow velocity, transport time and volume of soil affected at different road measurement sites. Depth penetration = depth where resistivity decreased by 25%. Time = time lapse between infiltration and start of resistivity measurements. Volume of soil affected = relative amount (%) of the model block at depth 1.2 m affected, i.e. where the resistivity had decreased by more than 25% (Paper III). The location of the test sites is shown in Figure 5 in Paper III

Measurement site Depth

penetration (m)

Time (h)

Percolation velocity (m/h)

Transport time to depth 1 m (min)

Amount of soil affected at 1.2 m depth (%)

Road 897, Skirsvad 1.57 0.5 3.1 19 35

Road 875, Berghem 0.92 2 0.5 130 0

Road 126 I, Torpsbruk 1.39 2 0.7 86 4

Road 126 II, Torpsbruk 0.84 1 0.8 71 0

Road 261 Lovön 1.52 0.5 3.0 20 18

Road 261 Ekerö Rasta 1.55 0.5 3.1 19 31

G:a Nynäsvägen, Jordbro 0.91 0.5 1.8 33 0

Road E4/E20 Salem 0.70 0.5 1.4 43 0

Road E4 Kista 1.38 0.5 2.8 22 10

Analytical calculations, considering a worst case scenario of fully saturated conditions, for different road materials showed a very long transport time down to a depth of 1 m for roads made from natural soils, such as clay and till. However, calculations using 1D CoupModel resulted in transport times of about 14 h for these roads. For sand and gravel, the analytical solution (max. 0.5 h) was significantly faster than the dynamic model 1D model solution (4 h) (Figure 21). The road shoulder in modern roads, made from coarse rock fragments, showed much more rapid percolation than the shoulder in roads based on natural soils. Therefore, in modern roads the time window for action such as digging after traffic accident-related spills is very limited. The 2D COMSOL modelling gave similar results. Spillage of water-soluble pollutants was simulated for two typical types of roads for up to 500 minutes (Figure 22).


32

Figure 20 a) Estimated chloride concentration changes for all measurements in comparison with the background measurement in April and b) percentage resistivity changes for all measurements in comparison with the background measurements (C1-C5 represent the measurement series performed 0.5-22 h after injection).


33

Figure 21 Calculated depth from one-dimensional CoupModel for the chloride front beneath the road surface in different materials at 1 and 4 hours (Paper III).

Figure 22 Two-dimensional simulation results for two road types in COMSOL Multiphysics 5.2, showing the saturation changes due to added flux at the shoulder for: (left) a modern road made from rather coarse material and (right) an old road consisting of local natural materials (Paper III).


34

A modern road was modelled for 100 minutes and an old road was modelled for 500 minutes. It took 5 minutes for the tracer to reach the sub-base material in the modern road, whereas in the old road it took about 30 minutes for the flow to reach similar depth. A depth of 1 m was reached in less than 1 h in the modern road, whereas the entire simulation time (500 minutes) was not enough for the flow to reach 1 m depth in the old road type. The importance of the surrounding geological material for groundwater vulnerability assessments along roads has been highlighted previously (e.g. Gontier and Olofsson 2002, Thunqvist 2003). However, a traffic accident may release pollutants which infiltrate in the coarse road shoulder and percolate downwards through the road embankment. In analysing pollutant spread in roads, it is not appropriate to use geological maps showing natural soils, since percolation velocities are usually much lower in natural soils than in the coarse material used to construct modern roads. Based on the resistivity measurements, the calculated transport times were of the same order of magnitude as the analytical calculations for sand, whereas for the road shoulder and the inner ditch slope the resistivity results resembled those from dynamic 1D modelling. Reasonable percolation velocities were obtained from the 2D COMSOL simulations for base and sub-base material, but generally longer times than the analytical calculations and shorter than 1D modelling values for both types of road. Dynamic modelling is an excellent tool for vulnerability assessment of roads if the structures of the road and material hydraulic properties are known. Tracer tests combined with resistivity measurements and 3D inverse modelling are excellent non-destructive methods for identifying flow paths and transport times. However the approach lacks accuracy, since it is based on average values of the resistivity conditions at the specific measurement time. Due to preferential flow, individual molecules will not be detected in the resistivity measurements due to fast downward transport.

4.4. Modelling heat and water content changes while considering phase changes (Paper IV)

Parameter estimation was performed based on the parameter ranges from Table 3. Fifteen simulations were made for each parameter, which resulted in the curves shown in Figure 23. The diagram shows the root mean square error (RMSE) from comparison of measured and simulated values of liquid water content and temperature at selected points, representing the location of the sensors. One parameter was optimised at a time and the other parameters were fixed based on literature values (Table 3) (De Vries 1963, Kömle et al. 2007, Zhang et al. 2016). Saturated and residual water content were fixed at 0.45 and 0.05 m3/m3 by volume, respectively. Figure 23a shows the RMSE values for different thermal conductivities, while Figures 23b and 23c show the results of simulations from different saturated hydraulic conductivity values (Ks) and Mualem-van Genuchten


35

parameter values (alpha), respectively. Divergence from the ideal simulation occurred on using higher values of alpha. Thermal conductivity of the soil, alpha and n were most influenced by the temperature changes. Figure 23d shows the RMSE values for Mualem-van Genuchten parameter n and Figures 23e the RMSE values for coefficient a. Coefficient a had no significant influence on the model results, while coefficient b (Figures 23f) was inversely proportional to RMSE values.

Table 3 Parameters used in parameter estimation. Some parameter ranges are from the literature (De Vries 1963, Kömle et al. 2007, Zhang et al. 2016)

Variable and parameter Initial fixed value (units)

Selected value (units)

Lower value

Upper value

Reference Sign Name

alpha Mualem-van Genuchten

1.6 (1/m) 2 (1/m) 1.6 1170 Hansson et al. (2005)

n Mualem-van Genuchten

1.4 (---) 1.5 (---) 1.37 4 Hansson et al. (2005)

lms Therm. cond. of soil matrix

8 (W/m/K) 2 (W/m/K)

0.4 10 De Vries (1963)

a Coefficient estimated by regression

10 (---) 10 (---) 0.1 500 Zhang et al. (2016)

b 0.5 (---) 0.5 (---) 0.1 1000 Zhang et al. (2016)

Ks Saturated hydraulic conductivity

10-5 (m/s) 10-6 (m/s) 10-4

10-13

Based on material size distribution

The heat simulations were quite accurate, as can be seen in Figures 24a and 24b, which present the measured and simulated temperatures for two points in the road layer at a depth 0.2 m and 1.0 m from the road surface. The results for all depths from 0.2 m to 2 m are compared in Figure 24c. The correlation coefficient of the heat simulations at the different depths is shown in Figure 24d. As can be seen from the diagram, the upper layer (0.2 m) and lower layer (2 m depth) has the lowest correlation coefficient, around 0.8. This was mainly due to uncertainties in the boundary conditions. Thus small changes in the boundary conditions can change the model results.


36

Figure 23 Root mean square error (RMSE) for moisture (left-hand diagrams) and temperature (right-hand diagrams), arising from varying the parameters: a) Thermal conductivity of soil matrix, b) saturated hydraulic conductivity, c) Mualem-van Genuchten’s alpha, d) Mualem-van Genuchten’s n, e) coefficient a and f) coefficient b.


37

Figure 24 Comparison of measured (T) and partial differential equation (PDE) model-simulated soil temperature (Ts) at: a) 0.2 m depth, b) 1 m depth and c) all depths from 0.2 m to 2 m, and d) Pearson correlation coefficient between simulated and measured temperature for different soil depths. The comparison between measured and modelled liquid water content showed good performance regarding the seasonal dynamics but the simulations could not accurately predict the average water content (Figure 25). There were some differences between the measured and simulated moisture values at different TDR sensors, which could be due to errors in sensor calibration or, more probably, material differences, since TDR-6 and TDR-7 are located close to the inner ditch slope with more fine-grained material, while TDR-5 it is located close to the pavement in the road shelter with coarser material. Due to assumptions and simplifications in the model, the differences between the sensors regarding physical properties were not taken into account and the simulated values for all points were quite similar.


38

Figure 25 Comparison of liquid water content measured by time domain reflectometry (TDR) sensors at test station E18 and simulated using partial differential equation (PDE) models for three different TDR sensor locations: a) Beneath the road shoulder (TDR-5) and b, c) at the side slope (TDR-6, TDR-7). The TDR sensors are all positioned in the unbound base layer, at a depth of 20 cm, but the outermost sensor is slightly closer to the ground surface due to side slopes. Unbound base material allows preferential flow in addition to frozen water in the pores that cannot be measured by TDR. This may explain the varying water content and the significant time fluctuations in the measured water content. The simulated liquid water content was generally lower than the measured values during summer at TDR-6 and TDR-7, but higher at TDR-5, probably due to differences in evaporation between the slope and the shoulder material which are not considered in the simulation model. Simulated liquid and frozen water content for some selected days out of 365 simulated days are shown in Figure 26. The selected


39

days are based on days with sudden local climate data, i.e. precipitation and temperature changes. Most of the significant changes occurred during January, April and December. In January the ice content in the structure increased, starting from the uppermost layer and continuing downwards until April depending on the actual soil temperature. In spring (12 April 2013), the temperature increased and high precipitation amounting to 17.3 mm was recorded.

Figure 26 Simulated liquid water content (left-hand diagrams) and frozen water content (right-hand diagrams) at some interesting times during the study year.


40

Figure 27 Frozen water content in winter (January) and spring (April) in selected scenarios involving changes to model parameters. The situations in April occurred after intensive rain of 17.3 mm.

The ice content in the frozen layers started to melt, especially under the shoulder, as the rain water infiltrated. Some simulated frozen water under the pavement was still observed after one week, despite the increase in temperature and heavy precipitation. After two days no ice content remained in the road structure according to the model until December, when new ice started to form again in the pores. The upper boundary condition was very important in simulation of water and ice content in road layers. In order to analyse the impact of changes in the upper boundary, several simulations were performed with various materials and layer thicknesses of the road shoulder and slopes. Figure 27 shows some of the selected scenarios regarding ice formation in the road structure between January and April. Adding a layer of clay decreased infiltration of water into the shoulder and, as a result, the ice layer extended beneath the shoulder for a longer period of time. During April and after intensive rain amounting to 17.3 mm, water percolation through the road structure decreased and the frozen layer melted more depending on the heat changes rather than due to the percolated precipitation. An increase in the outer slopes of the ditch caused less frozen water content in the shoulder and at the centre of the ditch, since more water infiltrated into the road and the ditch. When the surface was flat, however, the shoulder became totally frozen and the intensive rain could not melt the ice


41

formation immediately. Changing the thickness of the soil cover on the shoulder caused very small changes in the frozen water content. Changing the ditch slope also caused a different distribution of the ice formation in comparison with the background model. Therefore, it is clear that the distribution of the frozen water content can change based on road geometry and material properties. Snow ploughing, which causes accumulation of snow on the road shoulders, can be assumed to affect the melting situation but was not studied in this thesis, since it requires combined simulation of snowmelt on the surface and the current model beneath the snow.

5. FINAL DISCUSSION

The work described in this thesis comprised several stages, which linked to each other to achieve the general aim of understanding water and pollutant movements through road layers to the groundwater. In order to study the hydrological impacts on roads and the resulting pollution on groundwater, seasonal climate data and soil data were collected from a road in operation. Since it is not practical to interrupt or block traffic for data collection, non-destructive methods, such as electrical resistivity tomography (ERT), can be an effective tool for hydrogeological mapping and other environmental applications (Dahlin 1996, French and Binley 2004, Auken et al. 2006, Loke et al. 2014). In Paper III, a full-scale tracer test was applied on different road shoulders, but only during the summer season, to simulate the spread of water-borne pollutants from road accidents. Several geoelectrical arrays were tested in order to trace the simulated pollutant infiltration into the road structure. An electrode pattern of 8×8 electrodes was selected, since the spread direction and the extent of pollution were not known at the beginning of the survey. The tracer (water and salt) was instantly released, to simulate an accident and sudden release at the road shoulder. Based on the tracer test results and modelling of the percolation time and speed, the pollutant spread in different types of roads was studied during the summer season and for several road sections. Seasonal electrical resistivity changes and their correlation to seasonal variations and physical material properties required data collected from fixed ERT lines installed beneath the road lanes. These data were obtained from an environmental road test station on European highway E18 (Test site E18). Electrical resistivity was monitored generally across an entire 64 m long section perpendicular to the highway covering full depth down to the groundwater. From the results and time lapse analyses, it was clear that changes mostly take place in the uppermost part of the road shoulders, which confirms modelling results previously reported by Hansson et al. (2005). In addition, local climate conditions, i.e. precipitation, air temperature and ground temperature, directly or indirectly contribute to resistivity changes. The analyses also revealed that even during periods of frozen ground, moisture movements can take place in the road


42

shoulders (Paper I). Based on these findings and using experiences from resistivity and 2D inverse modelling, more detailed tracer tests at the road shoulders were performed in different seasons (Paper II). In this case the focus was on seasonal changes and flow pathways at a single location. A rectangular electrode pattern of 4×16 electrodes was used, since the spread of the tracer from the road lanes towards the ditch was known to be less than 2 m, based on previous tracer tests (Paper III). Since the objective of this work was to monitor seasonal changes, the tracer was applied uniformly using a 4 m long perforated pipe. The results based on tracer tests and ERT inversions provide a better understanding of the flow pathways in different seasons (Paper II). Frozen ground did not prevent the spread of water-borne pollutants, as was also seen in Paper I. Seasonal flow pathways of melting snow in natural soils have been studied previously by French and Binley (2004). In this thesis the focus was on the spread into the road structure and detailed analyses were needed to understand the flow while considering phase changes. A simulation model was used in order to simulate and understand the liquid water content, frozen water and vapour in different road layers (Paper IV). The simulation results accurately described the water and heat changes during different seasons and the water-ice content when some boundary changes were applied. Previous models have been one-dimensional, not focusing on roads or considering phase changes (e.g. Harlan 1973, Jansson & Halldin 1980, Flerchinger et al. 1996, Šimůnek et al. 1999, Hermansson 2002, Hansson et al. 2004, Karra et al., 2014; Zhang et al., 2016). The models in these studies have often been based on simplifications, since this kind of modelling is quite complicated. The focus of the modelling work in this thesis was mainly on a part of the road and shoulder. The model results were very sensitive to the upper boundary conditions, as previously concluded by Erlingsson et al. (2009). However, no simulations of surface flow and snow melt/accumulation on the surface were considered in the model developed in this thesis, so the model can be improved. The modelling work also showed that Richards’s equation may not be a suitable formula to solve for water movements in road materials. The upper boundary needs to be linked with a surface flow model and snow melt model during different seasons. This study provides some implications for practical use. The first implication is regarding the monitoring stage. An improved monitoring methodology and analyzing process for non-destructive road investigations is presented which can be useful for future road humidity studies. Time series data of pollutant spread can be collected in a more efficient way at critical and vulnerable sites along roads, for example close to drinking water supplies. Transport time of pollutants within the road structure and the percolation to groundwater after road accidents on different types of road, can be measured in a fast and non-destructive way. The flow model developed can be used presumptively at the


43

environmentally critical road sites for predictions of the infiltration and percolation of pollutants during different seasons as well for the risk for frost uphealing wintertime. The second implication for practice is regarding the construction and maintenance of roads. One of the mitigation measures suggested based on the thesis, is to provide a less permeable road shoulder in order to reduce the infiltration of stormwater and reduce the road humidity. This can be done by adding a fine grained material on top of the road shoulder, for example using a thin clay layer. A vegetation cover on the road shoulder and the ditch slope may preserve the clay layer and give a better aesthetic view.

6. CONCLUSIONS

Based on the results obtained in this thesis, the following conclusions were drawn:

Water and pollutant movements mostly take place at the road shoulder of modern roads, even during periods with frozen ground. Dividing the road structure into different zones enables more detailed study of water flow and pollutant spread. The monitoring methodology and analyses provided can be applied in future monitoring programs.

Electrical resistivity changes are directly or indirectly caused by local weather conditions, i.e. precipitation, air temperature and ground temperature. However, there are positive and negative correlations between electrical resistivity tomography (ERT) values and different weather factors, and the resulting correlation coefficients are not high.

Electrical resistivity tomography is a good method for monitoring tracer movements in the road shoulder and could be used for tracing flow pathways in road layers.

Water-borne pollutants reaching the road surface from accidental spills can infiltrate rapidly into the road shoulder and inner ditch of modern road types. It is thus often necessary to take action within 0.5-1 h after a road spill in areas vulnerable to pollution.

During spring or in wet ground situations during other seasons, water spread from the road lane is more uniformly distributed and forms a plume on or within the surface layers. In dry ground conditions, the spread on the ground surface is limited but percolation through the road structure is rapid and occurs along many possible pathways. In frozen ground conditions a pronounced spread of the tracer occurred, followed by preferential infiltration and downward flow, which caused a deep and fast penetration of the tracer.


44

Based on tracer tests results, some chloride is retained in road layers and hence accumulates during the year.

For risk assessments of existing roads, e.g. regarding spills resulting from traffic accidents, information on the hydraulic properties of the road material should be obtained in advance. Geological information from maps is of limited value, since it only shows the surrounding soils.

Tracer experiments using different analytical methods (dynamic flow modelling with 1D and 2D models, analytical calculations, resistivity measurements and simulations) can give comparable results regarding flow patterns and velocities, even in hydraulically heterogeneous environments. All these methods indicate rapid transport processes in the road shoulder and inner ditch slope.

The 2D flow model developed in this thesis, which uses partial differential equations (PDE) in COMSOL to consider phase changes and thermal and hydrological properties of road materials, is an appropriate tool for studying freezing/thawing processes in road structures.

Parameters of the 2D flow model are based on data from an operational road test station. The model performed well in heat simulations. However, due to limitations in the measured data, modelling of moisture changes was only carried out for the top layer of the base material in the road. Prediction of model dynamics performed better than prediction of the average water content in the thin top layer.

The 2D flow model accurately simulated changes in the frozen water content within all road layers and during different scenarios of varying hydraulic conductivity and layer thicknesses. Based on the results, seasonal temperature and humidity changes in the top layer are very important, as are the upper boundary conditions. Frozen layers generally persist for longer under road lanes than beneath the road shoulder.

One of the suggested mitigation measures for environmental protection as well as for sustainable road construction is to provide a less permeable road shoulder in order to reduce infiltration. This can be done by adding fine grained material on the top of the road shoulder, for example using a thin clay layer.

7. RECOMMENDATIONS FOR FUTURE STUDIES

Future studies on water and pollutant spread and flow pathways in roads should:

Determine the hydraulic properties of road materials.


45

Obtain more frequent ERT data, for better correlations with precipitation, moisture content and soil temperature data.

Validate the estimates of chloride concentration by analysing soil samples from different layers.

Install TDR sensors at different depths in the road structure, for better measurements of water content.

Carry out soil temperature measurements at different locations in the shoulder and in the nearby natural soil, since these data are very important for setting the upper boundary conditions.

REFERENCES

Aaltonen J. 2001. Seasonal resistivity variations in some different Swedish soils. European Journal of Environmental and Engineering Geophysics. 6:33-45.

Aaltonen J, Olofsson B. 2002. Direct current (DC) resistivity measurements in long-term groundwater monitoring programmes. Environmental Geology. 41:662-671.

Abdul-Nafiu A. K. Nawawi M.M.N., Abdullah K., Saheed I.K., Abdullah A. 2013. Effects of electrode spacing and inversion techniques on the efficiency of 2D resistivity imaging to delineate subsurface features. American Journal of Applied Science. 10(1):64-72.

Abu-Hassanein Z., Benson C., Blotz L. 1996. Electrical resistivity of compacted clays. Journal of Geotechnical Engineering. 122(5):397-406.

Aljazzar T., Kocher B. 2016. Monitoring of contaminant input into roadside soil from road runoff and airborne deposition. Transport Research Procedia. 14:2714-2723.

Archie G.E. 1942. Electrical resistivity log as an aid in determining some reservoir characteristics. Trans. AIME. 146:54-61.

Arya L.M., Paris J.F. 1981. A physioemperical model to predict the soil moisture characteristics from particle-size distribution and bulk density data. Soil Science Society of America Journal. 45:1023-1030.

Auken E., Pellerin L., Christensen NB. and Sorensen K. 2006. A survey of current trends in near-surface electrical and electromagnetic methods. Geophysics. 71(5), 249-260.

Baldwin G., Addis R., Clark J., Rosevear A. 1997. Use of Industrial By-products in Road Construction – Water Quality Effects. CIRIA report; 167.

Barker R.D. 1979. Signal contribution sections and their use in resistivity studies. Geophysical journal of the royal astronomical society. 59:123-129.

Barker R.D., Moore J. 1998. The application of time-lapse electrical tomography in groundwater studies. The Leading Edge. 17:1454-1458.

Beskow G. 1935. Soil freezing and frost heaving. Stockholm, Geological Survey of Sweden SGU. Report: Årsbok 26.


46

Besson A., Cousin I., Dorigny A., Dabas M., King D. 2008. The temperature correction for the electrical resistivity measurements in undisturbed soil samples: Analysis of the existing conversion models and proposal of a new model. Soil Science. 173(10):707-720.

Blomqvist G., Gustafsson M., Eram M., Ünver K. 2011. Prediction of salt on road surface – Tool to minimize use of salt, Transportation Research Record. Journal of the Transportation Research Board. 2258:131-138.

Brunet P., Clement R., Bouvier C. 2010. Monitoring soil water content and deficit using Electrical Resistivity Tomography (ERT) – A case study in the Cevennes area, France. Journal of Hydrology. 380:156-153.

Bryson L.S., Bathe A. 2009. Determination of selected geotechnical properties of soil using electrical conductivity testing. ASTM Geotechnical Testing Journal. 32(3):252-261.

Campbell 2015. TDR100 Instruction manual. Campbell Scientific Inc., retrieved from https://www.campbellsci.com/tdr100.

Cassiani G., Bruno V., Villa A., Fusi N., Binely A.M. 2006. A saline tracer test monitored via time-lapse surface electrical resistivity tomography. Journal of Applied Geophysics. 59:244-259.

Chambers J.E., Gunn D.A., Wilkinson P.B., Meldrum P.I., Haslam E., Holyoake S., Kirkham M., Kuras O., Merritt A., Wragg J. 2014. 4D electrical resistivity tomography monitoring of soil moisture dynamics in an operational railway embankment. Near Surface Geophysics. 12(1):61-72.

Claerbout J.F., Muir F. 1973. Robust modelling with erratic data. Geophysics. 38:826-844.

COMSOL Multiphysics® 2016, v. 5.2a. The PDE Interface. www.comsol.com. COMSOL AB, Stockholm, Sweden.

Cuong L.P., Tho L.V., Juzakova T., Re´dey. A´., Hai H. 2016. Imaging the movement of toxic pollutants with 2D electrical resistivity tomography (ERT) in the geological environment of the Hoa Khanh Industrial Park, Da Nang, Vietnam. Environmental Earth Science. 75(286):1-14.

Dahlin T. 1996. 2D resistivity surveying for environmental and engineering applications. First Break. 14(7):275-283.

Daily W., Ramirez A., Labrecque D., Nitao J. 1992. Electrical-resistivity tomography of vadose water-movement. Water Resources Research. 28:1429-1442.

Daily W., Ramirez A. 1995. Electrical resistance tomography during in-situ trichloroethylene remediation at the Savannah River Site. Journal of Applied Geophysics 33:239–249.

Dawson A. (ed.) 2009. Water in Road Structures – Movement, Drainage & Effects. Springer: Nottingham. 436 p.

De Vries D.A. 1963. Thermal properties of soil. In: Van Wijk W.R. (ed.), Physics of Planet Environment. (pp. 210-235). North Holland, Amsterdam.


47

Earon R., Olofsson B., Renman G. 2012. Initial effects of a new highway section on soil and groundwater. Water, Air, and Soil pollution. 223:5413-5432.

Edlefsen N.E., Andersen A.B.C. 1943 Thermodynamics of soil moisture. Hilgardia. 15(2):31–298.

Ekblad J., Isacsson U. 2007. Time-domain reflectometry measurements and soil-water characteristic curves of coarse granular materials used in road pavements. Canadian Geotechnical Journal. 44:858-872.

Flerchinger G.N., Hanson C.L. 1989. Modeling soil freezing and thawing on a Rangeland watershed. Soil and water Division. ASAE. 32(5):1551-1560.

Erlingsson S., Brenčič. M., Dawson A. 2009. Water flow theory for saturated and unsaturated pavement material. In: Water in Road Structures – Movement, Drainage & Effects, Dawson A. (ed.) Springer: Nottingham, 23-44.

Freden C. (ed.) 1994. National Atlas of Sweden - Geology. 208 pp. French H. K., Van der Zee S.E.A.T.M. 1999a. Field observations of

small scale spatial variability of snowmelt drainage and infiltration. Nordic Hydrology. 30:161-176.

French H. K., Van der Zee S.E.A.T.M., Leijnse A. 1999b. Differences in gravity dominant unsaturated flow during autumn rains and snowmelt. Hydrological Processes. 13(17):2783-2800.

French H. K., Hardbattle C., Binley A., Winship P., Jakobsen L. 2002. Monitoring snowmelt induced unsaturated flow and transport using electrical resistivity tomography. Journmal of Hydrology. 267:273-284.

French H. K., Binley A. 2004. Snowmelt infiltration: monitoring temporal and spatial variability using time-lapse electrical resistivity. Journmal of Hydrology. 297:174-186.

Ghazavi M., Roustaie M. 2010. The influence of freeze–thaw cycles on the unconfined compressive strength of fiberreinforced clay. Cold Regions Science and Technology. 61(2-3):125-131.

Gontier M., Olofsson B. 2002. Areell sårbarhetsbedömning för grundvattenpåverkan av vägförorening (In Swedish). Dept of Land and Water Resources Engineering, Royal Institute of Technology, TRITA-LWR- Report 3011, Stockholm.

Griffiths D.H., Turnbull J. 1985. A multi-electrode array for resistivity surveying. First Break. 3(7):16-20.

Gunn D.A., Chambers J.E., Uhlimann S., Wilkinson P.B., Meldrum P.I., Dijkstra T.A., Haslam E., Kirkham M., Wragg J., Holyoake S., Hughes P.N., Hen-Jones R., Glendinning S. 2014. Moisture monitoring in clay embankments using electrical resistivity tomography. Construction and Building Material. 95:82-94.

Guymon G.L., Luthin J.N. 1974. A coupled heat and moisture transport model for Arctic soils. Water Resources Research. 10:995– 1001.


48

Hansson K., Šimůnek J., Mizoguchi M., Lundin L.C., van Genuchten M. Th. 2004. Water flow and heat transport in frozen soil – numerical solution and freeze thaw applications. Vadose Zone Journal. 2(3):693-704.

Hansson K., Lundin L.C., Šimůnek J. 2005. Modeling water flow patterns in flexible pavements. Journal of the Transportation Research Board. 1936:133-141.

Harrison R.M., Wilson S.J. 1985. The chemical composition of highway drainage waters. Major ions and selected trace metals. The Science of the Total Environment. 43(1-2):63-77.

Harlan RL. 1973. Analysis of coupled heat-fluid transport in partially frozen soil. Water Resources Research. 9(5):1314–1323.

Hermansson Å. 2002. Modeling of frost heave and surface temperatures in roads. PhD thesis. Luleå University of Technology, Sweden. ISSN: 1402-1544. ISRN: LTU-DT 02/13SE.

Hermansson Å., Charlier R., Collin F., Erlingsson S., Laloui L., Srṧen M. 2009. Heat transfer in soils. In: Water in Road Structures – Movement, Drainage & Effects, Dawson A. (ed.) Springer: Nottingham; 69-79.

Jackson P.D., Northmore K.J., Meldrum P.I. Gunn D.A., Hallam J.R., Wambura J., Wangusi B., Ogutu G. 2002. Non-invasive moisture monitoring within an earth embankment – a precursor to failure. NDT & E International. 35(2):107-115.

Jansson P.E., Halldin S. 1980. SOIL water and heat model: Technical description. Swedish Coniferous Forest Proj. Tech. Rep. 26. Swedish University of Agricultural Sciences, Uppsala, Sweden.

Jansson C., Almkvist E. Jansson P.E. 2006. Heat balance of an asphalt surface: observations and physically-based simulations. Meteorological Applications. 13:203-21.

Jansson P.E., Karlberg L. 2001. Coupled Heat and Mass Transfer Model for Soil–plant–atmosphere Systems. Royal Institute of Technology, Department of Civil and Environmental Engineering, Stockholm. 325 pp.

Kalantari Z. 2014. Road structures under climate and land use change: Bridging the gap between science and application. PhD thesis. TRITA LWR PHD-2014:01, 31 p.

Karra S., Painter S., Lichtner P. 2014. Three-phase numerical model for subsurface hydrology in permafrost-affected regions (PFLOTRAN- ICE v1. 0). Cryosphere. 8(5):1935-1950

Keller G.V., Frischknecht F.C. 1966. Electrical Methods in Geophysical Prospecting. Pergamon Press: Oxford; 517 p.

Kroener E., Vallati A., Bittelli M. 2014. Numerical simulation of coupled heat, liquid water and water vapour in soils for for heat dissipation of underground electrical power cables. Applied Thermal Engineering. 70:510-523.

Kömle N.I., Bing H., Feng W.J., Wawrzaszek R., Hütter E.S., He P., Marczewski W., Dabrowski B., Schröer K., Spohn T. 2007.


49

thermal conductivity measurements of road construction materials in frozen and unfrozen state. Acta Geotechnica. 2(2):127-138.

Leroux V., Dahlin T. 2006. Time-lapse resistivity investigations for imaging saltwater transport in glaciofluvial deposits. Environmental Geology. 49:347-358.

Lindström, R. 2005. Groundwater vulnerability assessment using process-based models. PhD thesis. Royal Institute of Technology, Stockholm, TRITA-LWR PHD 1022.

Lindström R. 2006. A system for modelling groundwater contamination in water supply areas - chloride contamination from road de-icing as an example. Nordic Hydrology. 37(1):41-51.

Loke M.H., Barker R.D. 1996. Rapid least-squares inversion of apparent resistivity pseudosections by a quasi-Newton method. Geophysical Prospecting. 44:131-152.

Loke M.H. 2007. RES3DINV. Rapid 3-D Resistivity and IP inversion using the least-squares method. Manual, Geotomo Software, Malaysia.

Loke M.H., Dahlin T., Rucker D.F. 2014. Smoothness-constrained time-lapse inversion of data from 3D resistivity surveys. Near Surface Geophysics. 12:5-24.

Loke M.H. 2016. Tutorial: 2-D and 3-D electrical imaging surveys. Retrieved from http://www.geotomosoft.com/.

Looms M.C., Jensen K.H., Binley A., Nielsen L. 2008. Monitoring unsaturated flow and transport using cross-borehole geophysical methods. Vadose Zone Journal. 7:227-237.

Lundmark A., Olofsson B. 2007. Cl deposition and distribution in soils along a deiced highway – assessment using different methods of measurement. Water, Air, and Soil Pollution. 182(1-4):215-232.

Lytton R.L., Pufahl D.E., Michalak C.H., Liang H.S. Dempsey B.J. 1993. An Integrated Model of the Climatic Effects in Pavements. Federal Highway Administration, McLean, VA, U.S.A. Rep. FHWA, RD-90-033.

Mao D., Revil A., Hort R.D., Munakata-Marr J., Atekwana E.A., Kulessa B. 2015. Resistivity and self-potential tomography applied to groundwater remediation and contamination plumes: Sandbox and field experiments. Journal of Hydrology. 530:1-14.

Meinzer O.E. 1923. The occurrence of groundwater in the United States, with a discussion of principles. Ecological Water-Supply Paper 489. USA Printing office:Washingto; 321 p.

Minas M. 2010. Monitoring highway runoff using.2-D and 3-D resistivity methods. Master’s Thesis TRITA-LWR-EX-10-24.

Mualem Y. 1976. A new model for predicting the hydraulic conductivity of unsaturated porous media. Water Resources Research 12(3):513–522.

Nickman A. 2016. Road disasters? Modeling and assessment of Swedish roads within crucial climate conditions. PhD thesis. TRITA LWR PhD 2016(06): 43 p.

Ogilvy R.D., Kuras O., Meldrum P.I., Wilkinson P.B., Gisbert J., Jorreto S., Pulido-Bosch A., Kemna A., Nguyen F., Tsourlos P.


50

2007. Automated monitoring of coastal aquifers with electrical resistivity tomography. In: Coastal Aquifers: Challenges and Solutions, Pulido Bosch A., López-Geta J.A., Ramos González G. (eds.). Instituto Geológico y Minero de España, Madrid, 333-342.

Olofsson B., Sandström S. 1998. Increased salinity in private wells in Sweden - natural or man-made? In: Deicing and dustbinding - risk to aquifers (D&D98), Nystén T., Suokko T. (eds.), Nordic Hydrological Programme (NHP). 43:75-81.

Olofsson B., Jernberg H., Rosenqvist A. 2005. Tracing leachates at waste sites using geophysical and geochemical modeling. Environmental Geology. 46(5):720-732.

Olofsson B., Lundmark A. 2009. Monitoring the impact of de-icing salt on roadside soils with time-lapse resistivity measurements. Environmental Geology. 57(1)217-229.

OriginLab 2015. OriginLab Corporation. Software, Northampton, MA.

Philip JR, de Vries DA. 1957. Moisture movement in porous materials under temperature gradients. Transactions - American Geophysical Union. 38:222–231.

Piguet P. 2007. Road runoff over the shoulder difuse infiltration. Real-scale experimentation and optimization. PhD thesis. 3858, Université de Neucha`tel, 277 p.

Reynolds J.M. (ed.). 2011. An Introduction to Applied and Environmental Geophysics. 2nd ED. Wiley-Blackwell. UK.

Rhoades J.D., Raats P.A.C., Prather R.J. 1976. Effects of liquid-phase electrical conductivity, water content and surface conductivity on bulk soil electrical conductivity. Soil Science Society of American Journal. 40(5):651-655.

Riehm M. 2012. Measurements for winter road maintenance. TRITA, LWR PhD, ISSN 1650-8602 ; 1069, 45 pp.

Saito H., Simunek J., Mohanty B.P. 2006. Numerical analysis of coupled water, vapor, and heat transport in the vadose zone. Vadose Zone Journal. 5(2):784-800.

Salour F. 2015. Moisture influence on structural behaviour of pavements: Field and laboratory investigations. PhD thesis. TRITA-TSC-PHD 15-003.

Samouëlian A., Cousin I., Tobbagh A., Bruand A., Richard G. 2005. Electrical resistivity survey in soil science: a review. Soil and Tillage Research. 83:173-193.

Sarady M., Sahlin E.A.U. 2016. The influence of snow cover on ground freeze thaw frequency, intensity, and duration: An experimental study conducted in coastal northern Sweden. Norsk Geografisk Tidsskrift - Norwegian Journal of Geography. 70(2):82-94.

SCCV. 2007. Sweden facing climate change- threats and opportunities. Swedish Commission on Climate Change and Vulnerability. Swedish Government Official Reports. SOU:2007, 159 p.


51

Shevnin V., Mousatov A., Ryjov A., Delgado-Rodriquez O. 2007. Estimation of clay content in soil based on resistivity modelling and laboratory measurements. Geophysical Prospecting. 55:265-275.

Sheshukov A.Y. Nieber J. L. 2011. One-dimensional freezing of nonheaving unsaturated soils: Model formulation and similarity solution. Water Resources Research. 47(W11519):1-17.

Simonsen E., Janoo V. C. Isacsson U. 1997. Prediction of temperature and moisture changes in pavement structures. Journal of Cold Regions Engineering. 11(4):291-307.

Singha K., Gorelick S.M. 2005. Saline tracer visualized with three-dimensional electrical resistivity tomography: Field-scale spatial moment analysis. Water Resources Research. 41(W05023):1-17.

Singha K., Gorelick S.M. 2006. Effect of spatially variable resolution on field-scale estimates of tracer concentration from electrical inversions using Archie´s law. Geophysics. 71(3):G83-G91.

Slater L., Sandberg S. 2000. Resistivity and induced polarization monitoring of salt transport under natural hydraulic gradients. Geophysics. 65:408-420.

Slater L., Binley A. 2003. Evaluation of permeable reactive barrier (PRB) integrity using electrical imaging methods. Geophysics. 68:911-921.

SMHI 2017. Retrieved from http://www.luftwebb.smhi.se/. Swarzenski P.W., Burnett W.C., Greenwood W.J., Herut B., Peterson

R., Dimova N., Shalem Y., Yechieli Y., Weinstein Y. 2006. Combined time-series resistivity and geochemical tracer techniques to examine submarine groundwater discharge at Dor Beach, Israel. Geophysical Research Letters. 33(L24405):1-6.

Šimůnek J., Šejna M. van Genutchen M. Th. 1999. The HYDRUS-2D Software Package for Simulating Two-Dimensional Movement of Water, Heat, and Multiple Solutes in Variably-Saturated Media, Version 2.0, U.S. Salinity Laboratory, USDA, ARS, Riverside, California.

Telford W.M., Geldart L.P., Sheriff R. E., Keys D. A. 1990. Applied Geophysics (2nd Edition), Cambridge University Press. 770 pp.

Test site E18. 2017. Climate and test station data. Retrieved on Jan 31 2017 from www.testsitee18.se

Trafikverket 2005. Kapitel E Obundna material (In Swedish). VV Publ 2005:112 Retrieved from http://www.trafikverket.se/

Trafikverket 2015. From trafikverket annual archive data. Retrieved from http://www.trafikverket.se/

Thunqvist E.L. 2003. Estimating chloride concentration in surface water and groundwater due to deicing salt application. PhD thesis. Royal Institute of Technology (KTH), TRITA-LWR PHD 1006.

Thunqvist E.L. 2004. Regional increase of mean chloride concentration in water due to application of deicing salt. Science of the Total Environment. 325:29-37.

van Genuchten MT. 1980. A closed-form equation for predicting the hydraulic conductivity of unsaturated soils. Soil Science Society of America Journal. 44(5):892-898.


52

Wenner F. 1912a. The four-terminal conductor and the Thompson bridge. US Bureau of Standards Bulletin. 8:559-610.

Wenner F. 1912b. A method of measuring earth resistivity. US Bureau of Standards Bulletin 12:469-478.

Wilhelmsson H. 2017. Tjälgränsmätning. In Swedish, retrieved from https://www.vti.se/sv/Forskningsomraden/Tjalgransmatning/

Wilkinson P. B., Meldrum P. I., Kuras O., Chambers J. E., Holyoake S. J., Ogilvy R. D. 2010. High-resolution Electrical Resistivity Tomography monitoring of a tracer test in a confined aquifer. Journal of Applied Geophysics. 70:268-276.

Xu XZ., Wang JC., Zhang LX. 2010. Frozen Soil Physics. Science Press, Beijing.

Yisa J. 2010. Heavy metal contamination of road deposited sediments. American Journal of Applied Science. 7(9):1231-1236.

Zhang M., Wen Z., Xue K., Chen L., Li D. 2016. A coupled model for liquid water, water vapor and heat transport of saturated-unsaturated soil in cold regions: Model formulation and verification. Environmental Earth Science. 75(701):1-19.

Zhou B., Dahlin T. 2003. Properties and effects of measurement errors on 2D resistivity imaging. Near Surface Geophysics. 1(3):105-117.

Hedi Rasulkth.diva-portal.org/smash/get/diva2:1201373/FULLTEXT01.pdfTRITA-LWR DLT-188 ISBN...

Documents

Transcript of Hedi Rasulkth.diva-portal.org/smash/get/diva2:1201373/FULLTEXT01.pdfTRITA-LWR DLT-188 ISBN...