Hydronumerical Modelling of Nutrients and Oxygen in Ciénaga Grande De Santa Marta, Colombia

Master Thesis

Hydronumerical Modelling of Nutrients and Oxygen in Ciénaga Grande de Santa Marta, Colombia

In fulﬁllment of the requirements for the degree of Master of Science (M.Sc.) in Water Resources and Environmental Management At the faculty of Civil Engineering and Geodetic Sciences of Leibniz Universität Hannover

Marcos Carvajalino Fernández Mat. Nr: 3040980

Date of presentation: 30. September 2015

Supervisor: M.Sc. Oliver Lojek

Examiners: Prof. Dr.-Ing T. Schlurmann

Prof. Dr.-Ing A. Hildebrandt This page intentionally left blank.

II Dedicated to my grandmother, Amalia Isabel Sánchez Santiago. Who was always there for me and taught me that true happiness comes from loving and caring for someone else. You will always be by my side. This page intentionally left blank.

IV DECLARATION

I hereby declare that the topic handed out on September 30th of 2015 as Master Thesis, entitled "Hydronumerical modelling of nutrients and oxygen in Ciénaga Grande de Santa Marta, Colombia", was authored entirely and independently by me, Marcos Antonio Carvajalino Fernández (Mat. Nr: 3040980). This academic work is the outcome of a joint effort between Instituto de Investigaciones Marinas y Costeras "José Benito Vives de Andreis" (INVEMAR), Franzius-Institute for Hydraulic, Estuarine and Coastal Engineering and the author. It was developed under the supervision of professionals from Franzius-Institute and INVEMAR. However, the conclusions presented in this document correspond to my assessment of the facts and do not necessarily reflect the opinion of any of the involved institutions. I also declare that I have written this document without extracting sections from other sources (including electronic media and online sources). All the documents used as references for the present report have been thoroughly cited and identified. This work has not been submitted to any other examination board in this or any other university in order to obtain an academic degree. I am aware that a false declaration could have legal implications. Therefore, I agree to the submission of this document to external services for plagiarism check.

MARCOS CARVAJALINO FERNÁNDEZ Hannover, Germany. 30. Sept. 2015

V This page intentionally left blank.

VI ACKNOWLEDGEMENTS

First and foremost, I would like to thank INVEMAR and Franzius-Institute for placing their trust in me and accepting my research proposal. Betting on this project was a leap of faith since we all knew from start that information was scarce and scattered along many institutions and a considerable effort had to be made in order to develop an integrated tool like the one proposed. Thanks for encouraging young professionals into pursuing their own ideas. I would like to thank Deutscher Akademischer Austausch Dienst (DAAD) for financing my studies in Germany during the past two years. This has been the most nourishing experience of my life both academically and culturally. I am specially indebted to M.Sc. Luis Fernandes (Actionmodulers Inc.) for his constant technical support regarding MOHID modelling system and Eduardo Jauch (Technical University of Lisbon) for his help in compiling MOHID Water in Linux Operating Systems. My academic supervisor M.Sc. Oliver Lojek was always eager to help me with administrative procedures and coping with my ever increasing computational needs, even when it meant lending me his personal desktop PC. Thanks for easing all the thesis process so I could focus on the important aspects of the project. I would also like to express my gratitude to the staff from CAM and GEO programs at INVEMAR, specially Ostin Garcés and Carlos Carbonó for their help during the field campaigns. Thanks to Janet Vivas and Martha Bastidas as the direct contacts from both programs for facilitating the access to INVEMAR datasets. M.Sc. Jorge Corrales was so kind as to lend us his company’s flowmeter when we had problems with ours, without it we could have not finished the data gathering on time. Many people offered me their professional advice during the project, thanks to M.Sc. Jorge Mazenett, Luís Vanegas and Dr. Andrés Osorio Arias (Universidad Nacional de Colombia) for giving me some hints and sharing some of their experience with me. Special thanks to my colleagues Jesús Casado, Raúl Villanueva, Erika Álvarez, Fernanda Scholz and Larisa Tarasova for the company during all this experience and the countless coffee/döner breaks discussing our projects’ issues. Last but not least, I would like to thank my family for their constant support and advice throughout the unique (and sometimes scary) experience of living and studying in a foreign country. You were the true pillar of this achievement and I could have never done it without you.

VII This page intentionally left blank.

VIII ABSTRACT

A numerical model for the simulation of hydrodynamic, conservative substances and nutrient/oxygen transport inside a selected domain of Ciénaga Grande de Santa Marta lagoon complex (CGSM-LC) and its adjacent continental margin was developed in MOHID Water Modelling System as the first large scale water quality modelling initiative from INVEMAR in the Colombian Caribbean. The model performed satisfactorily well in long-term (yearly) simulations of water level fluctuations and salinity patterns inside the estuary area, reporting daily level changes of around 3 cm to 4 cm and velocities of around 1 m s−1 in the seawater inlet of the system and less than 5 cm s−1 in most of the lagoons. Good representation of monthly patterns reported for the system was achieved for wet and transition climatic periods, however the model still needs to be adjusted to adequately simulate the dry weather conditions. Flow patterns in the system shown a marginal tidal influence in the Northeastern part of the system with a tidal excursion of approximately 4 km during flood tides, with the salt wedge reaching the zone of El Boquerón island before the slack tide and water retreat. In the rest of the system, flow patterns are dominated by the freshwater inflows and wind shear stresses, showing microscale phenomena like water reversals and short-life eddies in the North and Southwest areas of the main lagoon. Flow in CGSM main lagoon is very slow, with residence times of more than 7 months and probably even reaching a year, this situation is even worse in Pajarales lagoon where almost no net water flushing can be observed, operating mostly as an evaporative system. These long residence times make both systems specially sensitive to the accumulation of recalcitrant pollutants. Nutrient and Oxygen simulations for 5 d periods in both dry and wet weather conditions were performed, representing daily biochemical cycles in concordance with previous descriptive works in the area, with a cyclic daily pattern of nutrient consumption by phytoplankton during the day and oxygen depletion during the night. One major drawback in the current state of the model is that it is not including the bacterial biomass that is intervening in the nutrient cycles due to lack of field information on this parameter. Without this element, the nitrification schemes are not complete and an Ammonia accumulation trend can be observed in the results. Further calibration and validation procedures, inclusion of the bacteria component, as well as long term simulations to verify that the model is able to replicate seasonal patterns for these non-conservative variables are planned extensions for the near future.

IX This page intentionally left blank.

X CONTENTS

1 Introduction1 1.1 Ciénaga Grande de Santa Marta Lagoon Complex...... 1 1.2 Motivation...... 5 1.3 Objectives...... 7

2 Domain description9 2.1 Institutional framework and previous studies...... 9 2.2 Physical description...... 10 2.2.1 Bathymetry...... 12 2.2.2 Bottom material...... 14 2.3 Environmental conditions...... 15 2.3.1 Atmospheric dynamics...... 15 2.3.2 Hydrology...... 17 2.3.3 Pollution sources...... 18 2.4 Ocean conditions...... 19 2.4.1 Internal waves...... 22

3 Materials and methods 23 3.1 Model domain and software selection...... 23 3.2 MOHID model description...... 25 3.3 Hydrodynamic model...... 27 3.3.1 Grid selection...... 27 3.3.2 Boundary conditions and input data...... 28 3.3.3 Residence time calculation...... 34 3.4 Nutrients and oxygen model...... 34 3.5 Simulations...... 35 3.6 Calibration...... 36

4 Results and discussion 37 4.1 Input data preprocessing...... 37 4.1.1 Bathymetry...... 37

XI 4.1.2 Atmospheric variables...... 39 4.1.3 Hydrological variables...... 41 4.1.4 Tide veriﬁcation...... 43 4.2 Stability and convergence...... 44 4.3 Temporal patterns of hydrodynamic variables...... 46 4.3.1 Water level...... 46 4.3.2 Water velocity...... 47 4.4 Flow patterns in the system...... 48 4.5 Residence times...... 50 4.6 Hydrodynamic calibration...... 51 4.7 Transport of conservative substances...... 53 4.8 Nutrient and oxygen transport...... 55

5 Conclusions and recommendations 59

Bibliography 62

Appendices 69

A Hydrodynamic and constituent transport in estuaries 71 A.1 Equations for hydrodynamics in shallow waters...... 71 A.1.1 Forcing functions...... 72 A.2 Equations for constituent transport...... 74

B Tidal constituent analysis 75

C MOHID WaterQuality module equations 85

XII LIST OF FIGURES

1.1 Location and components of CGSM-LC...... 1 1.2 Primary productivity of CGSM-LC among coastal systems in the world...... 2 1.3 Fishermen in CGSM-LC...... 3 1.4 Timeline of main interventions in CGSM...... 4 1.5 Fish deaths in CGSM-LC...... 6

2.1 Conceptual model of CGSM-LC by Deeb Sossa S en C(1993)...... 11 2.2 Bathymetric map of CGSM...... 13 2.3 Delta lobes in the continental margin of CGSM-LC...... 13 2.4 Sediment distribution in lagoons of CGSM-LC...... 14 2.5 Climatic periods in the study area...... 15 2.6 Movement of ITCZ...... 16 2.7 Ocean wind velocities over the Caribbean Sea...... 17 2.8 Pollution sources in CGSM-LC...... 18 2.9 Anthropic sources of pollution...... 19 2.10 Sea surface levels near CGSM-LC...... 20 2.11 Ocean waves in the Colombian Caribbean...... 20 2.12 Waves in front of barrier island...... 21 2.13 Longshore drift in CGSM-LC...... 21 2.14 Internal wind driven waves...... 22

3.1 Model domain extents and points of special interest in the system...... 24 3.2 MOHID conceptual model...... 26 3.3 Numerical grids...... 28 3.4 Compiled bathymetric information...... 30

4.1 Cross validation of bathymetry, GEBCO vs. COL-1203...... 37 4.2 Bathymetric proﬁle of the continental margin...... 38 4.3 Bathymetric gradient of the continental margin in front of CGSM...... 38 4.4 Bathymetric map of CGSM...... 39 4.5 Walter & Lieth diagram for CGSM-LC...... 40 4.6 Hydrometric measurements at Clarín and Aguas Negras creeks...... 42

XIII 4.7 Tide verification...... 43 4.8 Semidiurnal moving average fit...... 44 4.9 Spin-up period for CGSM-LC...... 45 4.10 Time series of water level in estuarine area...... 46 4.11 Temporal patterns of water level in CGSM-LC...... 47 4.12 Vector velocity map during ebb and flood at La Barra...... 48 4.13 Flow patterns during an Ebb-Flood cycle in CGSM-LC (April)...... 49 4.14 Flow patterns during an Ebb-Flood cycle in CGSM-LC (October)...... 50 4.15 Residence times in CGSM...... 51 4.16 Hydrodynamic calibration using horizontal eddy viscosity...... 52 4.17 Hydrodynamic calibration using bottom rugosity...... 53 4.18 Temporal patterns of salinity in CGSM-LC...... 54 4.19 Comparison of observed and simulated salinity in different stations...... 55 4.20 Nutrient and oxygen simulations for dry season...... 56 4.21 Nutrient and oxygen simulations for wet season...... 57

B.1 Observed tides in 2015, "Santa Marta" station...... 75 B.2 Observed tides in 2014, "Santa Marta" station...... 77 B.3 Observed tides in 2013, "Puerto Velero" station (1)...... 79 B.4 Observed tides in 2013, "Puerto Velero" station (2)...... 80 B.5 Observed tides in 2014, "Puerto Velero" station (1)...... 81 B.6 Observed tides in 2014, "Puerto Velero" station (2)...... 83

XIV LIST OF TABLES

1.1 Species diversity in CGSM-LC...... 3

2.1 Summary of studies in CGSM-LC regarding hydrodynamics...... 10 2.2 Design ﬂows for creeks connecting CGSM-LC and Magdalena River...... 17

3.1 DIMAR stations for tide analysis...... 31 3.2 IDEAM meteorological data for the project...... 32 3.3 IDEAM stations for river ﬂows...... 33 3.4 System conﬁgurations used for simulations...... 36

4.1 Mean monthly atmospheric conditions for CGSM-LC...... 40 4.2 Variables selected for monthly variations...... 41 4.3 Mean monthly river ﬂows into CGSM-LC...... 42

XV This page intentionally left blank.

XVI LIST OF ABBREVIATIONS

CAM Marine water quality program at INVEMAR

CFL Courant-Friedrichs-Lewy stability criteria

CGSM Ciénaga Grande de Santa Marta main lagoon

CGSM-LC Ciénaga Grande de Santa Marta lagoon complex

CORPAMAG Regional environmental agency of Magdalena department

DAAD Deutscher Akademischer Austausch Dienst

DIMAR General Marine Directorate

ENSO El Niño Southern Oscillation

FDM Finite Diﬀerences Method

GEBCO Global Bathymetric Chart of the Oceans

GEO Marine geosciences program at INVEMAR

GTZ German agency for technical cooperation

HAB Harmful Algal Blooms

HRT Hydraulic Retention Time

IBA Important Bird Area

IDEAM National Institute for Hydrology, Meteorology and Environmental Studies

INVEMAR Instituto de Investigaciones Marinas y Costeras "José Benito Vives de Andreis"

IOC Intergovernmental Oceanographic Commission

ITCZ Intertropical Convergence Zone

XVII This page intentionally left blank.

XVIII CHAPTER ONE

INTRODUCTION

1.1 Ciénaga Grande de Santa Marta Lagoon Complex

Located in the Northern part of Colombia, between 10◦440-11◦00 N and 74◦160-74◦380 W, CGSM-LC is the largest coastal lagoon system of the country with an overall extension of approximately 4280 km2 (Ibarra et al., 2015). Out of this area, approximately 730 km2 correspond to permanent and semi-permanent lagoons while the rest of the system comprises extensive mangrove swamps, sand bars, ﬂood plains, and a complex network of interconnection creeks between the main water bodies.

187296,הח׍ 348760,הח׍ 510224,הח׍ 671688,הח׍ 833152,הח׍

LA GUAJIRA ؁ؔ؎؁׳ ® ؁ؔؒ؁׭ BOLÍVAR

SUCRE MAGDALENA ,178631 ,178631 CESAR בב בב

Service Layer Credits: Esri, DeLorme,

,047672 Caribbean Sea ,047672 בב בב

؁،،؉ؕؑ؎؁ؒؒ؁ע ,916713 ,916713 אב אב

؁،،؅ؖ؉׳ ؉ؖ؅ؒײ׀؁؎؅،؁؇؄؁׭ ,785754 ,785754 אב אב

؁؃؁ؔ؁؃؁ؒס 02,5 5 10 15 20 km ؎ړ؉؃؁؄؎ؕצ 187296,הח׍ 348760,הח׍ 510224,הח׍ 671688,הח׍ 833152,הח׍ Fig. 1.1: Location and components map of CGSM-LC

1 The system receives freshwater inputs from the largest river in the country (Magdalena river) via connection creeks along the West margin, and from three additional major streams located along the South and East margins, namely Fundación, Aracataca and Sevilla rivers. A 200 m width seawater inlet (La Barra) exist in the North-Eastern corner of the main lagoon, which provides the only relevant exchange between the system and the Caribbean Sea. CGSM-LC is regarded as a strategic environment in the colombian caribbean region due to its extremely high net primary productivity, which has been reported to reach 990 g m−2 yr−1 of organic Carbon, three to four times the primary productivity reported by Brothers et al.(2013) for German lakes like Schulzensee (372 g m−2 yr−1) or Kleiner Gollinsee (264 g m−2 yr−1) near Berlin. Moreover, this net primary productivity is considered high even among most coastal and estuarine systems worldwide (Fig. 1.2), and defines the lagoon complex as hypertrofic according to Nixon’s classification scale (>500 g m−2 yr−1).

Fig. 1.2: Values of annual primary productivity for coastal and estuarine systems around the world, with CGSM-LC highlighted. GPP: gross primary productivity, NPP: net primary productivity, U: undeﬁned. Adapted from Cloern et al.(2014)

This high primary productivity translates into approximately 360 000 t/yr of available organic Carbon for secondary consumers (Deeb Sossa S en C, 1993), promoting a high biodiversity inside the system. Table 1.1 reports animal diversity inside the complex, from it, it can be evidenced that CGSM-LC shows a high diversity of fish, mollusks and birds (both permanent residents and migratory) which makes it a hotspot for commercial fisheries and conservation efforts.

2 Table 1.1: Species diversity and representative organisms in CGSM-LC. Summarized from Polanía et al.(2001) and Aguilera(2011)

Category Species Representative organisms/families Phytoplankton 300 Pennate, centric diatoms, dinoflagellates Invertebrates 102 Snails (Melampus coffeus), crabs (Uca rapax, Uca vocator), oysters (Crassotrea rhizophorae) Bony fish 144 Anchovies (Engraulididae), mullets (Mugilidae), mojarra (Gerreidae) Birds 199 Shorebirds (Limicolae), pelicans (Pelecanus occidentalis), herons (Ardeidae) Reptiles 26 Crocodiles, caiman, tortoises Mammals 19 Crab-eating racoon, prehensil-tail porcupine, ocelot, ring-tail monkey

According to Deeb Sossa S en C(1993), fisheries from CGSM-LC account for 60 % of the total fish captures in Magdalena department, all of them via traditional fishing arts (Fig. 1.3). By 2014 fish captures in the complex (including mollusks and crustaceans) yielded approximately 450 t/month, having ranged historically between 300 t/month to 1100 t/month (Ibarra et al., 2015). This figures include only the fishermen reporting to the official monitoring program and therefore could be much higher in reality. Fisheries in the system provide direct livelihood to approximately 40 000 inhabitant in the area (Zamora and Meza, 2013).

Fig. 1.3: Traditional ﬁsheries in La Barra sector, where a high abundance of both estuarine and sea ﬁsh species can be found.

Lands near the East and South-West margins of the main lagoon are highly fertile due to the deposition of nutrient rich sediments during ﬂood events in the system. Large banana plantations, African oil palm industries and subsistence crops, as well as cattle breeding areas, boost the local economy, but at the same time impose a high ecological stress onto the delicately balanced lagoon system due to nutrient enrichment and diversion of the freshwater bodies originally discharging in the main lagoons. Approximately 341 500 people were living inside the ecological region of CGSM-LC by 2005, most of them in nearby towns or dwelling directly inside the estuary in stilt villages. Around 43.5 % to 67.3 % of the population showed some level of unsatisﬁed basic needs (housing, sanitation, source

3 of income, access to education), among which drinking water coverage ranged from 50.8 % to 86 % (mostly untreated surface water) and only 10 % to 50 % had access to some kind of wastewater treatment facilities. Most household wastes and discharges end up entering the lagoon complex untreated, continuously polluting the main source of food and income for the population (Vilardy and González, 2011). The influence of armed groups involved in the Colombian internal conflict can be evidenced mostly in the South-West and East sectors of the complex, other minor criminal bands dwell in the cities that surround the largest water bodies. Even though the inhabitants of the complex are always eager to help the institutions that undergo conservation and environmental monitoring in the area, it is still very difficult to deploy automatic data loggers or long data gathering campaigns due to the safety and economical risks involved. CGSM-LC is considered to be an ecosystem undergoing recovery from several interventions during the last century, when it was heavily impacted by large infrastructure projects. Fig. 1.4 presents a timeline of the main interventions and milestones in the complex since 1950.

RN90-T07 RN27-T01 PRO-CIÉNAGACreek reopening 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015 Fig. 1.4: Timeline of main interventions in CGSM

The main interventions started in 1956 with the construction of a section of National Highway 90 (RN90-T07) along the coastline North of the complex, linking the capital cities of Santa Marta and Barranquilla but blocking multiple inlets that connected the system to the sea at the same time, leaving the actual inlet known as "La Barra" as the sole point for freshwater/seawater interchange in the complex. Between 1960 and 1970 a second major road (RN27-T01) was constructed on the Western part of the complex, along the right bank of Magdalena river. In order to preserve this road and provide ﬂood protection for the towns settled along the course of Magdalena river, a series of dikes has been constructed creating a continuous barrier that nowadays extends for more than 50 km (Deeb Sossa S en C, 1993). These dikes blocked several freshwater inlets and disrupted the seasonal 7-year ﬂooding pattern from the river into the complex, imposing even higher hydrological stress onto the system. The aforementioned changes to the water balance in the system generated an acute hypersalinization of the soils (more than 100 in interstitial water) that caused the loss of approximately 28 570 ha of mangrove forest betweenh 1956 and 1995 (Ibarra et al., 2015). In order to evaluate and implement corrective measures that avoided further destruction of the mangrove forest, a joint venture between the national government, the Regional environmental agency of Magdalena department (CORPAMAG) and the German agency for technical cooperation (GTZ) 1, called "PRO-CIÉNAGA", was formed. This joint venture concluded, after some detailed

1Now Deutsche Gesellschaft für Internationale Zusammenarbeit GmbH (GIZ)

4 studies, that the system should be reconnected to Magdalena river in order to balance the salt acummulation and therefore the redesign and maintenance of three creeks was performed. As to record the progress of the system after the creek reopening, INVEMAR has been implementing a monitoring program for water quality, mangrove recovery and ﬁshing resources for more than 15 years. To date, this program is the largest available source of information regarding constituent dynamics in the complex. Some of the ongoing threats for the lagoon complex, according to Botero and Mancera(1996) and Aguilera(2011), are:

• Intense resource exploitation due to poor living conditions of the inhabitants, including over- ﬁshing and logging activities in magrove swamps.

• Anthropogenic changes in the water balance, mainly:

– Interruption of the water exchange between the complex and the sea by highways construction.

– The uncontrolled construction of dikes, ﬂood protection and diversion structures in the right margin of Magdalena river, disrupting the seasonal ﬂooding dynamics and freshwater inputs into the lagoon system.

– The lack of maintenance of the artiﬁcial and natural creeks bringing freshwater into the system, allowing them to clog with ﬂoating vegetation and detritus from the connected water bodies.

• Hypersalinization of soils near the estuarine system due to reduced freshwater inputs, long retention times and very high hydric deﬁcit. Consequent loss of native mangrove forest and associated fauna.

• Water pollution due to lack of adequate sanitation in the nearby cities and inner stilt villages, nutrient enrichment from increased agriculture in the surrounding lands.

• Lack of information regarding system dynamics due to safety risks for developing intensive monitoring initiatives, undermining the ability of informed decision making in the area.

For a more detailed description of the environmental variables in the study area and their dynamics please refer to chapter 2.

1.2 Motivation

As it was mentioned in the previous section, the lack of continuous monitoring and information regarding the dynamics of CGSM-LC constitute one of the main threats for the correct equilibrium of the system, while it aﬀects the decisions that stakeholders in the area can take regarding infrastructure, land titles and water rights, which are at the same time causes for most of the other threats

5 to the complex. Therefore, developing decision support tools that help overcome this information gap should be one of the main focuses of research and development in the area. According to Neves(2007), numerical models provide enough insight to make major decisions in natural resources related problems. The use of numerical models as decision support tools for environmental management has been rapidly increasing since 1970, mainly due to the escalation of computer power in the past decades. Nowadays it is possible to simulate complex systems and compare multiple development scenarios in short time with the use of personal computers or simple cluster arrays, the use of parallelism techniques can furthermore reduce simulation times. By the aforementioned, developing a numerical model that adequately represents the hydrodynamics in CGSM-LC would provide institutions involved in the management of the complex with a tool to evaluate the response of physical variables like water levels, bottom shear stresses (consequently erosion/deposition dynamics), velocity ﬁelds and salinity against external phenomena or development scenarios under analysis. By coupling constituent transport models to the proposed hydrodynamics model, extensive studies regarding environmental impact of intervention policies in the ecological region, as well certain ecological phenomena whose causes still remain uncertain to this date could be made. One of such ecological phenomena are the sudden and apparently random massive ﬁsh deaths that have been registered in the complex in the past decades (Fig. 1.5).

Fig. 1.5: Fish deaths episodes in CGSM-LC. Left: October 2014, right: June 2015

Since 1971, massive fish deaths have been reported in the system by the fishermen and local press agencies, sometimes accounting for more than 20 tonnes of death specimens per event. Environ- mental control agencies usually visit the area of the event and gather point samples, most of the time declaring that reduced oxygen levels due to eutrophication or toxins from Harmful Algal Blooms (HAB) are the causes of the incident (Mancera and Vidal, 1994). As fisheries are by far the main source of livelihood for the population in the complex, a planning approach against fish deaths episodes would be desirable rather than the current reactive one. In order to achieve such approach, this project focused on the development of a coupled hydrodynamic

6 and nutrient/oxygen transport model that allows stakeholders to identify potential risk areas for fish death episodes and analyze the main drivers of this phenomenon in order to take future actions to control it. By the end of the project, the model will be handed out to INVEMAR, as the partner institution in Colombia in charge of the marine water quality monitoring in the system, for further improvement and operationalization. In order to achieve the project aim, the following research questions were addressed: Which are the processes that drive the hydrodynamics and nutrient transport/oxygen depletion inside CGSM-LC, how are the seasonal flow patterns inside the lagoon complex, which areas present a higher risk of presenting fish death episodes according to nutrient enrichment and oxygen concentrations?

1.3 Objectives

The main goal of the project is to develop a numerical tool to simulate nutrients and oxygen concentrations inside Ciénaga Grande de Santa Marta lagoon complex. In order to achieve this goal, the following speciﬁc objectives were deﬁned:

• Gather and analyze the existing information regarding environmental conditions and forcing functions in the system.

• Develop a hydrodynamic model for the selected domain

• Design a nutrient and oxygen transport model

7 This page intentionally left blank.

8 CHAPTER TWO

DOMAIN DESCRIPTION

„There is no way to simulate in suﬃcient detail the ecosystem behavior without an in-depth treatment of the full cycle of inorganic matter“ Neves(2007)

2.1 Institutional framework and previous studies

Several public and private institutions are involved in managing the resources and monitoring the environmental conditions inside CGSM-LC: Meteorological and hydrology related variables are measured by the National Institute for Hydrology, Meteorology and Environmental Studies (IDEAM), oceanographic variables (sea level, waves, winds) are monitored by a the Ministry of Defense via the General Marine Directorate (DIMAR), continuous water quality and ecological monitoring is done by INVEMAR. Some of this information is handed out to the public institutions in charge of decision making in the area, namely CORPAMAG and the national parks management unit. In spite of this cooperation scheme, there are still issues regarding competences and coordination between the involved agencies, and therefore a centralized decision making process with a holistic view of the system is not yet implemented. Even though CGSM-LC has been regarded as a strategic ecosystem by Colombia, with the declaration of a national park and a flora and fauna sanctuary, and as RAMSAR site, IBA and UNESCO biosphere reserve by the international community, the amount of studies regarding hydrodynamics is surprisingly low, specially by taking into account the complexity and high variability of an estuarine system of this size. Most official baseline studies were done around 1960 and 1991 when the largest highways were constructed and the biggest impacts to the system due to hypersalinization became evident respectively. Subsequent specific projects in recent decades have been grounded on the assumption that the conditions portrayed by these baseline studies remain valid. Most of the information regarding hydrology and hydrodynamics in the lagoon complex stems from national and foreign university research, with the Faculty of Civil Engineering of Universidad de los Andes in Bogotá being the largest contributor regarding numerical modelling of the system. Specific studies in hydrodynamic and salinity transport (Camacho, 1991), hydraulic effect of creek reopening (Toro and Gómez, 1997), non conservative substances (Lozano, 2003) and modern numerical

9 schemes (Flórez, 2013), are among the most important contributions. A summary of the the main studies done on the complex related to hydrodynamics or physical parameters that inﬂuence water movement is provided in Table 2.1.

Table 2.1: Detailed summary of existing studies in CGSM-LC regarding hydrodynamics

Id. Year Author Description 1 1968–1972 Laboratoire Central D’Hydraulique de Several hydraulic studies, velocity distribu- France tions and flow patterns 2 1978 Universidad de los Andes (CETIH) Hydric and salinity diagnosis 3 1989 Estudios y Asesorias Numerical experiments for calculating water balance in CGSM and Pajarales lagoon 4 1989–1991 Universidad de los Andes / Camacho, L. Hydrodynamic study of CGSM with 2D FDM 5 1997 Deeb Sossa S en C / PRO-CIENAGA Hydraulic rehabilitation and creek reopening plan 6 1997 Toro, F & Gómez, E 2D simulation of the effect of reopening Clarín creek using efficient element method 7 2002 CIOH-CORPAMAG-INVEMAR Bathymetric survey in CGSM, Pajarales and coastal area (dx ≈ 1km) - COL1203 8 2003 Lozano, J. Coupling of a solute transport model, DBO and Nutrients 9 2003 Tuchkovenko, Y. 2D model of oxygen and eutrophication 10 2009 INVEMAR Bathymetric and sediment study 11 2013 Flórez, J. Modelling by Lattice-Boltzmann methods 12 2014 Hylin, A. Water budget in CGSM-LC 13 2015 INVEMAR Bathymetric survey inside CGSM (dx=100m) with detailed survey in Boca de la Barra (dx=50m)

1 2 3 4 5 6 7 8,9 10 11 12 13 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015

By analyzing the modelling reports cited in Table 2.1, it could be observed that most of them refer to the same set of water levels and current velocities measured by Camacho(1991) in 1989. Moreover, all simulations have been made for short periods (days) and do not reﬂect the seasonal dynamics of CGSM-LC, a topic of massive importance for water resources and ecological management of the system.

2.2 Physical description

CGSM-LC is composed of approximately 16 main water bodies (lagoons/swamps), several interconnection creeks, freshwater inputs and a sole seawater/estuarine water interchange inlet. The system has two main permanent water bodies connected by a large navigable channel known as Caño Grande, a brief description of these lagoons is presented below:

10 • Ciénaga Grande de Santa Marta main lagoon (CGSM): By far the largest lagoon in the complex (≈450 km2), it is shaped as a triangle oriented NE→SW with a length of approximately 26 km and width in its mid section of around 24 km. Mean depth in this lagoon is around 1.5 m, therefore steady state water volumes would approximate 675 hm3.

• Ciénaga de Pajarales or simply Pajarales: The second largest lagoon in the system (≈120 km2), it is located west from CGSM and houses the two largest stilt villages in the complex, namely Nueva Venecia and Buenavista. Water depths in Pajarales are shallower than CGSM, with mean depth of around 1 m, therefore the lagoon would contain around 120 hm3 of water.

The rest of the system comprises many permanent and semipermanent small lagoons and eventual ﬂooding areas. Responding to the season, water deﬁcit and large scale climatic phenomena like El Niño Southern Oscillation (ENSO), these secondary water bodies can merge into single lagoons or disappear completely for irregular periods of time. Therefore, the morphology of the system is changing constantly. The only complete conceptual model of the system was developed by Deeb Sossa S en C(1993) and due to its importance is reproduced in Fig. 2.1.

Fig. 2.1: Conceptual model of CGSM-LC (adapted from Deeb Sossa S en C(1993)). The green area corresponds to agricultural lands surrounding the mean lagoon, water ﬂuxes in hm3 yr−1

The barrier island (Isla de Salamanca) is located in the Northern part of the system, this 500 m width sandy bar separates the estuary from the Caribbean sea and hosts some of the largest ﬁshermen towns in the system. The zone is highly inﬂuenced by winds, with the formation of dunes and erosive patterns at the beach due to reversals in longshore drift during certain parts of the year. The

11 vegetation is composed mostly by grasslands and individual mangrove patches that become more abundant to the West. On the East and South margins of the complex a large area of agricultural lands locally known as Zona Bananera can be found, this area extends in a belt of around 10 km to 25 km surrounding the initially saline soils on the margins of the main lagoon (Deeb Sossa S en C, 1993). This area is irrigated by a network of rivers that belong to the catchment in the Western face of the massif Sierra Nevada de Santa Marta, out of which Fundación, Aracataca and Sevilla rivers are the larger streams discharging in the estuary. The zone East from the complex is the most urbanized one, with some major cities and small towns in the path of the main road connecting the Caribbean coast and inner regions of the country. Only some of these cities have wastewater treatment facilities, and those in operation rarely include more than primary treatment trains, therefore high loads of organic matter and nutrients are still being discharged into the streams or directly to the estuary. Moreover, due to the aforementioned agricultural activities in the area, the streams discharging to the estuary are also influenced by diffuse nutrient loads and effluents from extraction plants, specially African palm oil industries. The West margin of the system hosts most of the small lagoons included in the complex, this area is particularly suitable for mangrove development due to the semi permanent nature of these water bodies and deposits of detritus from older paths of the Magdalena river, therefore, the largest and most historic mangrove clusters can be found here. However, this was also the heaviest impacted area during the hypersalinization event and therefore the vegetation is still undergoing a recovery process nowadays.

2.2.1 Bathymetry

As for many sedimentary coastal environments in the tropics, water bodies in CGSM-LC are extremely shallow (Fig. 2.2), with reported bottom depths ranging between 0.5 m to 2.1 m for CGSM and 0.2 m to 1.3 m for Pajarales lagoon (Bernal and Betancur, 1996; CIOH et al., 2001; Posada et al., 2009). Deep waters have been reported in the largest connection of the system (Caño Grande: 6 m) and in the seawater inlet at La Barra (7 m), where higher velocities due to flow area constriction generate higher shear stresses over the soft bed materials. Bed morphology of CGSM shows a general deepening in the direction E→W, specially in the northernmost part of the lagoon with depths of around 0.8 m in the East and around 2 m near the West margin. This deepening responds to the deposition dynamics near the mouths of the rivers that feed the system from the arable lands in the direction of the seawater inlet. Seafloor in front of the complex can be divided in two main areas, defined as sedimentary delta lobes by von Erffa(1973): To the East (in front of CGSM) a 14 km to 15 km wide platform with soft bottom slope (vonErffa’s lobe I) and to the West (disappearing near the mouth of Magdalena river) a platform of (8 km to 10 km) with traces of a rocky core from coral origin (Vilardy and González, 2011). These areas are divided by a steep slope narrow ridge that projects to the coast with depths of 50 m just 5 km from the shoreline, known geographically as the Salamanca Canyon (Fig. 2.3). The lobe I has been reported to contain a couple of reef risings between isobaths 20 to 30, that

12 Fig. 2.2: Most up-to-date published bathymetric map of CGSM, depths in meters. Taken from Posada et al.(2009)

Fig. 2.3: Seafloor configuration in front of CGSM-LC and deltaic lobes according to von Erffa(1973)

13 serve as hotspots for benthic, coral and algae diversity and therefore as a high relevance zone for oﬀ coast ﬁsheries related to the system. A complete description of the species associated to these risings can be found in Bula-Meyer and Díaz-Pulido(1995).

2.2.2 Bottom material

Recent descriptions of bed materials for the main lagoons in the system were done by Bernal and Betancur(1996) and Posada et al.(2009), concluding that mixtures of clayey and silty sediments of around 2 m to 4 m over hardened peat deposits are dominant in most of the water bodies, with presence of sandy deposits in the shore of the barrier island North of the main lagoon and the mouths of the rivers draining the agricultural lands on the East (Fig. 2.4). The coarsest bottom materials inside CGSM are located along the inner margin of the barrier island, mainly due to tidal influence from the seawater input to the East where fine and medium sands are common, and degrading to silty sands in the West bank where lower to no tidal influence can be appreciated. Sedimentation rates between 10 cm yr−1 to 13 cm yr−1 have been reported for CGSM (Vilardy et al., 2011).

Fig. 2.4: Sediment distributions in the lagoons of CGSM-LC according to Bernal and Betancur (1996). Left: CGSM, right: Pajarales lagoon. Modiﬁed from the original.

Pajarales lagoon shows a more complex distribution with coarser material in the inputs of the connection creeks and the mouth of Caño Grande and clay-silt deposits in the West surrounding the stilt village of Nueva Venecia.

14 Organic deposits are common in CGSM, mostly shell conglomerates in the central part and highly degraded plant debris in most of the creeks and outer margins. Although Pajarales also shows some shell deposits, these are a lot smaller and disperse than the main lagoon. Organic matter content for the sediments was reported by Bernal and Betancur(1996) to range from 0.48 % to 14.68 % in weight with the highest values near the river mouths, the bathymetric depression NW of the main lagoon and in the middle of Pajarales lagoon. Comparing these results with the ones from Betancourt et al.(2013), who obtained organic matter values from 17.5 % to 35 % for the same points, there seems to be an impact on this parameter after the hydraulic rehabilitation projects. The coastal margin of the system comprises mostly litoclastic muds with less than 15 % of calcium carbonate and coarser grain material in the wave breaking zone, detailed maps of the surface layer composition and stratigraphic depositional patterns can be found in Posada and Henao(2008) and Posada et al.(2009) respectively.

2.3 Environmental conditions

2.3.1 Atmospheric dynamics Weather in the region can be categorized as tropical-arid, showing one dry (December-April) and one rainy (September-November) period per year, in-between, a smaller rainy period from May-June and a sharp dry period around July, known locally as "Veranillo de San Juan", can occur (Fig. 2.5). Due to its location near the Equator, in the zone of highest global radiation levels, the study area is not inﬂuenced by climatic seasons (i.e. spring/winter cycle).

Fig. 2.5: Climatic periods in the study area. During the transition period the short rainy season is depicted in blue while "Veranillo de San Juan" is in red, August (in white) can be a dry or a wet month depending on major climatic phenomena in the area.

As the region is located inside the movement path of the Intertropical Convergence Zone (ITCZ) from South to North during the year’s progression (Fig. 2.6), heavy rain is concentrated in the third and fourth trimesters of the year, when the convective belt is located in its Northernmost position hovering over the Caribbean Sea. The rest of the year precipitation responds to local phenomena like low pressure or frontal systems. The region is heavily aﬀected by large scale and mid-level tropical climatic phenomena like ENSO and cold atmospheric fronts, these phenomena can increase the threats to human life and infrastructure,

15 Fig. 2.6: Movement of the ITCZ during a year. The convective belt hovers the Caribbean around the end of the third and start of the fourth trimester. Taken from Ortiz et al.(2013)

impose severe rain and drought periods as well as causing local storm surges (Blanco et al., 2006; Ortiz et al., 2013). The eﬀect of hurricanes and tropical storms on the area has been almost null, with only tropical storm Bret (1993) and hurricane Joan (1988) having some impact over the the region (Ortiz, 2007). The area is categorized as a medium to low vulnerability zone for hurricanes (Ortiz, 2012). Most storm surges in the area have been caused by frontal systems (cold fronts) (Ortiz et al., 2013)

Atmospheric pressure in the area registers some of the lowest values in the Caribbean region (≈1009 bar) with a mild peak between June and July (≈1013 bar) (Posada and Henao, 2008). Mean air temperature throughout the year ranges from 27 ◦C to 28.5 ◦C and relative humidity from 76 % to 84 %.

Inland winds in the area have a mean velocity around 3 m s−1 to 4.5 m s−1 while ocean winds can show peaks during January and February of 8.8 m s−1 to 11 m s−1, with the lowest velocities registering during the wet season (Fig. 2.7). Wind direction follows a marked seasonal pattern due to the Northeasterly trade winds from the Azores-Bermuda semi permanent high pressure system and the movement of the ITCZ, being NE and N the prevailing directions most of the year, but ﬂuctuating between E, S and SW during the period of September-November, due to the eﬀect of weak equatorial winds coming from the South. During this period, the lowest velocities are also registered (Ruíz- Ochoa and Bernal, 2009; Thomas et al., 2012).

Extremely high instantaneous wind velocities of 111 km h−1 to 130 km h−1 can be registered during the hurricane season in the Caribbean sea (June-November). However, as aforementioned, the Colombian Caribbean coast is just mid to lowly susceptible to hurricane impacts and local jets from cold fronts are much more important for the area (Ortiz et al., 2013).

16 Fig. 2.7: Ocean wind velocities and directions over the Caribbean Sea at 10 m, we can observe the marked eﬀect of the Northwesterly trade winds with higher intensity at the start of the year.Left: January, right: October. Taken from Posada and Henao(2008)

2.3.2 Hydrology Historically, CGSM-LC was an estuary created by the mouth of Magdalena River entering the Caribbean Sea, this river is the one with longest path and largest flow in Colombia (MQ: 7100 m3 s−1). Around 6000 to 3500 years ago the river mouth started migrating to the North West until reaching its actual location in the area known as Bocas de Ceniza, leaving the CGSM as a depositional low- land subjected to the effects of tides, river inputs and periodical floods from the new path of the river West from the system (Posada et al., 2009). Freshwater inputs from the Magdalena are still being received via gate regulated creeks in the North West of the main lagoon (Caño Clarín) and West from Pajarales (Caño Aguas Negras). Another creek (Renegado) connects in the South West of the complex with some of the secondary lagoons of the system. According to the original design by Deeb Sossa S en C(1993) the three creeks were designed to input a maximum of 140 m3 s−1 to the system (Table 2.2), but it has been assessed that the lack of maintenance and uncontrolled vegetation growth has reduced their capacity drastically and they have been only sporadically operative (Hylin, 2014).

Table 2.2: Design ﬂows for creeks connecting CGSM-LC and Magdalena River according to the original design by Deeb Sossa S en C(1993). These conditions where calculated for a Mean Flow of 9300 m3 s−1 in Magdalena River (November-December conditions).

Creek Connection Design ﬂow [m3 s−1] Clarín Northwest of CGSM 20 Aguas Negras Southwest and Northwest of Pajarales 60 Renegado-Condazo Ciénaga La Aguja/Tamacá 60

In normal conditions, CGSM-LC shows a hydric deﬁcit of about 1150 mm yr−1 in its Northern margin due to the very high evaporation in the area compared to the scarce precipitation, the same pattern repeats in the Southern margin, however, higher precipitation in this area reduces the hydric deﬁcit to levels around 210 mm yr−1 (Blanco et al., 2006). This hydric stress condition is caused by the

17 area’s high surface radiation throughout the year and its location leeward from the tallest coastal mountain range of the world (Sierra Nevada de Santa Marta: 5775 m) which receives most of the rain contained in the clouds as orographic precipitation. Flow patterns in the lagoon have been described to follow a counterclockwise pattern driven mostly by the rates of river discharge and inﬂuence of the tidal action restricted to the area nearby the tidal inlet in La Barra (Wiedemann, 1973). These patterns are consequent with the inner bottom sediment patterns described in subsection 2.2.1. Due to the high amount of clay in the system, groundwater ﬂows to or out from the system can be neglected.

2.3.3 Pollution sources Main pollution sources to the system have been described in detail by Vivas-Aguas et al.(2013) and are summarized in Fig. 2.8. Pollution sources are highly related to economical activities of the local population, however inﬂuence of other economical sectors and industries as mining, tanneries or chemical products is received due to the connection with Magdalena river, which runs along some of the largest productive regions of the country. Therefore, it is not uncommon to ﬁnd complex pollutants like heavy metals in the system’s waters and sediments (Parra and Espinosa, 2007; Espinosa et al., 2011).

Pollution Sources

Industry

Discharge

Dead mangrove

Fisheries

Palm oil industry

Permanent crops Transitory crops Dwellings and stilt village

Urban areas

Water bodies

Rivers Department main roads Department sec. roads Unpaved roads

Fig. 2.8: Pollution sources in CGSM-LC. Modiﬁed from Vivas-Aguas et al.(2013)

The use of fertilizers and pest control agents during agricultural activities, as well as associated processing industries, slaughterhouses nearby the lagoon complex and municipal wastewater discharges are the main nutrient and organic matter inputs to the aﬄuent streams into the system, driving the nitrogen, phosphorus and oxygen biogeochemical cycles. Pollutant concentrations in the tributaries

18 are highly dependent on the climatic conditions, with higher loads during the start of the rainy season due to surface wash-off and subsequent load attenuation due to the higher river flows. The highest inorganic nutrients and suspended solids input to the system has been reported to be Clarín creek in the Northwest of the main lagoon due to the high nutrient pollution of Magdalena river waters draining into it (Vivas-Aguas et al., 2013). Despite the importance of diffuse pollution sources into eutrophization sensitive systems like CGSM- LC, most diagnosis of the complex have approached this component qualitatively. Consequently, not enough information is available nowadays to quantify the specific impact of this kind of pollution into the area of study. Human dwellings generate one of the most important aesthetic and microbiological impacts in the area while most of the population in the shoreline towns and stilt villages don’t or very seldom have access to any kind of sanitation services (Fig. 2.9). Untreated discharges increase coliform bacteria in the neighborhood of the stilt villages way over the national limits for bathing waters and impose risks health risks on the inhabitants, being diarrea and gastroenteritis the most common illnesses reported by the region’s health care systems (Vilardy and González, 2011; Vivas-Aguas et al., 2013).

Fig. 2.9: Anthropic sources of pollution inside CGSM-LC. Left: Accumulation of solid wastes in the Northern shore near Tasajera, right: Stilt villages not connected to waste collection or wastewater treatment services (Nueva Venecia).

As it was mentioned in previous sections, INVEMAR has been monitoring water quality conditions of CGSM-LC for the past 15 years via the "Monitoreo de las condiciones ambientales y los cambios estructurales y funcionales de las comunidades vegetales y recursos pesqueros durante la rehabilitación de la Ciénaga Grande de Santa Marta" project and the REDCAM national monitoring program (Ibarra et al., 2015).

2.4 Ocean conditions

Tidal amplitude in the Colombian Caribbean has been reported to fluctuate between 10 cm to 30 cm, classifying the system domain as a microtidal estuary following a mixed semidiurnal pattern (Fig. 2.10). The influence of tides has been reported to affect only the inlet at La Barra with maximum surface velocities of around 1 m s−1 (Posada et al., 2009).

19 Sea level rise tendencies for the period between 1950 - 2000 has been reported to be approximately 5 mm yr−1, meaning that the lowlands and mangrove swamps that are characteristic of this region and the main cause of its ecosystem richness could be aﬀected in the upcoming 100 years if this trend keeps constant (Rangel-Buitrago et al., 2015).

Tidal Curve - DIMAR/IOC Santa Marta (11.235156,-74.221572) 50 40 30 20 10 0 -10 Relative Height [cm] -20 -30 Jan 02 Jan 06 Jan 10 Jan 14 Jan 18 Jan 22 Jan 26 Jan 30 Date 2014 Fig. 2.10: Example of free sea surface levels near CGSM-LC for January/2014. Minute data taken from IOC operational monitoring service based on DIMAR gauging stations.

Thomas et al.(2012) reported average signiﬁcant wave heights for a 2◦ x 2◦ cell in the Caribbean sea containing the study area, with values around 1 m to 2.5 m, periods from 5.5 s to 7.5 s and directions following the wind patterns previously described, i.e. mostly NE the whole year with E, S and SW during rainy season (Fig. 2.11). The largest wave heights are registered during December- March and June-July, corresponding with the dry periods, while the lowest are registered between September-October during the rainy season.

Fig. 2.11: Signiﬁcant wave height and direction in the Colombian Caribbean basin for January (left) and October (right) from ICOADS 1◦ x 1◦ data from 1992 to 2008. Taken from Thomas et al. (2012)

Regarding ocean currents, the zone is dominated by the Caribbean current that moves E→W most

20 of the year with an inﬂuence of the Colombia counter current (W→E) during the rainy season when the winds get weaker (Posada and Henao, 2008). Martínez and Molina reported that along the coastline of the barrier island the waves impact almost frontally, with heights of 0.6 m to 1 m and periods of around 6 s (Fig. 2.12), while from La Barra inlet all the way East until the city of Santa Marta, heights decrease to a range of 0.15 m to 0.4 m with a period of approximately 8 s (as cited in Posada et al., 2009).

Fig. 2.12: Waves in front of the barrier island at CGSM-LC

Longshore drift follows the typical E→W direction observed in the Colombian caribbean coastline all the way from the system’s eastern limit until approximately the middle point of the barrier island (El Torno sector). However, from this point heading West until the mouth of Magdalena river, shifts in the wave direction tend to generate a W→E littoral drift pattern that creates a convergence point around the start of vonErﬀa’s Lobe 1 (Fig. 2.13). This phenomena is much more frequent during the wet period with the shift on wind directions.

Fig. 2.13: Longshore drift in CGSM-LC, the convergence point of the two drift patterns is located approximately in the middle of the barrier island (El Torno sector). Blue: Littoral drift, red and black: currents at 3 m to 8 m and 8 m to 20 m isobaths respectively, x-axis in UTM-18N coordinate system. Taken from Posada et al.(2009)

The aforementioned northeast drift reversal during the wet season has been reported to generate high frequency waves inducing strong coastal erosion in deﬁned sectors along the barrier island, mean erosion rate for the sector has been calculated as 3 m yr−1 with a maximum of 19 m yr−1 in the point

21 known as Km19 where a strip of RN90 highway is at serious risk of being destroyed (Posada and Henao, 2008; Rangel-Buitrago et al., 2015).

2.4.1 Internal waves Circulation and mixing patterns in the system are completely wind driven due to its shallow nature and the large exposed area susceptible to wind inﬂuence (wind fetch). These conditions are ideal for the generation of internal waves that can become one major factor inﬂuencing dispersion and turbulence parameters in the system (Fig. 2.14).

Fig. 2.14: Internal wind driven waves registered during the ﬁeld campaign on April 14th of 2015

The inﬂuence of wind driven waves on the nutrient cycles of the system can be very important due to enhanced oxygen dynamics in the water-air interface, sediment resuspension and consequent nutrient desorption and impact on microbiological rates.

22 CHAPTER THREE

MATERIALS AND METHODS

It is critical that modelling initiatives in the system start including long term simulations instead of the phenomena-specific short term simulations performed until this date. Moreover, they should be based in more up to date measurement campaigns, like the ones hold by INVEMAR each two months. Unfortunately, these campaigns do not include physical or hydrodynamic variables like water levels, velocities or tracer studies for diffusion coefficients, but rather water quality variables and ecological indicators. As it was previously stated in this report, data availability in CGSM-LC is limited due to difficulties on physical access to sampling sites, budget restrictions, institutional competence issues and security risks. This is specially true for hydrodynamic variables as water levels and velocities that are normally required for model calibration and validation but imply more complex sampling methodologies, more costly equipment and larger field work schemes. In order to address these data limitations, the project intended to simulate a reference situation (generic year) under the most representative boundary conditions for the model domain, according to the throughout description of the study area and system’s behavior available in previous studies and observed trends recorded during INVEMAR monitoring program for water quality variables. This generic year approach is mentioned by Neves(2007) as the first requisite in order for a model to be used as a Decision Support Tool at institutional level.

3.1 Model domain and software selection

Due to the small size or semi permanent nature of many of the creeks and streams inside CGSM-LC, there is not enough available information nowadays to construct an integrated model comprising all the existing interrelations between the water bodies inside the lagoon complex. Therefore, a selection criterion for water bodies, based mostly on the existence of bathymetric information but also in the relevance of the zones for ﬁsheries, economical activities, ecological resources and historical importance for the local communities was applied. The ﬁnal selected model domain is presented in Fig. 3.1, with its most important components described below:

23 • Coastal area: Extending from the line of low tide of the barrier island (Isla de Salamanca) and encompassing the continental margin until approximately 36 km oﬀshore, the northernmost limit of the domain is an open boundary where tidal conditions are to be imposed during the simulations.

• Estuarine area: Includes the two main lagoons of the system (CGSM and Pajarales) as well as the connecting channel know as Caño Grande and sole seawater inlet know as La Barra. This area holds most of the ﬁshing resources of the complex and coincides with a local administrative ﬁgure called exclusive reserve area, declared on 1978 for the exploitation of hydrobiological species (Vilardy and González, 2011).

• Tributaries: Only the largest freshwater inputs from the two contributing areas are included in the model, these are:

– West face of Sierra Nevada de Santa Marta mountain range: Fundación river, Aracataca river, Sevilla river. – Magdalena river delta: Clarín creek, Aguas Negras creek.

-74,788424 -74,643658 -74,498892 -74,354126 -74,209360

Caribbean Sea ,190540 ,190540 11 11 ,071486 ,071486 11 11 ׭׳קף ,952432 ,952432 10 10 ,833378 ,833378 10 10

׳ץ׬סײסתסװ 714324, 714324, 10 -74,788424 -74,643658 -74,498892 -74,354126 -74,209360 10

؁ؒؒ؁ע׀؁׬ ؎ڍؒ؁،ף׀؏ڑ؁ף ؄؅؎؁ؒק׀؏ڑ؁ף ؓ؁׮؅؇ؒ׀ؓ؁؇ؕס׀׎ף Fig. 3.1: Model domain extents and points of special interest in the system. Right: Main water bodies, bottom: system connections and inlets. In the map, coastal area is depicted in a lighter blue than estuarine area

As for the selection of the modelling suite to use for the project, an INVEMAR internal report by Bastidas-Salamanca et al.(2012) thoroughly analyzed the advantages and disadvantages of several

24 state of the art numerical models for their implementation in the institute according to their numerical schemes, data requirements, available institutional resources and deﬁning selection criteria regarding the water quality problem to be addressed and transport processes to be included. According to the intended simulation aims (i.e. non-conservative substances), the required model had to be able to simulate both hydrodynamics and transport and have the possibility to include nutrient cycle related processes. Furthermore, due to the already mentioned data limitations in the area a structured grid approach was preferred over unstructured grids due to better consistency and accuracy of the former. In this order of ideas, DELFT3D and MOHID modelling suites were the most likely options for the project. The ﬁnal decision was to use MOHID modelling suite because the personnel at INVEMAR received a training course on the use of this model by ActionModulers Inc. in 2015, which ensures the institutional capacity for further developing and adapting the intended model, as well as using it in new modelling proposals in the near future. Thus increasing the impact of the project on the area of study.

3.2 MOHID model description

MOHID Water Modelling System (from "MOdelo HIDrodinâmico" in Portuguese, translates to Hy- drodynamic Model) is a finite volumes hydrodynamic and water quality modelling platform developed in ANSI FORTRAN 95 by Instituto Superior Técnico (IST) at the University of Lisbon, Portugal. To be more precise, the model used in this project corresponds to the component known as MOHID Water, while the MOHID modelling system comprises also two additional models for catchment simulations (MOHID Land) and in-stream river processes (MOHID River) that can be coupled to Mohid Water in order to achieve an integrated catchment-to-coast modelling approach. In behalf of simplicity, from now on when this document refers to MOHID it should be understood as refering to MOHID Water model only. The model uses an object oriented structure and is capable of simulating hydrodynamics, waves, sediment transport, conservative and non-conservative substances transport (with eulerian and lagrangian approaches), oil spills (including beaching and free vertical movement) and eco-hydrological processes (Ascione Kenov et al., 2014). MOHID works exclusively with structured grids for both the horizontal and the vertical domain, however, different types of grids are available for each plane based on the most common morphological phenomena that can be found on each of them (e.g. straits, water surface fluctuation, sedimentation). In the horizontal domain, grids can be laid out as constant spacing (regular), variable width, nested domains or curvilinear meshes, while in the vertical domain both the classic Sigma and Cartesian grids, as well as a series of adaptive grids like harmonic or Lagrangian grids (mostly used in reservoir models) can be implemented. Physical compartments of the comprehensive water system (i.e. Water column, sediment matrix, benthos and atmosphere) are managed as independent sets of properties and parameters in MOHID via a modular approach (each compartment has a separate input .dat file), while the interactions

25 between them are handled by separate "interfaces" where main forcing functions are calculated, as presented in Fig. 3.2.

Air-Water Interface

Own velocity Physical processes Water properties Water quality Dissolved Waves properties Adsorption Currents Advection Di usion Oil Turbulence Particulate properties Macroalgae

Benthic Interface

Physical processes Sediment properties Diagenesis Dissolved Bioturbation Interstitial properties Waves water ow Advection Consolidation Particulate Di usion Seaweeds processes properties

Fig. 3.2: Conceptual description of MOHID showing the main compartments and processes. Adapted from Ascione Kenov et al.(2014)

Numerical methods and transport phenomena are also treated as separated modules inside the model, some of the most important modules for this project are the ones containing parametrizations for the turbulence closure problem (Turbulence.dat), eulerian (WaterProperties.dat) and lagrangian (Lagrangian.dat) transport and biogeochemical cycles (WaterQuality.dat). As all surface water numerical models, MOHID’s hydrodynamic module is based on the 3D Navier- Stokes conservation equations for incompressible ﬂuids. For the particular case of MOHID, the following assumptions are made:

• Hydrostatic distribution of the pressure in z direction.

• Boussinesq approximation for buoyant ﬂuids.

• Reynolds averages approach for turbulent ﬂuctuations in the momentum equations. A throughout description of the primitive equations for ﬂuid motion and their main assumptions is beyond the objective of the present report, however, a summary of these concepts is provided in AppendixA for further reference.

26 The following generic diﬀerential conservation equation is used by MOHID to describe constituent accumulation and transport (Ascione Kenov et al., 2014):

∂β ∂ ∂β = − uj β − ϑ + (So − Si ) (3.1) ∂t ∂xj ∂xj

Where:

β: Water property xj : Distance in j direction ϑ: Eddy diﬀusivity t: Time

uj : Velocity in j direction So − Si : Diﬀerence between sources/sinks

In Equation 3.1, the first term on the right hand side stands for the advective/diffusive transport of the substances by water movement while the sources/sinks term encompasses all the additional processes for mass transport/conversion inside the control volume and through the interfaces of the model. Sources and sinks are highly pollutant-dependent and their definition for each major constituent involved in water quality modelling has been addressed by many authors in the last decades. For sources and sinks of nutrients an oxygen in estuarine systems, and particularly CGSM- LC, the works of Statham(2012) and Tuchkovenko and Calero(2003) can be consulted. In order to ease the preprocessing of the spatial information, numerical grids and input files for MOHID, as well as the postprocessing of the model outputs (HDF5 files and time series), Action- modulers Inc. provided the author with an academical license for their propietary user interface (MOHID Studio v 2.2).

3.3 Hydrodynamic model

3.3.1 Grid selection

As it was previously mentioned, MOHID model works with structured grids in the horizontal and vertical domains. For this project two different grids were used: A regular grid of 500 m resolution and a variable grid with finer discretization in the main constrictions of the system, namely La Barra and Caño Grande (Fig. 3.3). The regular grid was used as a preliminary approach to assess the response of the model to the addition of new forcing functions and solve initial processing issues, this method proved to be very efficient while a grid of this resolution implies very low computational effort and allows the modeler to debug the code in short time. However, this resolution implies very high numerical instabilities and could hardly represent the processes in the constrictions of the system, where the size of the representative elementary volumes were much larger than the scale of the channels. In order to run representative simulations according to the system’s shape, a variable resolution grid was constructed using the work of Camacho(1991) as a reference, where a 500 m to 200 m grid of this type was used to simulate flows in Boca de la Barra inlet, refining the grid gradually across 10 elements. As the constrictions of the system are narrower than 200 m and a more detailed bathymetry

27 Fig. 3.3: Numerical grids implemented in the project. Left: Regular grid with 500 m resolution, right: variable grid with gradient to finer resolutions (50 m) in the main straits. than the one used by Camacho(1991) was available, the grid for this project was constructed with a 500 m to 50 m gradient implemented gradually across 20 elements (22.5 m reduction per element) for Boca de la Barra and a 500 m to 250 m gradient across 10 elements for Caño Grande (25 m reduction per element). As it has been mentioned in preceding chapters, water bodies inside CGSM-LC are very shallow and highly influenced by wind stresses, therefore no salinity or temperature gradients in the vertical direction are to be expected (Wiedemann, 1973). Accordingly, the model was designed following a 2D depth-averaged configuration with only one Sigma vertical layer for the whole domain.

3.3.2 Boundary conditions and input data In MOHID architecture, variables and boundary conditions can be initialized in several ways according to the nature of the process in the area. Variables can be initialized either as constant values (CONSTANT), time dependent values (TIMESERIE), spacially distributed values (RASTER, BOXES) or space-time dependent values (HDF5). Code 3.1 shows an example of a Dirichlet time dependent flow boundary condition in one of the rivers feeding the system. Due to the nature of some physical processes (e.g. bathymetry), they can be only input using certain type of initialization scheme (i.e. RASTER). However, the model is flexible enough to let the user start the variables in the simplest way (CONSTANT) and change the initialization condition afterwards to cope with data availability and the trial and error procedure that characterizes a numerical model development. The main boundary conditions applied to CGSM-LC according to the selected domain extents and simulation specifications were:

• Open boundary condition: Forced by triangulated tidal harmonics (amplitude, phase) from random positions in a zig-zag pattern across the open sea limit of the model domain (recommended method from MOHID manual).

28 • Bottom boundary: Defined via a bathymetry file (RASTER), in this boundary advective fluxes were considered null while diffusive transfer of momentum is driven by the shear stresses calculated using a no-slip condition and a quadratic law of the wall. The evolution of the velocity profile is calculated using a constant or space variable bed roughness coefficient (CONSTANT or BOXES). Mass transfer through this boundary is possible via the Interface Sediment-Water module.

• Free surface boundary: A null advective ﬂuxes boundary with diﬀusive transfer of momentum via wind surface stresses imposed via wind velocity and direction values (CONSTANT or TIMESERIES). Mass transfer through this boundary is possible via the Interface Water-Air module.

• Land boundary: Defined as impermeable closed boundaries with zero normal and diffusive momentum water fluxes (free-slip condition).

• Dirichlet boundaries: Deﬁned in the river and creek mouths by imposing constant or time- dependent ﬂow and concentration values (CONSTANT or TIMESERIES).

For the developed model, no moving boundaries (intertidal areas) were taken into account.

Code 3.1: Example code of a time dependent streamﬂow initialization !RIOS NAME :Rio_Sevilla DESCRIPTION : Mean multiannual monthly values I_CELL : 44 J_CELL : 134 DEPTHBYGRID : 1 K_CELL : 1 DATA_BASE_FILE : ../General Data/Boundary Conditions/ Rio_sevilla.srr FLOW_COLUMN : 2 TIME_SERIE_COLUMN : 1 DISCHARGE_UNIFORM : 1

In accordance to the strong seasonal patterns in the study area and the large time scales required for nutrient transport processes in estuarine systems, mean multiannual monthly values were used for the time dependent boundary conditions and main system variables in CGSM-LC and its tributaries. Monthly variations were used as adequate estimators for the behavior of the system in a generic year due to data availability and the system climatic patterns described in section 2.3. In the following subsections, a description of main input data used to conﬁgure the model and the applied processing techniques will be presented.

29 Bathymetry

For the main lagoon (CGSM) and the inlet known as La Barra, data from a bathymetric survey realized by INVEMAR in 2015 was used, this data was gathered using a multi-frequency single beam ecosounder ODOM HYDROTRAC Model HT97001 in 100 m transects with a finer resolution of 50 m in the inlet. Data for Pajarales lagoon was extracted from bathymetric chart COL-1203 (CIOH et al., 2001) which is the most up to date information for this area. In order to obtain bathymetric information for the open sea areas included in the model domain, data with a 3000 resolution was obtained from Global Bathymetric Chart of the Oceans (GEBCO) project from the 300 m-Isobath to the open boundary. Between the coastline and the 300 m-Isobath data from COL-1203 was preferred due to possible satellite inconsistencies near the coast. All the information (12 037 data points, Fig. 3.4) was processed in ArcGIS via an iterative finite difference interpolation (TopoToRaster) using the coastline as solid boundary. The output raster file was loaded in MOHID Studio as base bathymetry for the domain.

-74,777410 -74,632644 -74,487878 -74,343112 -74,198346 ® ,196436 ,196436 11 11 ,077382 ,077382 11 11 ,958328 ,958328 10 10 ,839274 ,839274 10 10 Data source (Nr. of points)

COL-1203 (339) GEBCO (1.779) Invemar_2015 (9.919) ,720220 ,720220 10 10 -74,777410 -74,632644 -74,487878 -74,343112 -74,198346 Fig. 3.4: Bathymetric information compiled from diﬀerent sources as input for MOHID Studio

For some areas of the system, where no recent bathymetric information was available, old bathymetric maps from the 70’s and 80’s where consulted in INVEMAR’s document center during the ﬁeld campaigns between March and April of 2015 in order to obtain coarse estimations of bed gradients. The most representative case of the use of this historical gradients is Caño Grande connection where vertical proﬁles surveyed in 1972 had to be used to get a picture of the approximate depth pattern

30 along the length of the channel where no further information could be found. This information was supplemented with local knowledge from inhabitants of the area. Based on the theory that in deep water conditions swells and wind generated waves do not generate bottom shear stresses and no strong bottom currents have been identified in the coastal area of the model domain, a cross validation procedure against the data reported in bathymetric chart COL- 1203 for the outer coast zone was used to verify the precision of GEBCO project data for the study area and discard any outliers in both data sources. Different interpolation procedures included in MOHID Studio (i.e. IDW, triangulation, mean values) were used according to the desired grid configuration in each simulation, the sedimentary behavior of the complex and the minimal depths registered in subsection 2.2.1, all interpolations were limited to a minimum water depth of 10 cm.

Tides and open sea boundary

Input data for the model in the form of tidal components (harmonics) for several random points across the open boundary were obtained from FES2004 global tidal solution (Lyard et al., 2006). These points were triangulated by the model to obtain a complete tidal ﬁeld forcing the outer elements of the open boundary. Absolute water levels for two gauging stations inside and nearby the model domain (Table 3.1) were obtained from DIMAR’s mareographic monitoring network via the sea level monitoring service from Intergovernmental Oceanographic Commission (IOC). These data were analyzed using MATLAB’s T_Tide routine developed by Pawlowicz et al.(2002) in order to extract the tidal components and verify/adjust the harmonics reported by FES2004 with local conditions (results in AppendixB).

Table 3.1: DIMAR gauging stations used for local tide analysis

Station Lat/long Frequency Data period Santa Marta 11.235118/-74.2216 Minute 2014-01-01 – 2015-08-02 Puerto Velero 10.944722/-75.04083 Hourly 2013-08-24 – 2014-12-31

As the data from the gauge inside the model domain has a very short recording frequency (i.e. every minute), much noise is to be expected on the general tide signal. Therefore, moving average methods with diﬀerent time spans (hourly, semi-diurnal, diurnal) were used to ﬁlter the tidal signal before comparing it with the one simulated with FES2004. In order to convert absolute recorded values to relative surface levels (zero-centered), monthly mean levels were used after a visual exploration of the recorded series trends.

Meteorological data

Daily records of meteorological variables for the area were requested from the Colombian weather monitoring network managed by IDEAM. Reported stations, variables and years covered are summarized in Table 3.2.

31 Table 3.2: IDEAM meteorological data used as input for the project. SH: Sunshine hours, WD: wind direction, EVR: real monthly evaporation, RH: relative humidity, PREC: precipitation, RV : cumulative wind traveled distance, TEM: temperature (max, min, mean), WS: Wind speed.

Station Lat/long Variables Recorded period1 A.E CORTISSOZ [29045020] 10.8834 / -74.7798 SH, WD, EVR, RH, 1941-2015 PREC, RV, TEM, WS APTO SIMON BOLIVAR [15015050] 11.1283 / -74.2289 SH, WD, EVR, RH, 1952-2013 PREC, RV, TEM, WS BONGO EL [29060030] 10.6488 / -74.3755 PREC 1975-2014 COCOS LOS [29060080] 11.0082 / -74.6869 PREC 1967-2014 ENANO EL [29060160] 10.8909 / -74.1884 PREC 1974-2013 ESPERANZA LA [29060180] 10.7425 / -74.3063 PREC 1975-2014 FLORES LAS [29045120] 11.0339 / -74.8198 SH, WD, EVR, RH, 1980-2014 PREC, RV, TEM, WS PADELMA [29065020] 10.7211 / -74.1997 SH, RH, PREC, RV, 1967-2013 TEM PALMA LA [29060210] 10.9651 / -74.2034 PREC 1967-2014 PALO ALTO [29060270] 10.7225 / -74.2719 PREC 1967-2014 POLY LA [29060220] 10.8167 / -74.1833 PREC 1972-2002 PRADO SEVILLA [29065030] 10.7642 / -74.1547 SH, WD, EVR, RH, 1970-2013 PREC, RV, TEM, WS PUNTA BETIN [15015030] 11.2500 / -74.2167 SH, RH, PREC, TEM 1969-1980 RUBY EL [29060550] 10.8451 / -74.1882 PREC 1984-2014 SAN ISIDRO [29060280] 10.8837 / -74.2126 PREC 1967-2014 SAN RAFAEL [29060540] 10.5906 / -74.6637 PREC 1982-2014 SARA LA [29060230] 10.8353 / -74.1637 PREC 1971-2011 SEVILLANO [29060310] 10.9334 / -74.2525 PREC 1970-2014 TASAJERA [29060120] 10.9762 / -74.3618 PREC 1965-2014 YE LA [15015020] 10.9924 / -74.2111 SH, EVR, RH, PREC, 1969-2013 RV, TEM

Even though complete daily logs were available for certain variables in speciﬁc years, most meteorological datasets showed large amounts of missing data and even lack of measurements for several months/years in some areas, moreover, not all datasets had the same record lengths and therefore a complex additional component of time series analysis would have been necessary in order to ensure these data consistency and complete time-dependent inputs for the atmospheric compartment. Such an analysis exceeds the expected aim of this work but nevertheless would be important for subsequent model optimization initiatives. In order to impose an adequate representation of the meteorological conditions during the "generic year" approach in the area of study, the following procedure was implemented:

• Extract mean monthly values for all the variables in their respective recording periods

1Not constant for all variables in the station

32 • Compare graphically the multiyear behavior of each variable and identify possible outliers and impacts from extreme weather phenomena like ENSO, La Niña or cold fronts, exclude these data from further analysis. References to the impacts of extreme weather conditions in the area like Blanco et al.(2006) and Ortiz et al.(2013) were particularly useful in this step.

• Extract synthetic monthly weather time series for standard conditions in a generic year.

• Assess if the synthetic time series show enough internal variability to keep them as a time dependent boundary conditions or simplify them to constant values throughout the simulation. For this assessment a simple statistical parameter, the coeﬃcient of variation, deﬁned as the ratio between standard deviation and mean value, was used.

For easing the management and analysis of the raw data from IDEAM’s format during the ﬁrst step of the aforementioned procedure, a software tool to extract monthly meteorological means was developed in Visual Studio 2013.

Hydrological data

Mean monthly ﬂow data for the main rivers draining in the Eastern part of the system were provided by INVEMAR from datasets acquired from IDEAM’s national hydrological monitoring network for previous projects (until 2014 national environmental raw data was not free of charge to the public in Colombia). Table 3.3 shows the data periods and stations used for the analysis.

Table 3.3: IDEAM monthly ﬂow data provided by INVEMAR for the area of study

Station Stream Lat/long Data period CANAL FLORIDA [29067050] Sevilla river 10.7558/-74.0863 1965-2005 RIO FRIO [29067070] Frío river 10.9054/-74.1541 1967-2013 PTO RICO HDA [29067060] Fundación river 10.5000/-74.1333 1965-2002 GANADERIA CARIBE [29067150] Aracataca river 10.5749/-74.1267 1965-2005

A similar procedure to the one mentioned for the meteorological data was used to analyze the mean multiannual monthly ﬂow values for the rivers and create synthetic time series to feed the model in a "generic year". Hydrometric measurements were carried out for the two creeks in the model domain using a SEBA type F1 propeller ﬂow meter mounted on a 2 m wadding rod and a guiding rope across a representative cross section of the stream with verticals every 1 m. For Clarín creek in-stream measurements using a two-point method were used while for the deeper Aguas Negras creek, on-board measurements from the vessel were necessary since the available wadding rod was not large enough to reach the bottom of the stream and consequently the one-point method was preferred, in order to measure the bottom depth and select the measuring position, a sounding line with a plummet was used.

33 Water quality data

Data from the national marine water quality monitoring program and the speciﬁc monitoring program for CGSM-LC were provided by INVEMAR for the last 15 years. Mean monthly values from these datasets were imposed as initial conditions for the modelling domain according to the simulations start dates. A BOXES initialization method was used for the water quality variables in order to impose space dependent values for the ocean, CGSM and Pajarales lagoon, also variables for the tributary rivers were extracted from the database. The main variables that were used for the project were: Salinity, water temperature (◦C), Ammonia (mg L−1 N), Nitrate (mg L−1 N), Nitrite (mg L−1 N), dissolved Oxygen (mg L−1), inorganic phosphorus (µg L−1) and Chlorophyll alpha (µg L−1).

3.3.3 Residence time calculation

Hydraulic retention time (also called residence or flushing time) is one of the most important parameters in estuarine constituent transport while it defines the minimum time the pollutants will stay in the water body and will be subjected to the conversion processes. In MOHID, a lagrangian approach is used to calculate hydraulic retention times, the method is based on releasing a certain number of particles representing water parcels from zones of interest inside the system (emission boxes) in a way their collective volume equals the initial volume of the water body. During the simulation, this particles are followed individually in order to identify the remaining volumetric fraction from the tracers still inside the monitoring boxes (these can be different from the emitting ones). The following equation is used to calculate the remaining volumetric fraction of tracer volume in box i from origin j:

Vi,j (t) fi,j (t) = (3.2) Vi,i (0) Where:

i: Monitoring box Vi,j (t): Instant volume of tracers.

j: Emitting box Vi,i (0): Initial volume of box i

Hydraulic Retention Time (HRT) is calculated as the time needed for 80 % of the initial water volume to leave the system (Braunschweig et al., 2003)

3.4 Nutrients and oxygen model

According to MOHID paradigm, biogeochemical cycles are modeled using a 0-D approach in the WaterQuality module, in this module the interconnections between cycles and diﬀerent biological compartments can be activated/deactivated via keywords. As little information is currently available for most biochemical parameters inﬂuencing the nutrient, oxygen and biological cycles in CGSM-LC, a conservative approach including only phytoplankton and zooplankton was implemented.

34 Initial conditions for the lagoons, the coastal area and the rivers flowing into the system where taken from INVEMAR databases as mentioned in previous subsections. In the particular case of phytoplankton, values of Chlorophyll alpha were converted to plankton carbon content multiplying by a constant 50 g g−1 Redfield ratio. Values for the main biochemical parameters in the nutrient cycles and biochemical equations (see AppendixC for further details) were obtained from Gordon, Jr. et al.(1996) which is one of the most extended sources in the field. An additional work by Tuchkovenko and Calero(2003) with local calibrations for some of the parameters in the complex seemed to have a high potential but could not be implemented on this stage of the modelling.

3.5 Simulations

Accounting for the probable long retention times in the system and the "generic year" approach described mentioned at the start of this chapter, simulations were run in a one year window starting in April. April was chosen as the start month for the simulation in order to coincide with the end of the dry season in the region, when the harsher ecological conditions are to be expected. As this was the ﬁrst numerical modelling project based on INVEMAR databases on the area, a wide range of conditions were simulated in order to evaluate the main drivers of system circulation and transport patterns. The most important tests performed where:

• Initial short-term simulations with only tidal forcing in order to determine an appropriate spin-up period for model convergence.

• Hydrodynamic simulations according to bottom rugosity:

– Constant roughness coeﬃcient – Space dependent roughness coeﬃcient according to Camacho(1991) and Flórez(2013),

with Chèzy coeﬃcients (Cz ) of 125 for the main constraints and 60 for the lagoons. – Space dependent roughness coeﬃcients according to literature and bottom materials.

• Hydrodynamic and conservative transport according to river and creek ﬂows:

– Mean multiannual monthly ﬂow values according to IDEAM with reduced streamﬂow in the East margin due to water withdrawal from Zona Bananera, according to (Bernal and Betancur, 1996, pag. 59).

– Constant low ﬂow conditions in the creeks during the year (simulating a strong ENSO phenomena, common in the latter years for the area).

– Constant high ﬂow conditions in the creek (general approach used in previous models). – Random ﬂow distribution in the creeks based on regular climatic periods in the area.

• Hydrodynamic and conservative transport simulations by meteorological variables:

35 – Constant hydric deﬁcit in worst case scenario (−1150 mm yr−1). – Space dependent rain values according to IDEAM. – Constant wind and velocity direction using the most frequent values in the area. – Time dependent wind and velocity direction (monthly).

• Lagrangian tracer simulations for residence time inside CGSM and Pajarales systems.

Simulations were run on different system configurations (Table 3.4) in order to analyze computation times and machine effort according to available system resources, as well as minimum requirements for running the model in acceptable times for decision making.

Table 3.4: System conﬁgurations used for simulations during the project

Conﬁguration OS System characteristics Linux Server Ubuntu Linux 14.04 2 x Quad-Core AMD R OpteronTM Processor 2356 (8 cores x 2.3 GHz), 32 GB RAM Windows desktop Microsoft Windows 7 64 bit Intel R CoreTM i3 x 2.93 GHz, 3.96 GB RAM Windows laptop Microsoft Windows 7 64 bit Intel R CoreTM i5 - 2430M (2 cores x 2.4 GHz), 4 GB RAM

3.6 Calibration

Hydrodynamic calibration in the project was a major challenge due to the already mentioned lack of hydrodynamic measurements and monitoring inside the system. Calibration attempts for the hydrodynamic model were performed by modifying two main parameters: Bottom roughness coefficient (CZ ) and horizontal eddy viscosity (NH ). Results were compared graphically against a depth-averaged velocity dataset derived from a 28 hours (minute frequency) velocity measurement survey performed at La Barra strait by Marine geosciences program at INVE- MAR (GEO) program from INVEMAR during July of 2015. Another fit verification implemented in the project was to compare the annual salinity patterns observed in different areas of CGSM-LC during INVEMAR’s monitoring program versus the modelling results for long term simulations and adjust probable sources of uncertainty in the input data as boundary conditions.

36 CHAPTER FOUR

RESULTS AND DISCUSSION

4.1 Input data preprocessing

4.1.1 Bathymetry The comparison between GEBCO data and ﬁeld data from bathymetric chart COL-1203 (Fig. 4.1) revealed the existence of an important outlier of 561 m registered inside the chart, which could be a very local accident measured during the survey. The rest of the data showed a very good congruence with the satellite data and therefore it was concluded that GEBCO project data are very precise for the study area and can be included without further consideration into the bathymetric dataset of the model domain for its deeper areas.

Fig. 4.1: Cross validation process of bathymetric data from GEBCO and bathymetric chart COL- 1203. An excellent coherence between both data sets was observed, the outlier from COL-1203 was excluded from the ﬁnal bathymetry

After interpolation inside MOHID, it was common for the final bathymetry to show undefined elements according to the grid resolution and the scarcity of the data in certain areas of the system like Pajarales or Sevillano lagoon. These cases had to be addressed independently and decide which bed value could fit the element the best according to nearby patterns and ensuring the connectivity between all the areas in the system. At the end of the procedure, an additional tool from MOHID Studio (Filter_bathymetry.exe) was used to smooth the bed gradients based on a maximum accepted slope between two neighboring elements (0.1 % to 0.2 %). As it can be appreciated in Fig. 4.2, bathymetric contours for the coastal and open ocean areas of the domain clearly maintain the geomorphological features described by von Erffa(1973) with the

37 three lobe conﬁguration almost intact after 40 years from their initial description.

-74,798500 -74,538425 -74,278350

50 1000 900 800 800 700

600

500

400

600 300 50 250 700 150 50

450

,175466 350 ,175466 11 11

200 100 50 20 µ 40 30 ,979555 ,979555 10 10 ,783644 ,783644 10 10

02 4 8 12 16 Kilometers

-74,798500 -74,538425 -74,278350

Fig. 4.2: Bathymetric contours of the continental margin in front of CGSM-LC produced in this project and comparison with von Erﬀa(1973) delta lobes.

According to the interpolated bathymetry, the continental shelf directly in front of La Barra inlet shows a soft gradient from 0 m to 40 m in the ﬁrst 15 km (bottom slope: 0 26% presumably caused by deposition of soft material from previous displacements of the mouth of Magdalena river and the sediment output from CGSM (Posada et al., 2009). After 15 km (continental slope) the gradient increases a ten-fold reaching a depth of 1000 m after approximately 36 km (Fig. 4.3).

Fig. 4.3: Bathymetric gradient of the continental margin in front of CGSM

The integrated inner bathymetry of CGSM-LC, including the new data from INVEMAR’s recent survey in 2015 is presented in Fig. 4.4. The observed deposition patterns in the system give a fists approximation of the prevailing flow patterns inside the estuary. Sediment accumulation starting at the mouth of Fundación river and following the Eastern bank all the way to a large sediment bed from the mouth of San Joaquín creek to La Barra can be observed. This sediment bank can be considered an indicator of the historical position of the null zone, where horizontal velocities tend to zero due to the clash of landward tidal currents and seaward river flows, which is usually considered a zone of maximum sedimentation (Martin and McCutcheon, 1999).

38 4,5 4 2 4 10 1 12 11,5 1,5 11 10,5 2 1 9 9,5 2 3 7,5 8,5 1 2,5 3,5 7 8 1 2 3 5 6,5 6 5,5 2,5 4,5 2,5 1 2 Caño Clarín 2 2,5 2,5 2 2 1 1,4 1,8 1 1 1 1 3,2 1 1 1 3,5 1 1,51,4 1 1 3 2 2,6 1 1 1 1,5 1,4 2 1 1,5 1 1 2,5 1 1 2 1,5 2,8 2,6 5 1,5 Caño1,2 1 Bristol ó Sucio 1,5 2,2 4 2,4 3,8 3 1,2 1,5 2 1,5 1,6 1 1,6 Covadoó 3,2

1,8 ElIndio 2,8

Caño 0,6

2,5 2 1,2

1 0,6 1 0,6 1 1 2,4 2 1 0,8 1 1 0,5 0,8 1 0,8 2,5 1 1,4 2 1 1,8 1 1,2 1 1 0,6 1,6 1,5 2 0,8 1 1,5 1 1 1,5 2 2 1,2 1,6 1 2 0,8 2 1,5

0,5

1 0,6 1,8 2 2 0,8 2 Caño 2 1 Hondo 2 2 0,8 0,8 0,6 1,4 2,2 2 1 2

2 2 2 1,5 1,5 Caño 1,5 El Solo 1 1,5

1 1,5 0,5 0,5

0,5 1,5 0,5 1,5 2 1,5 2 Río Frío

0,5

2 2 1,5 0,5 1,5 1 0,5 1 2 0,5

1 1 1 0,5 1,5 1,5

Río Sevilla 0,5 1

0,5

0,5 0,5 1 Caño San Joaquin

1 0,5 1 1

0,5 1

1 1 2

Micos 0,5 Caño Los 1 1,5

2 1 1 Caño Pájaro 1,5 1 1 1 1 1,5

0,5 Alfandoque 1 Caño Caño Doncellas Río Aracataca 1 ó Paraíso

1 1 1

Fig. 4.4: Bathymetric map of CGSM using data from GEBCO, COL-1203 and new survey from INVEMAR. Red numbers correspond to 50 cm contour lines

Similar to the sediment bank in the Northeastern part of the system, the evident depression in the Nortwestern corner of the lagoon could be an indicator of very fast water movement and high bottom shear stress induced by the mouth of Clarín creek. However, this hypothesis cannot be veriﬁed just with the bathymetric information while this creek is known to be aﬀected by sedimentation problems along it’s path from Magdalena river, considerably reducing its sediment load by the time it arrives to CGSM, these latter arguments were reinforced during the hydrometric survey when a greenish color due to colloidal organic material in the water and a high amount of organic bottom debris where evidenced near the mouth of the creek instead of coarse sediment. Consequently, the depression on the Northwest can be only the natural pristine original bottom depth of the system.

4.1.2 Atmospheric variables

After the analysis of the yearly patterns from the weather datasets by the procedure described in section 3.3.2, the following monthly synthetic time series were deﬁned for the atmospheric drivers of the system during a generic year: Data for wind direction was much more scarce than the data for wind velocity, therefore for this variable a distribution based on literature description of its behavior for the study area was used. The pattern used for wind direction was a prevailing wind of the NE between December and August with a ﬂuctuation to the S in September, SW in October and S again in November (wet season).

39 Table 4.1: Mean monthly conditions for atmospheric variables in CGSM-LC

Variable Units JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC Mean temperature ◦C 27.252 27.648 28.086 28.423 28.415 28.352 28.252 28.087 27.818 27.504 27.553 27.368 Wind velocity m/s 3.997 4.190 4.219 3.917 2.940 2.622 2.877 2.785 2.414 2.252 2.495 3.386 Relative humidity % 77.654 76.261 76.195 77.636 79.881 80.140 79.758 80.652 81.545 82.681 81.732 79.737 Cloud cover % 26.588 29.883 40.185 43.465 49.389 46.388 44.244 46.750 52.287 51.632 45.813 36.429 Evaporation mm/month 161.420 163.751 188.770 176.698 157.876 145.220 155.713 150.727 132.500 125.326 121.959 140.172 Precipitation mm/month 2.100 1.555 5.964 40.901 120.451 111.826 87.551 126.845 168.423 200.913 103.409 23.643

In order to verify whether the deﬁned synthetic series keep the original structure of the meteorological periods in the study area during standard years (no inﬂuence of major climatic phenomena), a Walter & Lieth climogram was prepared Fig. 4.5. From the analysis of the climogram, an excellent match with the characteristic climatic patterns of the zone can be observed with a very arid period between December and April and the bimodal rain pattern during the rest of the year.

10°44-11°00N 74°16-74°38W °C CGSM-LC mm Altitude: 4m 400

300

200

100

40 80

30 60

20 40

10 20

0 0

-10 JFMAMJJASOND Fig. 4.5: Walter & Lieth diagram (climogram) for the region of CGSM-LC.Red line: mean air temperature, blue line: mean precipitation, yellow areas indicate arid periods while dark blue areas represent wet periods, light blue areas denote semi-arid conditions

In the last step of the analysis, the final decision regarding to initialize the atmospheric variables as constant or time dependent values was taken based on their internal variation, calculated by a very simple method using the coefficient of variation. This is an important step when using a "generic year" for the simulation while including seasonal fluctuations in variables that really don’t deviate that much from their mean can add noise to the model’s results rather than important information, besides, the extra complexity of the model can unnecessarily increase the computational effort of the simulation if more than the minimum sources of uncertainty are added to the configuration (Occam’s razor). Selection of the threshold value for the coefficient of variation in order to determine which variables should be included as time series in the model was a trial and error procedure, based on simulating the inputs as constants first and afterwards as time variables keeping all the other variables equal. This procedure was not applied to all the variables due to the long computation times for the model,

40 however the tests shown that variables deviating more than 15% from their yearly mean normally had more inﬂuence in the model results when used as time series. The ﬁnal decision over which variables to use as time series inputs is summarized in Table 4.2.

Table 4.2: Analysis of atmospheric variables for annual ﬂuctuation and initialization methods. CV : Coeﬃcient of variation, variables with CV>15% (red) are initialized as monthly variable inputs inside the model.

Variable Units Range Mean CV Initialization method Mean temperature ◦C 27.25-28.42 27.9 2% CONSTANT Wind velocity m/s 2.37-4.64 3.42 24% TIMESERIES Relative humidity % 76.2-82.68 79.5 3% CONSTANT Cloud cover % 26.6-52.4 43 19% TIMESERIES Evaporation mm/month 122-188.7 151.5 13% CONSTANT Precipitation mm/month 1.56-200.9 82.66 82% TIMESERIES

Results show that both wind velocity and wind direction fluctuations during the year are important to the system, this can be explained due to the large fetch of the lagoons and the strong influence of these parameters over the system’s mixing. Change in cloud cover can highly influence light availability inside the lagoon and consequently the growth of phytoplankton in the system, which is particularly important in this complex due to its previously mentioned high primary productivity and its influence on oxygen production/depletion. Due to the fact that CGSM-LC is located close to the Equator and therefore is exposed to high amounts of solar radiation (high multiannual mean value), evaporation fluctuations through the year did not show much relevance to the system, however precipitation showed the highest variability among the variables and therefore should be included as a time series in the system.

4.1.3 Hydrological variables

During the analysis of the river ﬂows, it could be observed that all the gauging stations from the national hydrometric network are located at mid catchment level (very far away from the river mouths into the system), and between them and CGSM stand the larger arable land and crops extensions in the ecological region. Therefore, the eﬀect of water extraction and intensive agriculture would not be taken into account if the model were to be fed with the raw data provided by IDEAM, overestimating the freshwater inputs to the system. In order to address this issue, a simple reduction factor approach was proposed:

Qmouth = (1 − kr )Qmeas (4.1) where kr is the flow reduction factor defined as the percentage difference between the extracted water from water rights and the returned water from discharges. In order to apply this procedure, information about water intakes and discharges along the stream is needed, which in the case of CGSM could only be found for Frio river which crosses the largest plantations around the system

41 before becoming a tributary for Sevilla river, and where kr is in the order of 55% until a maximum of 5.1 m3 s−1. Final ﬂow time series analyzed in a similar fashion to meteorological data can be found in Table 4.3.

Table 4.3: Mean monthly river ﬂow time series contributing to CGSM-LC along the East bank

River JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DIC Fundación 16.200 13.730 13.760 19.550 29.150 28.330 22.520 25.470 32.020 38.780 39.740 24.960 Aracataca 9.355 7.100 6.557 10.520 20.270 19.940 14.3500 17.080 24.000 29.780 26.100 16.29 Sevilla 9.325 7.014 6.463 8.784 17.602 26.001 26.470 29.545 40.664 45.727 35.684 16.126

The two creeks flowing into the system in the Northwestern corner of CGSM and West from Pajarales are ungauged and therefore very little observed flow data was available in the literature. Most previous modelling initiatives in the complex have assumed the design flows presented in Table 2.2 as representative for these streams, however Deeb Sossa S en C(1993) indicates that these flows are associated to a flow of 9300 m3 s−1 in Magdalena river, which are the normal conditions for November-December (largest wet season in the year) and therefore are not applicable to the rest of the year. In order to verify the span of the flows in the creeks, the author and personnel from INVEMAR performed a joint hydrometric field survey during April 2015 (Fig. 4.6) to get an idea of the low flow conditions in these tributaries.

Fig. 4.6: Hydrometric survey at the creeks connecting Magdalena river and the model domain on April of 2015. Left: measurements at Clarín creek, right: Aguas negras creek .

Results from the field survey showed extremely low flow values of the creeks into the system: 4.166 m3 s−1 for Aguas Negras and 0.616 m3 s−1 for Clarín creek. This was no surprise since, according to IDEAM’s emergency bulletins, during 2014 and 2015 the region has been experiencing a very strong drought due to an ENSO phenomena with a rain anomaly of −59.9 %. Even when adjusting the values using the rain anomaly in the actual ENSO conditions, this would mean that Aguas Negras creek would be discharging around 10.4 m3 s−1 and Clarín around 1.54 m3 s−1 by the end of the dry season on a standard year. These values are very different from the respective

42 design ﬂows of 60 m3 s−1 and 20 m3 s−1 that have been used as constants in previous modelling initiatives in the area. As the previous result is very interesting for the study of worse case scenarios in the system, due to previous experiences regarding hypersalinization, the measured ﬂows during the hydrometric cam- paing were used as the constant boundary conditions for the creeks as worst case scenario for the system.

4.1.4 Tide verification Fit of the tidal fluctuations inside the model domain, generated by the imposed in the open boundary conditions from FES2004 global model dataset, were cross validated against a one year observed level time series from Santa Marta gauging station (01/Jan/2014 to 01/Jan/2015). In order to transform the absolute values provided by the gauging station (including the depth of the sensor) into relative levels that could be compared to the simulated series and to remove any seasonal effects during the measurement period, a moving average using different windows was subtracted from the original series (Fig. 4.7).

Annual mean Monthly mean

0.2 0.2

0 0

Relative Level [m] Relative Level [m] Observed -0.2 -0.2 FES2004

Jun 02 Jun 07 Jun 12 Jun 17 Jun 22 Jun 27 Jun 02 Jun 07 Jun 12 Jun 17 Jun 22 Jun 27 Date 2014 Date 2014 Daily mean Semidiurnal mean 0.3 0.3

0.2 0.2

0.1 0.1

0 0

Relative Level [m] -0.1 Relative Level [m] -0.1

-0.2 -0.2 Jun 02 Jun 07 Jun 12 Jun 17 Jun 22 Jun 27 Jun 02 Jun 07 Jun 12 Jun 17 Jun 22 Jun 27 Date 2014 Date 2014 Fig. 4.7: Veriﬁcation of tides from FES2004 global model versus observed data in Santa Marta station using diverse moving average windows. June-July period is presented as an example of the ﬁt for enhanced visualization of the patterns.

From the results, it can be observed that the selected averaging window has an important effect in the fit between observed and simulated series, and that the closest this parameter gets to the major tidal constituents of the system (M2 for the region, see AppendixB) a better fit can be achieved. Therefore, a semidiurnal (≈ 12 h) moving average window is selected for comparison (enlarged in Fig. 4.8). Even though there is a good overall correspondence between simulated and observed series, some underestimation of the secondary peaks can still be observed, this effect could be caused either by a lower amplitude for the specific constituent in FES2004 or by the effect of atmospheric tides not

43 0.25

Observed 0.2 FES2004

0.15

0.1

0.05

0 Relative Level [m]

-0.05

-0.1

-0.15 Jun 02 Jun 04 Jun 06 Jun 08 Jun 10 Jun 12 Jun 14 Jun 16 Jun 18 Jun 20 Jun 22 Jun 24 Jun 26 Jun 28 Jun 30 Date 2014 Fig. 4.8: Veriﬁcation of tides from FES2004 global model versus observed data in Santa Marta station using semidiurnal moving average. June-July period is presented as an example included in the forcing of the model (the observed series corresponds to free surface levels, not only tidal components).

4.2 Stability and convergence

MOHID uses the a semi-implicit version of the Alternating Direction Implicit (ADI) method to solve the differential equations for advective quantities. As the model uses a semi-implicit instead of a fully explicit scheme, the Courant-Friedrichs-Lewy stability criteria (CFL) can be more than one without limiting the operation, being a general recommendation from the developers to try to attain a CFL lower than 6 in most cases and at most 8 for acceptable results (personal communication with Luis Fernandes). Initial tests with the rectangular grid allowed the use of time steps from 30 s to 60 s, however for the variable grid an adaptative time step was used with initial time steps from 5 s to 10 s (CFL: 13.1) and afterwards from 2 s to 5 s (CFL: 5.2). The latter configuration was the one finally used for the results presented in this report. The use of local refinements in the system straits could have improved model stability without such small time steps as the ones used, the closest mechanism to achieve this in structured grids would have been through nested grids. Even though MOHID is capable of creating nested grids, this configuration was not used in this project while the communication between a father/son model is unidirectional in structured grid approaches, therefore, in order to accurately simulate the processes in Boca de la Barra, a minimum grid spacing of 50 m would have to be imposed in the whole estuarine area of the model domain. This condition is restrictive in the light of the current lack of detailed bathymetries for Pajarales and Caño Grande sectors. In case a nested grid could be implemented, Camacho(1991) mentioned that they are highly susceptible to numerical instabilities if the resolution gradient in the father-son boundary is too sharp.

44 In order to avoid this issue, the following design criteria are to be followed (Personal communication with Luis Fernandes, ActionModulers Inc.):

• The scaling between father and son grids should not be higher than 1:3.

• A buﬀer zone of at least 10 cells of the son domain should be left between the father-son boundary and the area of interest where detailed results are desired. This will allow a smooth numerical transference between the models.

Provision of an appropriate spin-up period for model convergence in the initial stages of the simulation is necessary to allow for the connection of the forcing terms and the results to reach an statistical equilibrium around an acceptable level of relative error, with this, large ﬂuctuations or no physical meaning of the output values can be avoided. As large residence times inside the system were to be expected, spin-up periods of 15 d and 30 d were tested for convergence in several areas of the system (Fig. 4.9).

Fig. 4.9: Spin-up periods tested for CGSM-LC in diﬀerent sectors. The system is considered to have converged when water levels ﬂuctuations stabilize around zero and both graphs coincide (vertical red line).

It can be concluded that convergence evolves spatially following the path from the tidal forcing from the open ocean through the inlet (convergence achieved after 4 d) into the main lagoon (5 d for equilibrium in the center and 6 d for the Southwestern corner) and ﬁnally into Pajarales system (convergence reached after 8 d). From these results, the simulations could be started with a spin- up period of approximately 8 d in order for the complete system to reach convergence, however a spin-up period of around 15 d for the area was ﬁnally selected in order to give a security factor and assure the reliability of the numerical results. A similar time was found by Betancur-Pérez(2014) when modelling a large water body (Riogrande II) reservoir in Colombia using MIKE3 and DELFT3D models.

45 4.3 Temporal patterns of hydrodynamic variables

4.3.1 Water level

Level fluctuations inside the system show a clear cyclic pattern responding to tidal currents generated in La Barra. However, due to the already low tidal amplitude in the region and the strong tidal modulation when entering the lagoon, the changes in water level are very small in the estuarine area, with relative daily fluctuations of around 4 cm in CGSM and 3 cm in Pajarales during the same month. Taking this into account, water levels inside the lagoon complex would not be good calibration parameters for future model optimization or new modelling initiatives in the region, while they don’t present a large enough fluctuation that allow to segregate between particular processes in the system. From the yearly pattern in Fig. 4.10, it can be observed that even though tides are the main source of variability in short periods (i.e. days or hours), other phenomena are also influencing the water levels in the long term dynamics of the system. Changes in mean relative water level inside the estuary are not drastic (1 cm to 2 cm) but show a marked seasonal pattern of higher waters during the wet season, when the system receives a larger volume of freshwater via its tributaries and the largest precipitations of the year.

Fig. 4.10: Water levels in the central areas of the two main lagoons of CGSM-LC during a one year simulation.

In Fig. 4.11 the seasonal spatial distribution of water column depths between shallow waters and the 5 m isobath is presented. Water level oscillations cannot be easily perceived in the presented seasonal maps due to the ﬂuctu- ations’ small magnitude regarding the total water column of the lagoons (between 1 m and 2.5 m) .

46 Fig. 4.11: Water level during a generic year in CGSM-LC between isobaths 0 m to 5 m

4.3.2 Water velocity

The highest water velocities in the system were found in the main constriction at La Barra inlet, in accordance to the previous modelling studies realized in the complex. Maximum horizontal velocity values were observed during the flood tide with a magnitude of around 1.1 m s−1, which is consistent with results from previous modelling projects in the area and available descriptions for the zone. In the rest of the lagoons, very slow horizontal flows are dominant, with magnitudes of less than 5 cm s−1 and local gradients in the river mouths and Caño Grande. Fig. 4.12 shows the flow patterns in La Barra strait for the largest flood and ebb currents during a tidal cycle in June as an example of behavior at the transition period. From the spatial velocity distribution at La Barra it is evident that, during high and low tides, maximum velocities are concentrated in the narrowest parts of the strait (approximately midway between the sea and the lagoon), this area is located nearby the foundations of a highway bridge that crosses the strait and directly below an elevated gas pipe crossing. Due to these high velocities, bed shear stresses are important in this area, causing a localized deepening of around 7 m that was detailed in the bathymetric analyses. The mentioned deepening is used as a traditional hotspot for fishermen while during flood and ebb events, fish density in the area surrounding the pit increases

47 1.64 Velocity [m/s] 1.07 0.50 Fig. 4.12: Vectorial velocity maps in La Barra inlet for ﬂood (left) and ebb (right) conditions. dramatically. An interesting feature that was observed in Fig. 4.12 is the generation of local eddies nearby the estuary mouth both in the outer coastal area (during ebb) and near the entrance channel to Sevillano lagoon (during ﬂood), so far no mention of such phenomena had been found in the existing literature for the complex and further discussion with the inhabitants and experts could be useful to either discard the observation as a numerical phenomena or further analyze it on the light of this project results.

4.4 Flow patterns in the system

From the bathymetric contours and sediment distribution inside the main lagoon, a flushing path along the East margin was to be expected, transporting the waters fed by the rivers draining on this area of the complex and depositing their sediments along their path from their specific mouth to the ocean inlet. However, flow patterns inside the system were not as straightforward as expected due to the important velocity gradient between slow flows inside the lagoons and the comparably high velocities during the tide reversals in La Barra. Several microscale phenomena were observed during the long term simulations like the evident eddy registered during April in Fig. 4.13 This microscale phenomena are not permanent in the system and can completely disappear during some climatic periods where new or more intense forcings are imposed to CGSM-LC. This is the case of the aforementioned eddy near El Boquerón island, which is observed very seldomly during the wet season (Fig. 4.14) compared to the dry season patterns. Even though the magnitude of these events is not high as to directly alter the maritime traffic inside the complex or the daily life of the families dwelling in the banks, they might have a notable influence in the biological processes inside the system, like fish banks daily movement patterns or areas of preferable benthic settlements. Moreover, this additional turbulence elements can influence other

48 Fig. 4.13: Flow patterns during an Ebb-Flood cycle in CGSM-LC during April (dry season). estuarine processes like oxygen surface or sediment interface ﬂuxes. The system’s long term ﬂow patterns can be summarized as follows:

• During the wet season (September-November), due to the increase in rain and consequent larger ﬂows from the rivers feeding the system, as well as the shift of the winds from the North-East to the South and South-West, ﬂuxes from CGSM to the sea tend to increase (Posada et al., 2009).

• As Pajaral lagoon shows little influence from the tidal forces exerted in La Barra and receives inputs from two high capacity creeks in its West margin, it is permanently transferring water to the CGSM with very seldom direction reversals. Changes of circularion flor inside this system are only evident when the magnitude of flood current entering the system are relatively large, which is typical during the dry season.

As it was previously mentioned in this report, one of the most important drivers to the system flow patterns is the wind. Most of the year wind influences the complex from a NE or N direction, coinciding with the preferential flow path from the ocean into the lagoon complex. Adding the fact that CGSM-LC is a very shallow water body there can be a high probability of a "water entrapment" phenomena in the opposing side of the main lagoon and in Pajarales complex, theory that seems to be backed by some flow reversal areas and eddies West from Fundación river mouth.

49 Fig. 4.14: Flow patterns during an Ebb-Flood cycle in CGSM-LC during October (wet season).

4.5 Residence times

A Lagrangian tracer approach was used in order to calculate residence times as the time required for an initial volume of water (represented as discrete water parcels) to leave a defined zones of the complex. Fig. 4.15 shows the evolution of Lagrangian tracers remaining in each water body in time during a residence time experiment, in this case CGSM was fed with 1300 non reactive tracers (simulating water parcels of 500 000 m3) while Pajarales was fed with 915 non reactive tracers (50 000 m3 water parcels). From the results, it can seen that the whole system has a very weak flushing dynamics, however, CGSM has a much higher exchange gradient than Pajarales lagoon which barely transfers any water during the simulation. Results from Lagrangian tracers simulations seem to confirm the flow patterns mentioned in 4.4, showing that most of the water renewal occurs over the East margin while North-West zones tend to stay still and accumulate parcels from other areas like Pajarales complex (Fig. 4.15). However, there’s no evidence of a long term water entrapment in the Southwest margin as suggested in the previous section. CGSM presents really high retention times, taking almost 7 months to renew half of its original

50 1.1

0.9

0.8

0.7

0.6 Retained volume fraction

0.5 CGSM Pajarales 0.4 May Jun Jul Aug Sep Oct Nov Dec Date

Fig. 4.15: Lagrangian tracer distribution and remaining volumes for residence time calculation. Top left: Particle distribution at the start of the simulation, top right: 8 months after, bottom: Volume retention curves volume. According to the adopted definition for flushing time in this report, the final residence time for the large lagoon could be extrapolated from the results to around one year. This result indicates that most pollutants entering the system will remain in it enough time to be converted via biological or chemical processes. Furthermore, this low flushing dynamics makes CGSM very susceptible to the accumulation of recalcitrant pollutants like heavy metals. The main factors influencing the retention times in the system seem to be the freshwater flows and wind direction, most of the year the wind blows from the North or North-East, in a direction opposed to the flushing path into the sea. Therefore, water masses tend to get retained in the South and North-West shores of the system most of the year. It is important to notice that the flushing times calculation does not include the additional water surface flux due to evaporation, which might not be important in temperate climates but is very strong in the area of study, further inclusion of this parameter in future analysis might shed some light over the water exchange dynamics in Pajarales lagoon..

4.6 Hydrodynamic calibration

Calibration of hydrodynamics in the system with the available ADCP velocity measurements at La Barra did not perform well in terms of ﬁt. This happened most probably due to the enormous diﬀerence in scales between the observed dataset (minute data) and the modelling approach used in the simulations by variating the focing functions in a monthly basis.

51 Results for calibration test for horizontal eddy viscosity (NH ) are presented in Fig. 4.16, while calibration results using bottom roughness coeﬃcient can be found in Fig. 4.17.

Fig. 4.16: Horizontal eddy viscosity (NH ) calibration test keeping rugosity coeﬃcient constant (CZ =125).

According to Martin and McCutcheon(1999), horizontal roughness coefficient in estuarine systems can fluctuate between 10−2 and 102, moreover, according to MOHID developer team, this parameter should be around 10 % of the grid size (personal communication). Therefore values between 1 and 100 were used for the calibration. According to the model results, velocities in La Barra are better represented with a high horizontal eddy viscosity, however not even the maximum value of 100 properly fitted the observed ADCP measurements. One way to further analyze the issue was to implement Smagorinsky’s horizontal eddy viscosity model (AppendixA, Equation A.3) to calculate a spatially distributed parameter based on velocity gradients between grid elements. With this approach, results of approximately 50 were found for La Barra, however, in the slow moving estuarine zone, this value dropped to 1 or even less. Moreover, results with the Smagorinsky method yielded much higher velocities than the ones expected according to the observed data. Bottom roughness values were obtained from previous modelling studies in the area (Camacho, 1991; Flórez, 2013), traditional roughness coefficient tables from hydraulics literature and from a review document of MIKE21 model for the selection of roughness coefficients. According to the calibration results, simulated velocities approach better the field values by consid- ering a rougher bed, this agrees with the sedimentological surveys done in the system where coarser material was to be expected in the inlet due to tidal action. However, even the typical values for sandy beds show scarce fit with the observed velocities. By further analyzing the field measurements used as reference values for the calibration, it can be observed that it does not reflect the characteristic mixed semidiurnal tidal pattern from the study area. Therefore, further detailed revision of the data consistency and calibration test should be realized.

52 Fig. 4.17: Bottom roughness coeﬃcien (Chèzy) calibration test. Horizontal eddy viscosity was calculated for all test via a Smagorinsky model

By addressing a side note by Ascione Kenov et al.(2014), for this kind of validations, it is important to take into account that discrepancies between the spatio-temporal resolutions of observed and simulated dataset are to be expected in data scarce highly dynamic systems, therefore comparison of these sets must usually be addressed in a qualitative rather than quantitative basis.

4.7 Transport of conservative substances

Salinity is commonly used in hydrodynamic simulations as a reference conservative constituent for estuarine and oceanic systems. Due to the conflicting results in the hydrodynamic calibration section, the analysis of this variable was used to further verify whether the model was simulating the seasonal patterns in a generic year satisfactorily. Salinity patterns for specially relevant moments of the year are presented in Fig. 4.18. By analyzing the previous maps, an good congruence between the model results and the theoretical behavior of the system can be observed: During the larger dry period (February) the system is "resalinizing" after the strong rains from the previous rainy period, therefore a progressive salt wedge from the inlet can be observed, it is spected for the saline wedge to cover most of the larger lagoon by the end of the dry period (April) when the highest values in all stations are registered according to INVEMAR’s monitoring program. During the small rainy season, a new freshwater wedge can be observed extending from the South of the system due to river runoff, by July both fronts are in apparent equilibrium in the system until a new rainy season disrupts it and impose complete freshwater conditions in most of the estuary. It is important to observe that Pajarales lagoon is not very affected by the aforementioned yearly process and it’s salinity variations are therefore more influenced by the inflow from Aguas Negras creek and the high surface water flux due to strong evaporation potential and very still hydrodynamics in the lagoon (discussed previously in section 4.5).

53 JUNE 1

Fig. 4.18: Temporal patterns of salinity during a generic year in CGSM-LC. February reﬂects the conditions during the larger dry period, June the end of the small rainy period, July the small dry period (Veranillo de San Juan) and October the major rainy season in the region.

In order to further asses if these values really reflect the salinity patterns observed in CGSM-LC, comparison between simulated results and observed monthly mean multiannual salinities for 4 representative stations in the system was performed. Selected stations were La Barra (North), Centro de la Ciénaga (midpoint), Frente a Fundación (Southwest) and Nueva Venecia (Pajarales). Results from this comparison are summarized in Fig. 4.19. Fig. 4.19 shows a very good fit of the stations located inside CGSM main lagoon for the wet and transition periods, however, a gap starts with the strong dry period at the end of the year where the model shows a progressive divergence from observed values and a consequent lag in the "resalinization" process. The main cause of this behavior can be the overestimation of the river inflows discussed previously on this chapter, adding too much water even during the dry season and slowing the transition back to salt water conditions. In the case of La Barra inlet, general patterns are correctly represented but magnitudes of the salt concentration are overestimated. By analyzing the whole domain, it was discovered that Sevillano lagoon (South from the inlet) is being modelled as a Hypersaline zone due to the very low water depth imposed during the bathymetry definition (no data were available for it) and the high evapo-

54 Fig. 4.19: Comparison of observed and simulated salinity values in different stations across the estuarine area Blue: Observed mean multiannual data, red: Model results transpiration. As this lagoon connects directly to the inlet, in each ebb tide this hypersaline water mixes with the water in La Barra, influencing the salt levels in the latter. The last panel presents the conditions at Pajarales lagoon, showing that, in its current state inside, the model cannot simulate the transport of conservative substances for this water body. This was expected since there’s no seasonal information available regarding flows from Aguas Negras creek into the lagoon, which is the main driver of this system.

4.8 Nutrient and oxygen transport

As conservative substance transport described in the previous section showed to be congruent with the general circulation patterns in CGSM, the next and ﬁnal step was to integrate the nutrient and oxygen transport model into the modelling system and obtain the general daily patterns of nutrient and oxygen production/destruction dynamics inside the estuary. In the case of nutrient and oxygen, Tuchkovenko and Calero(2003) indicated that in such a productive ecosystem as CGSM-LC, nutrient and oxygen patterns show a short term (hourly to daily) variability instead of the commonly reported long term changes described for temperate climates, this mainly due to the lack of climatic seasons in the Colombian Caribbean. Accordingly, simulations for these components will be performed only for short term simulations (i.e. 5 d). Results are presented for the dry and wet periods in the station located at the center of the main lagoon (Fig. 4.20 and Fig. 4.21 respectively) in order to check for diﬀerences between the water quality daily patterns in each climatic conditions. As it was previously mentioned in section 3.4, simulations of the nutrient and oxygen model will include phytoplankton as the largest biological compartment in the system and zooplankton in order to induce the predatory dynamics for controlling the phytoplankton growth. There are several

55 other biological components in such a rich ecosystem as CGSM-LC like Diatoms, bacteria, cilliates, macrofauna and so, however as no detailed information about this trophic levels and their dynamics for the system were available, the simplified approach using only phytoplankton and zooplankton seemed appropriate as a preliminar version of the model. As MOHID allows for a different time step for the biochemical model in the WaterQuality module, simulations were held in 900 min time steps in order to maintain the stability of the system but reduce the computational effort needed.

Fig. 4.20: Preliminar 0-D model results for nutrients and oxygen in CGSM-LC during dry season.

From the dry weather period results, a cyclic biological production/consumption pattern following the day/night dynamics is easily identified. From the graphical result of this model configuration, inorganic phosphorus appears to be the main nutrient used in the phytoplankton growth dynamics, with peak values during nightime (due to accumulation from the river loads) and preferential consumption during sun hours by the phytoplankton with the consequent oxygen production. In these conditions, a mean oxygen level fluctuating around 5.5 mg L−1 seems to be attained. During the rainy season, the same daily dynamics described for the dry season is maintained with phosphorus consumption and oxygen production during daytime by phytoplankton, the main difference in this period is a higher accumulation of Ammonia in the system due to higher river loads and a higher steady state oxygen concentration around 6 mg L−1. Both results (wet and dry conditions) show a common pattern for nitrogen dynamics with a clear Ammonia accumulation inside the system and no apparent nitrite or nitrate production. By analyzing this trend it can be concluded that the system is not reflecting the nitrification/denitrification processes adequately due to the lack of the bacteria component in the current model configuration (Ammonia should also follow a marked daily consumption pattern). Future work will need to include this component because a permanent Ammonia accumulation as the one currently ongoing will end up inhibiting phytoplankton growth at some point of the simulations.

56 Fig. 4.21: Preliminar 0-D model results for nutrients and oxygen in CGSM-LC during wet season.

The observed patterns reproduce very good the results obtained by Tuchkovenko and Calero(2003) for a similar 0-D modelling approach, except for the mentioned issue with nitrogen cycle and the bacterial component. Once this issue is addressed, calibration and validation against INVEMAR’s monitoring programs databases is still to be performed via long term simulations like the ones for conservative substances.

57 This page intentionally left blank.

58 CHAPTER FIVE

CONCLUSIONS AND RECOMMENDATIONS

As in any modelling project, the first aspect that has to be emphasized during these conclusions is that no model can reply flawlessly the processes occurring in a natural system. Models provide mathematical representations of the phenomena based on the information given by the users and therefore, they can be just as accurate as the quality and availability of the input data allows them to be. The model developed during this project is able to adequately simulate the hydrodynamic, conservative substances and nutrient/transport dynamics of CGSM-LC in the seasonal resolution, with further adjusts to be made in order to better reproduce the velocity fields and nitrogen cycles. Difficulties arisen during the calibration phase and further work is needed before a full implementation at INVEMAR is to be attained, however, this initial model constitutes a big step in the implementation of decision support tools for coastal resource management in the Colombian Caribbean and opens the path for further projects in other sensitive ecosystems of the region. CGSM-LC is a well documented system due to its ecological relevance and previous environmental crises suffered, therefore there is a large amount of available references for environmental variables and biological processes that greatly help with the setup of the modelling project. Nevertheless, the greatest difficulty when trying to model this system was precisely the lack of available information pertaining hydrodynamic variables like water levels, velocities or water fluxes, which impeded a good calibration of the hydrodynamic compartment to be performed. In order to improve the model and deploy it for real consultancy projects or decision making in the area, INVEMAR will have to plan some field surveys to capture real data on hydrodynamic behavior, not only in La Barra inlet but also in the inner waters of the lagoon for both climatic conditions. With these data, model parameters can be adjusted to the real field conditions in the complex. In the long term, the proposed hydrodynamic surveys of the system should be hold in a regular basis for quality control of the model performance, therefore it could be necessary to include them in the protocols of the ongoing water quality monitoring program in the area. According to the results from the field survey in April 2015 and the poor fit of the results in Pajarales lagoon, it is evident that the temporal variability of flows through the creeks connecting Magdalena river and CGSM-LC cannot be overlooked during modelling studies and that the use of design values

59 for them conduces only to biased results tending to non realistic freshwater conditions in the system. Therefore, the author will strongly recommend the gauging of the three major creeks draining into the system along the West margin, or at least carrying out flow measurements in these systems during both climatic conditions each year in order to have realistic input data for the model. A considerable amount of time in this project was used to obtain meteorological and oceanographic data from separate institutions in Colombia that are part of the same environmental information system, each one of them with different (not always clear) data acquisition procedures. It would be highly recommendable in order to simplify administrative procedures in this type of projects, to integrate the environmental information collected from all the institutions involved in the monitoring of the complex in the same place, so researchers can know from the start of the initiatives which information exists and which should be included in the project activities. The resolution of all the implemented grids is too coarse to simulate small scale and highly variable processes like oxygen deficit in intertidal areas, which, according to Nobre et al.(2005), is one of the key drivers of nutrient related macroalgal blooms. Therefore, future expansions of the proposed solution could consider methods to analyze intertidal and periodically flooded areas of the complex (e.g. small-scale tide pool models), as well as grid resolutions smaller than 100 m for these areas. MOHID interface, as well as all the structured grid models available in the market, has the drawback that variable grids cannot be refined in specific areas of the domain (e.g. Boca de la Barra) without affecting the whole rows/columns including the elements in that zone. This generates issues with numerical stability (i.e. high CFL) while the extension of the refinement to open sea waters could create small resolution cells with high depths if a pure 2D depth-averaged approach is used. As this was the case for the model domain, further research on the implementation of unstructured grids in order to reduce the timestep of the simulations could be desirable. The initial hypothesis that mean monthly multiannual values for the boundary conditions would represent the general trends of the system for a generic year hold up well enough in the case of conservative substances and nutrients in CGSM-LC. Even though information on seasonal patterns in a generic year is of high relevance for system parametrization, the users also can be interested in using the model to evaluate point processes or events in shorter time resolutions. MOHID model allows the user to easily change the time series input from monthly to daily or hourly data, but the main concern in the study area is to ensure environmental data completeness and consistency, which in this project could was evidently an issue. Therefore, when starting modelling initiatives in the area, previous discussion with the institutions in charge of measuring atmospheric and environmental variables should be hold in order to stress the negative impact of missing data in the expected results and achieve commitments regarding data integrity and completeness. As the area is affected directly by ENSO, running the model with detailed atmospheric and streamflow data at daily level would be preferable for detailed studies. Specially if the objective is to simulate short term events during particularly influenced weather conditions like the ones present during 2014 and 2015 in the area (strong ENSO event). Current databases and research regarding nutrient behavior in estuaries is heavily biased towards temperate systems (Statham, 2012), therefore, little information is available regarding nutrient and

60 oxygen rates in tropical estuaries like CGSM-LC. More studies related with biochemical cycles and biological reaction rates in the system would help to greatly improve this model’s accuracy. As information regarding water extraction and discharges in the rivers East of the system could just be found for Sevilla river, only this stream could be corrected during its transit through the agricultural lands, therefore, the effect of freshwater input from the East bank into the complex can be overestimated, this being one of the possible causes for the lack of fit in salinity distributions during the large dry climatic period. Accordingly, either closer to the discharge flow monitoring stations should be installed or detailed studies regarding water abstraction/discharge in the arable lands nearby CGSM are needed. One interesting approach to address this problem could be the implementation of hydrological models like SWAT to model water and nutrietnts dynamics in these areas and couple them to MOHID model (integration routines are available from MOHID’s webpage) in order to obtain a continous catchment-to-coast modelling scheme. In order to obtain a better nutrient and oxygen model for the system, the inclusion of the bacterial component in the model configuration is necessary. A study for the throughout description of nitrifying/denitrifying bacteria inside the system would be important for future model optimization. Computational effort for long simulations like the ones needed for this project makes it difficult to implement them on regular desktop or laptop computers. Even though the simulations ran at rates of 0.1 s/iteration to 0.4 s/iteration (for 5 seconds iterations), more than 250 hours of computer effort were needed for a one year simulation in the lagoon complex (without the nutrients/oxygen transport). In order to optimize computer and personal efforts, as well as to give some flexibility to run multiple simulations at the same time, it is highly recommended that INVEMAR dedicates and configures a server terminal with at least 8 cores and 10 GB RAM for modelling purposes, preferably with remote access for the researchers involved in modelling projects in the near future.

61 This page intentionally left blank.

62 BIBLIOGRAPHY

Aguilera, M. (2011). Habitantes del agua: El complejo lagunar de la Ciénaga Grande de Santa Marta. Technical Report 144, Banco de la República, Cartagena, Colombia. Documentos de trabajo sobre economía regional.3,5

Ascione Kenov, I., Campuzano, F. J., Franz, G., Fernandes, R., Viegas, C., Sobrinho, J., de Pablo, H., Amaral, A., Pinto, L., Mateus, M. D., and Neves, R. (2014). Advances in modelling of water quality in estuaries. In Finkl, C. W. and Makowski, C., editors, Remote sensing and modelling: Advances in coastal and marine resources, volume 9 of Coastal Research Library, chapter 10, page 494. Springer International Publishing. 25, 26, 27, 53

Bastidas-Salamanca, M., Vivas-Aguas, L. J., García-Garay, J. G., Pierini, J., and Acosta, J. (2012). Convenio Interadministrativo MADS - INVEMAR No. 33: Guía Nacional de Modelación de Calidad de Aguas Marinas. Informe Técnico Final. Technical report, INVEMAR - Programa GEO, Santa Marta, Colombia. 24

Bernal, G. and Betancur, J. (1996). Sedimentología de Lagunas Costeras: Ciénaga Grande de Santa Marta y Ciénaga de Pajarales. Boletín de Investigaciones Marinas y Costeras, (25):49–76. 12, 14, 15, 35

Betancourt, J., Parra, J., and Villamil, C. (2013). Emisión de metáno y óxido nitroso de los sedimentos de manglar de la Ciénaga Grande de Santa Marta, caribe Colombiano. Boletín de Investigaciones Marinas y Costeras, 42(1):131–152. 15

Betancur-Pérez, G. (2014). Metodología para la Selección de Modelos Hidrodinámicos Tridimen- sionales. Master thesis, Universidad Nacional de Colombia - Sede Medellín. 45

Blanco, J., Viloria, E., and Narváez B., J. (2006). ENSO and salinity changes in the Ciénaga Grande de Santa Marta coastal lagoon system, Colombian Caribbean. Estuarine, Coastal and Shelf Science, 66(1-2):157–167. 16, 17, 33

Botero, L. and Mancera, J. (1996). Síntesis de los cambios en los últimos 40 años de CGSM. Revista Academia Colombiana de Ciencias Exactas, 20(78):465–474.5

Braunschweig, F., Martins, F., Chambel, P., and Neves, R. (2003). A methodology to estimate renewal time scales in estuaries: the Tagus Estuary case. Ocean Dynamics, 53(3):137–145. 34

63 Brothers, S. M., Hilt, S., Meyer, S., and Köhler, J. (2013). Plant community structure determines primary productivity in shallow, eutrophic lakes. Freshwater Biology, 58(11):2264–2276.2

Bula-Meyer, G. and Díaz-Pulido, G. (1995). Macroalgas del banco de las ánimas y nuevos reg- istros para el caribe colombiano. Anales del Instituto de Investigaciones Marinas de Punta Betín, (24):173–183. 14

Camacho, L. A. (1991). Implementación de un modelo hidrodinámico y de transporte de contaminantes bi-dimensional y su aplicación al caso de la Ciénaga Grande de Santa Marta. Master thesis, Universidad de los Andes, Bogotá.9, 10, 27, 28, 35, 44, 52

CIOH, INVEMAR, and CORPAMAG (2001). Mapa batimétrico Complejo Lagunar Ciénaga Grande de Santa Marta. 12, 30

Cloern, J. E., Foster, S. Q., and Kleckner, a. E. (2014). Phytoplankton primary production in the world’s estuarine-coastal ecosystems. Biogeosciences, 11(9):2477–2501.2

Deeb Sossa S en C (1993). Plan de recuperación del complejo lagunar de la Ciénaga Grande de Santa Marta - Diseño Obras Hidráulicas. Technical report, DNP, CORPAMAG, GTZ, Santa Marta, Colombia.XIII,2,3,4, 11, 12, 17, 42

Espinosa, L. F., Parra, J. P., and Villamil, C. (2011). Determinación del contenido de metales pesados en las fracciones geoquímicas del sedimento superﬁcial asociado a los manglares de la Ciénaga Grande de Santa Marta, Colombia. Boletín de Investigaciones Marinas y Costeras, 40(1):7–23. 18

Falconer, R., Lin, B., and Kasheﬁpour, S. (2005). Modelling water quality processes in estuaries. In Bates, P., Lane, S., and Ferguson, R., editors, Computational Fluid Dynamics: Applications in Environmental Hydraulics, chapter 12, pages 305–328. John Wiley & Sons, 1st edition. 71

Flórez, J. (2013). Simulación de los patrones de profundidad, velocidad y salinidad del agua en la Ciénaga Grande de Santa Marta por el método de Lattice Boltzmann. Master thesis, Universidad de los Andes, Bogotá. 10, 35, 52

Gordon, Jr., D. C., Boudreau, P. R., Mann, K. H., Ong, J.-E., Silvert, W. L., Smith, S. V., Wattayakorn, G., Wulﬀ, F., Yanagi, T., and D. C Gordon, J. (1996). LOICZ Biogeochemical Modelling Guidelines. 35

Hylin, A. (2014). Water budget estimation on a data limited wetland: The case of the Ciénaga Grande de Santa. Master’s thesis, University of Stockholm. 17

Ibarra, K. P., Gómez, M., Viloria, E., Arteaga, E., Cuadrado, I., Martínez, M., Nieto, Y., Rodríguez, J., Licero, L., Perdomo, L., Sánchez, S., Romero, J., and Rueda, M. (2015). Monitoreo de las condiciones ambientales y los cambios estructurales y funcionales de las comunidades vegetales y los recursos pesqueros durante la rehabilitación de la Ciénaga Grande de Santa Marta. Technical report, INVEMAR, Santa Marta, Colombia.1,3,4, 19

64 Kämpf, J. (2010). Ocean Modelling for Beginners. Springer Berlin Heidelberg, Berlin, Heidelberg. 72

Lozano, J. C. (2003). Desarrollo de un modelo bidimensional hidrodinámico y de calidad en cuerpos de agua poco profundos. Master thesis, Universidad de los Andes, Bogotá.9

Lyard, F., Lefevre, F., Letellier, T., and Francis, O. (2006). Modelling the global ocean tides: modern insights from FES2004. Ocean Dynamics, 56(5-6):394–415. 31

Mancera, J. and Vidal, L. (1994). Florecimiento de microalgas relacionado con mortandad masiva de peces en el complejo lagunar Ciénaga Grande de Santa Marta, Caribe Colombiano. Anuario del Instituto de Investigaciones Marinas Punta de Betín, (23):103–117.6

Martin, J. and McCutcheon, S. (1999). Hydrodynamics and transport for water quality modeling. CRC Press, 1st edition. 38, 52, 72

Neves, R. (2007). Numerical models as decision support tools in coastal areas. In Gonenc, I., Koutitonsky, V., Rashleigh, B., Ambrose, R., and Wolﬂin, J., editors, NATO Securiy through Science Series: Assessment of the Fate and Eﬀects of Toxic Agents on Water Resources, chapter 8, pages 171–195. Springer.6,9, 23

Nobre, A., Ferreira, J., Newton, A., Simas, T., Icely, J., and Neves, R. (2005). Management of coastal eutrophication: Integration of ﬁeld data, ecosystem-scale simulations and screening models. Journal of Marine Systems, 56(3-4):375–390. 60

Ortiz, J. (2007). Huracanes y tormentas tropicales en el Mar Caribe colombiano desde 1900. Boletín Cientíﬁco CIOH, (25):54–60. 16

Ortiz, J. (2012). Exposure of the Colombian Caribbean coast, including San Andrés Island, to tropical storms and hurricanes, 1900-2010. Natural Hazards, 61(2):815–827. 16

Ortiz, J., Otero, L., Restrepo, J., Ruíz, J., and Cadena, M. (2013). Characterization and eﬀects of cold fronts in the Colombian Caribbean Coast and their relationship to extreme wave events. Natural Hazards and Earth System Sciences Discussions, 1(4):3659–3687. 16, 33

Parra, J. P. and Espinosa, L. F. (2007). Acumulación de Pb, Cd y Zn en sedimentos asociados a Rhizophora Mangle, en el Río Sevilla, Ciénaga Grande de Santa Marta, Colombia. Revista Academia Colombiana de Ciencias Exactas, 31(120):347–354. 18

Pawlowicz, R., Pawlowicz, R., Beardsley, R. C., Beardsley, R., Lentz, S., and Lentz, S. (2002). Classical tidal harmonic analysis including error estimates in MATLAB using T_TIDE. Computers & Geosciences, 28(8):929–937. 31

Polanía, J., Santos Matínez, A., Mancera Pineda, J., and Botero Arboleda, L. (2001). The Coastal Lagoon Cienaga Grande de Santa Marta, Colombia. Ecological Studies, 144:33–45.3

65 Posada, B. and Henao, W. (2008). Diagnóstico de la erosión en la zona costera del caribe colombiano. Technical report, INVEMAR, Santa Marta. 15, 16, 17, 21, 22

Posada, B., Morales, D., Neiza, N., Idarraga-García, J., and Henao, W. (2009). Estudio batimétrico y sedimentológico de la Ciénaga Grande de Santa Marta y la plataforma somera al frente de la Barra de Salamanca (Magdalena, Colombia). Technical report, INVEMAR, Santa Marta, Colombia. 12, 13, 14, 15, 17, 19, 21, 38, 49

Rangel-Buitrago, N. G., Anfuso, G., and Williams, A. T. (2015). Coastal erosion along the Caribbean coast of Colombia: Magnitudes, causes and management. Ocean & Coastal Management, 114:129–144. 20, 22

Ruíz-Ochoa, M. and Bernal, G. (2009). Variabilidad estacional e interanual del viento en los datos del reanálisis NCEP / NCAR en la cuenca Colombia , mar Caribe. Avances de recursos hidráulicos, (20):7–20. 16

Statham, P. J. (2012). Nutrients in estuaries - an overview and the potential impacts of climate change. The Science of the total environment, 434:213–27. 27, 60

Thomas, Y., Nicolae-Lerma, A., and Posada, B. (2012). Atlas climatológico del mar Caribe colombiano. Technical report, INVEMAR, Santa Marta, Colombia. 16, 20

Toro, F. and Gómez, E. (1997). Simulación numérica del efecto del caño clarín en los patrones de circulación de la Ciénaga Grande de Santa Marta. Avances en recursos hidráulicos, (4):73–90.9

Tuchkovenko, Y. and Calero, L. (2003). Modelo numérico de calidad de aguas para la ciénaga grande de santa marta. Boletin de Investigaciones Marinas y Costeras, (32):145–167. 27, 35, 55, 57

Vilardy, S. and González, J., editors (2011). Repensando la Ciénaga: Nuevas miradas y estrategias para la sostenibilidad en la Ciénaga Grande de Santa Marta. Universidad del Magdalena and Universidad Politécnica de Madrid, Santa Marta, Colombia.4, 12, 19, 24

Vilardy, S. P., González, J. a., Martín-López, B., and Montes, C. (2011). Relationships between hydrological regime and ecosystem services supply in a Caribbean coastal wetland: a social- ecological approach. Hydrological Sciences Journal, 56(8):1423–1435. 14

Vivas-Aguas, L. J., Espinosa, L., and Parra, L. (2013). Identificación de fuentes terrestres de contaminación y cálculo de las cargas contaminantes en el área de influencia de la Ciénaga Grande de Santa Marta, Caribe Colombiano. Boletín de Investigaciones Marinas y Costeras, 42(1):7–30. 18, 19 von Erffa, A. (1973). Sedimentation, transport und erorsion an der Nordküste Kolumbiens zwischen Barranquilla und der Sierra Nevada de Santa Marta. Mitteilungen aus dem Instituto Colombo- Alemán de Investigaciones Científicas Punta de Betín, 7:155–209. 12, 13, 37, 38

66 Wiedemann, H. (1973). Reconnaisance of the Ciénaga Grande de Santa Marta, Colombia: Physical parameters and geological history. Mitteilungen aus dem Instituto Colombo-Alemán de Investiga- ciones Cientíﬁcas Punta de Betín, (7):85–119. 18, 28

Zamora, A. and Meza, A. (2013). Comportamiento de los rendimientos económicos de la pesquería artesanal de la Ciénaga Grande de Santa Marta y el Complejo de Pajarales, Caribe Colombiano. Boletín de Investigaciones Marinas y Costeras, 42(1):91–99.3

67 This page intentionally left blank.

68 Appendices

APPENDIX ONE

HYDRODYNAMIC AND CONSTITUENT TRANSPORT IN ESTUARIES

A.1 Equations for hydrodynamics in shallow waters

Free surface water movement in an incompressible ﬂuid is described by the system of 3D Navier- Stokes equations. The General conservative forms of the 3D Reynolds equations for mass and momentum conservation (in the x direction) are respectively (Falconer et al., 2005):

∂u + ∂v + ∂w = 0 (A.1) ∂x ∂y ∂z ∂u ∂u2 ∂uv ∂uw 1 ∂P ∂u0u0 ∂u0v 0 ∂u0w 0 + + + = X − − − − (A.2) ∂t ∂x ∂y ∂z |{z} ρ ∂x ∂x ∂y ∂z |{z} |{z} | {z } 4 | {z } | {z } 1 2 3 5 6 Where:

u, v, w: Velocity components in x, y and z ρ: Fluid density t: Time P: Fluid pressure X: Body forces in x direction u0u0, u0v 0, u0w 0: Reynold stresses

In the previous moment conservation equation, term 1 stands for the temporal variation of the velocity in the system (acceleration parameter), terms 2, 3 and 6 accounts for Reynold averaged viscous forces and term 5 to the pressure gradient (comprising barotropic and baroclinic components). In turbulent systems (most of the natural water bodies) the Reynold averaged viscous effects are further developed into the widely extended concepts of eddy viscosity coefficients - Nx , NY . There are many formulations to approximate the values of eddy viscosity coefficients in order to parametrize the turbulence in a moving fluid (commonly called closure problem). For horizontal movement, Smagorinsky model (Equation A.3) is one of the most extended turbulence formulations

71 while for vertical turbulence κ and κ − models are some of the key referents in the literature.

s ∂u 2 ∂v 2 1 ∂u ∂v 2 N = c ∆x∆y + + + (A.3) H s ∂x ∂y 2 ∂y ∂x Where:

NH : Horizontal eddy viscosity cs : Smagorinsky coeﬃcient (0.1-0.2)

Numerical stability in a model indicates the correspondence of the selected grid and timestep dimen- sions with the physical scale of the process to be resolved, this usually implies that mass conservation for the control volumes holds and the transport of mass is limited by its availability inside the element. This condition is usually evaluated using the CFL which general deﬁnition for each element is:

∆t c = c (A.4) n d ∆x Where:

√ cd : Wave celerity ( gH in shallow waters) ∆x: Length in main direction

According to Kämpf(2010), for 2D shallow water equations with irregular dimension elements, CFL can be better deﬁned as: ∆t c = p2gh (A.5) n max min(∆x, ∆y) Where:

hmax : Maximum water depth in the domain ∆y: Length in the secondary axis

From Equation A.5 we can analyze that the worst cases for numerical stability in variable width grids would be very fine cells in deep waters. In order to reduce the numerical diffusion and instabilities due to truncation of the higher order terms in the modified conservative equations, CFL should restricted to values close to one for explicit numerical schemes (Martin and McCutcheon, 1999), however, for implicit or semi-implicit numerical solvers, this threshold is a lot more flexible and CFLs all the way to 10 are common.

A.1.1 Forcing functions

Forcing functions are mathematical representations of the body forces affecting flow and transport processes in the system, they stand for the X term in the unidimensional conservation equation presented in Equation A.2. Water level variation due to tidal movements can be defined as the sum of a series of superimposed oscillations at certain known frequencies according to the cyclic nature of the movement of the sun

72 and the moon, as well as other celestial bodies. The general equation to describe this process is:

N X ζ(t) = ζ0 + fi hi cos(ωt + αi ) (A.6) i=1 Where:

N: Number of representative harmonics hi : Amplitude

ζ0: Mean level over vertical datum ωt : Angular velocity

fi : Year correction factor αt : Phase

Another forcing of importance to water bodies, and specially in the case of shallow estuarine areas, is the shear stress caused by the wind when it incides horizontally over a large enough water surface (wind fetch). The wind stress can be described by the following equations:

= p 2 + 2 (A.7) τs,x CD Uw Vw Uw = p 2 + 2 (A.8) τs,y CD Uw Vw Vw

CD: Wind drag coeﬃcient Uw , Vw : X and Y wind speed components.

Bottom friction inﬂuences the magnitude of the velocity vector in the model elements and therefore is one of the main calibration terms used in hydrodynamic models. The equation for bottom shear stress is:

√ 2 2 τb,x = CB u + v u (A.9) √ 2 2 τb,y = CB u + v v (A.10)

Where:

CB: Bottom drag coeﬃcient u, v: Fluid velocity components in x and y

There are many empirical approximations to determine drag coefficients for both bottom and surface wind stresses based in parameters like bottom rugosity or theoretical wind distributions. However, the main trend in modelling studies is to use this variables as calibration parameters for the hydrodynamic compartment. An optional force influencing velocity vector direction is the Coriolis effect due to Earth’s rotation, this effect is just relevant for large water bodies where Coriolis greatly overcomes inertial forces, which be assessed by using the adimensional Rossby number:

u Ro = (A.11) Lf f = 2Ω sin ϕ (A.12)

73 Where:

L: Representative length of water body Ω: Earth’s angular frequency f : Coriolis frequency ϕ: Latitude

A.2 Equations for constituent transport

The transport and transformation of substances in the water matrix due to hydrodynamics and domain speciﬁc processes can be represented in a 1-D fashion by the following equation:

∂C ∂C ∂ ∂C = −u + D + S + R + Q (A.13) ∂t ∂x ∂x ∂x Where:

C: Constituent concentration D: Dispersion coeﬃcient t: Time S: Mass change by settling x: Distance R: Reactive mass change u: Advection velocity in x direction Q: Loads/sinks

Two main types of transport are possible in water quality modelling according to the nature of the substance to be transported, if the substance is not aﬀected by consumption/production processes inside the model domain (R = 0), the substance is considered conservative, this is the case of parameters like salinity or heavy metals, on the other hand, substances that are aﬀected by chemical or biological processes inside the domain (R 6= 0) are called non conservative and require that the processes driving these conversions have a mathematical description in order to include them in the model, this is the case of nutrients and oxygen cycles in natural waters.

74 APPENDIX TWO

TIDAL CONSTITUENT ANALYSIS

Tidal analysis for Santa Marta - Year 2015 3

2 Original Time series 1 Tidal prediction from Analysis Original time series minus Prediction 0 Elevation (m)

-1 0 20 40 60 80 100 120 140 160 180 200 Days in 2015

100 Analyzed lines with 95% significance level Significant Constituents SSA PHI1 MKS2 95% Significance Level TAU1K1 MSM O1 P1 MM NU2M2 -2 MSFMF 10 Q1RHO1BET1THE1 N2 K2 NO1CHI1 UPS1 2N2 L2 MU2 LDA2S2MSN2 2Q1 J1SO1 OQ2

Amplitude (m) SIG1 EPS2 ALP1 OO1 MO3M3 2MK5 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 frequency (cph)

360 Analyzed Phase angles with 95% CI Significant Constituents 270 180 90 0

Greenwich Phase (deg) 0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 frequency (cph)

Fig. B.1: Observed tidal components for station "Santa Marta" in year 2015, extracted with MAT- LAB T_Tide toolbox

number of standard constituents used: 59 Points used: 304333 of 304407 Using nonlinear bootstrapped error estimates Generating prediction with nodal corrections , SNR is 1.000000 percent of var residual after synthesis/var original: 56.81 % −−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−− date : 25−Aug−2015 nobs = 304407, ngood = 304333, record length (days) = 211.39 start time: 01−Jan −2015 rayleigh criterion = 1.0 x0= 2, x trend= 0

75 var(x)= 0.015945 var(xp)= 0.0068869 var(xres)= 0.009058 percent var predicted/var original= 43.2 %

tidal amplitude and phase with 95% CI estimates tide freq amp amp_err pha pha_err snr ∗SSA 0.0002282 0.0548 0.000 155.38 0.52 1.2e+04 ∗MSM 0.0013098 0.0264 0.000 45.33 1.07 3.1e+03 ∗MM 0.0015122 0.0162 0.000 346.39 1.80 1.1e+03 ∗MSF 0.0028219 0.0094 0.000 239.35 2.80 3.9e+02 ∗MF 0.0030501 0.0075 0.000 26.55 3.63 2.3e+02 ∗ALP1 0.0343966 0.0008 0.001 178.62 46.21 2.3 ∗2Q1 0.0357064 0.0019 0.001 289.70 19.78 9.2 ∗SIG1 0.0359087 0.0012 0.001 176.32 33.13 4.1 ∗Q1 0.0372185 0.0067 0.001 50.93 4.62 1.2e+02 ∗RHO1 0.0374209 0.0064 0.001 129.79 6.16 85 ∗O1 0.0387307 0.0261 0.001 40.97 1.33 1.8e+03 ∗TAU1 0.0389588 0.0275 0.000 277.78 0.84 4.7e+03 ∗BET1 0.0400404 0.0075 0.001 157.35 5.15 1.1e+02 ∗NO1 0.0402686 0.0045 0.000 57.76 5.12 1.5e+02 ∗CHI1 0.0404710 0.0048 0.001 154.36 8.48 71 ∗P1 0.0415526 0.0204 0.000 76.02 1.52 1.8e+03 ∗K1 0.0417807 0.0289 0.001 305.98 1.13 2.7e+03 ∗PHI1 0.0420089 0.0699 0.001 79.29 0.44 1.9e+04 ∗THE1 0.0430905 0.0059 0.001 299.28 5.44 96 ∗J1 0.0432929 0.0025 0.001 85.68 13.56 20 ∗SO1 0.0446027 0.0024 0.001 130.36 15.65 13 ∗OO1 0.0448308 0.0009 0.001 310.87 44.78 1.3 ∗UPS1 0.0463430 0.0045 0.001 80.68 12.93 24 ∗OQ2 0.0759749 0.0021 0.000 319.30 14.01 20 ∗EPS2 0.0761773 0.0012 0.000 72.94 21.63 7.9 ∗2N2 0.0774871 0.0037 0.000 254.56 7.59 59 ∗MU2 0.0776895 0.0027 0.001 117.13 10.05 27 ∗N2 0.0789992 0.0057 0.000 322.38 4.82 1.4e+02 ∗NU2 0.0792016 0.0148 0.000 158.28 1.801e+03 ∗M2 0.0805114 0.0185 0.000 16.74 1.57 1.5e+03 ∗MKS2 0.0807396 0.0536 0.001 135.40 0.60 7.1e+03 ∗LDA2 0.0818212 0.0030 0.000 129.87 9.00 36 ∗L2 0.0820236 0.0046 0.000 262.50 7.60 85

76 ∗S2 0.0833333 0.0027 0.001 291.03 9.78 29 ∗K2 0.0835615 0.0074 0.001 10.23 4.96 94 ∗MSN2 0.0848455 0.0032 0.000 7.28 9.04 47 ∗MO3 0.1192421 0.0007 0.001 207.89 46.79 1.6 ∗M3 0.1207671 0.0008 0.000 168.02 37.14 3.2 ∗2MK5 0.2028035 0.0004 0.000 121.50 65.241

Tidal analysis for Santa Marta - Year 2014 3

2 Original Time series 1 Tidal prediction from Analysis Original time series minus Prediction 0 Elevation (m)

-1 0 50 100 150 200 250 300 350 Days in 2014

100 Analyzed lines with 95% significance level Significant Constituents 95% Significance Level K1 SSA O1 P1 M2 MSM TAU1 PHI1 MKS2 MMMF RHO1 THE1 N2 -2 MSF NU2 S2 10 Q1 BET1NO1CHI1 J1 LDA2L2 K2 2Q1 OO1 MSN2ETA2 2N2MU2 EPS2 Amplitude (m) SIG1 SO1 OQ2 UPS1 SK3 M3SO3 ALP1 MO3 0 0.02 0.04 0.06 0.08 0.1 0.12 frequency (cph)

360 Analyzed Phase angles with 95% CI Significant Constituents 270

180

Greenwich Phase (deg) 0 0.02 0.04 0.06 0.08 0.1 0.12 frequency (cph)

Fig. B.2: Observed tidal components for station "Santa Marta" in year 2014, extracted with MAT- LAB T_Tide toolbox

number of standard constituents used: 59 Points used: 520851 of 520857 percent of var residual after lsqfit/var original: 68.21 % Using nonlinear bootstrapped error estimates Generating prediction with nodal corrections , SNR is 1.000000 percent of var residual after synthesis/var original: 68.21 % −−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−− date : 25−Aug−2015 nobs = 520857, ngood = 520851, record length (days) = 361.71 start time: 01−Jan −2014 00:01:00 rayleigh criterion = 1.0 x0= 2.01, x trend= 0 var(x)= 0.013003 var(xp)= 0.0041335 var(xres)= 0.0088692 percent var predicted/var original= 31.8 %

77 tidal amplitude and phase with 95% CI estimates tide freq amp amp_err pha pha_err snr ∗SSA 0.0002282 0.0350 0.000 70.03 0.65 8.3e+03 ∗MSM 0.0013098 0.0206 0.000 293.83 1.03 3.2e+03 ∗MM 0.0015122 0.0146 0.000 215.87 1.41 1.7e+03 ∗MSF 0.0028219 0.0074 0.000 112.23 2.53 3.9e+02 ∗MF 0.0030501 0.0172 0.000 309.85 1.17 2.4e+03 ∗ALP1 0.0343966 0.0005 0.000 277.64 52.64 1.6 ∗2Q1 0.0357064 0.0038 0.000 199.95 6.38 90 ∗SIG1 0.0359087 0.0015 0.000 167.19 18.60 12 ∗Q1 0.0372185 0.0061 0.000 198.01 4.10 2.1e+02 ∗RHO1 0.0374209 0.0116 0.000 140.73 2.54 6.5e+02 ∗O1 0.0387307 0.0275 0.000 152.17 0.94 3.9e+03 ∗TAU1 0.0389588 0.0219 0.000 204.93 0.93 4.6e+03 ∗BET1 0.0400404 0.0059 0.001 291.26 4.14 1.4e+02 ∗NO1 0.0402686 0.0068 0.000 211.56 3.40 2.2e+02 ∗CHI1 0.0404710 0.0060 0.000 81.72 4.56 1.9e+02 ∗P1 0.0415526 0.0276 0.000 95.35 0.76 5.8e+03 ∗K1 0.0417807 0.0480 0.000 122.29 0.59 1.3e+04 ∗PHI1 0.0420089 0.0206 0.000 340.18 0.95 2.9e+03 ∗THE1 0.0430905 0.0102 0.000 93.29 2.44 6.4e+02 ∗J1 0.0432929 0.0054 0.000 348.95 4.55 1.3e+02 ∗SO1 0.0446027 0.0015 0.000 49.25 18.33 8.8 ∗OO1 0.0448308 0.0031 0.001 256.30 11.45 19 ∗UPS1 0.0463430 0.0009 0.001 308.39 42.01 1.5 ∗OQ2 0.0759749 0.0013 0.000 34.06 15.77 15 ∗EPS2 0.0761773 0.0017 0.000 77.52 12.91 17 ∗2N2 0.0774871 0.0023 0.000 180.80 9.65 29 ∗MU2 0.0776895 0.0024 0.000 274.02 8.03 32 ∗N2 0.0789992 0.0108 0.000 270.64 2.11 8.8e+02 ∗NU2 0.0792016 0.0079 0.000 236.81 2.74 5.4e+02 ∗M2 0.0805114 0.0324 0.000 260.70 0.65 7.1e+03 ∗MKS2 0.0807396 0.0215 0.001 142.83 1.45 1.8e+03 ∗LDA2 0.0818212 0.0067 0.000 336.14 3.24 3.1e+02 ∗L2 0.0820236 0.0058 0.000 57.42 3.693e+02 ∗S2 0.0833333 0.0075 0.000 158.31 2.90 4.2e+02 ∗K2 0.0835615 0.0068 0.000 79.51 3.84 2.1e+02 ∗MSN2 0.0848455 0.0037 0.000 143.74 5.591e+02 ∗ETA2 0.0850736 0.0032 0.001 199.02 9.32 34

78 ∗MO3 0.1192421 0.0004 0.000 256.43 59.501 ∗M3 0.1207671 0.0006 0.000 141.39 34.92 3.2 ∗SO3 0.1220640 0.0006 0.000 310.22 45.87 1.4 ∗SK3 0.1251141 0.0012 0.000 104.14 21.34 9.5

Tidal analysis for Puerto Velero - Year 2013

0.2

Original Time series Elevation (m) -0.2 Tidal prediction from Analysis Original time series minus Prediction 25-Aug 01-Sep 08-Sep Date

100 Analyzed lines with 95% significance level Significant Constituents K1 O1 M2 95% Significance Level S2 -2 MSF 10 M3 Amplitude (m)

0 0.02 0.04 0.06 0.08 0.1 0.12 frequency (cph)

360 Analyzed Phase angles with 95% CI Significant Constituents 270

180

Greenwich Phase (deg) 0 0.02 0.04 0.06 0.08 0.1 0.12 frequency (cph)

Fig. B.3: Observed tidal components for station "Puerto Velero" between August 24th and Septem- ber 11th of 2013, extracted with MATLAB T_Tide toolbox

PUERTO VELERO : PARTE 1 24−08−2013 19:00:00 a 2013−09−11 2 1 : 0 0 : 0 0 number of standard constituents used: 17 Points used: 433 of 435 Using nonlinear bootstrapped error estimates Generating prediction with nodal corrections , SNR is 1.000000 percent of var residual after synthesis/var original: 16.67 % −−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−− date : 25−Aug−2015 nobs = 435, ngood = 433, record length (days) = 18.13 start time: 24−Aug−2013 19:00:00 rayleigh criterion = 1.0 x0= −0.0541, x trend= 0 var(x)= 0.0070745 var(xp)= 0.0058944 var(xres)= 0.0011792 percent var predicted/var original= 83.3 %

tidal amplitude and phase with 95% CI estimates

79 tide freq amp amp_err pha pha_err snr ∗MSF 0.0028219 0.0094 0.005 348.76 29.94 4.3 ∗O1 0.0387307 0.0559 0.006 170.28 6.04 84 ∗K1 0.0417807 0.0729 0.005 176.11 4.53 1.9e+02 ∗M2 0.0805114 0.0645 0.004 352.87 4.06 2.1e+02 ∗S2 0.0833333 0.0281 0.005 251.18 9.09 36 ∗M3 0.1207671 0.0062 0.005 269.67 40.51 1.8

Tidal analysis for Puerto Velero - Year 2014 1.5

1 Original Time series 0.5 Tidal prediction from Analysis Original time series minus Prediction 0 Elevation (m)

-0.5 01-Nov 01-Dec Date

100 Analyzed lines with 95% significance level K1 Significant Constituents O1 95% Significance Level M2 N2 MSF S2 -2 MM 10 Q1 Amplitude (m)

0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 frequency (cph)

360 Analyzed Phase angles with 95% CI Significant Constituents 270

180

Greenwich Phase (deg) 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 frequency (cph)

Fig. B.4: Observed tidal components for station "Puerto Velero" between October 24th and De- cember 31st of 2013, extracted with MATLAB T_Tide toolbox

PUERTO VELERO − 2013 2013−10−24 1 6 : 0 0 : 0 0 −− ’2013−12−31 23:00:00’ number of standard constituents used: 35 Points used: 1636 of 1640 Using nonlinear bootstrapped error estimates Generating prediction with nodal corrections , SNR is 1.000000 percent of var residual after synthesis/var original: 36.56 % −−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−− date : 25−Aug−2015 nobs = 1640, ngood = 1636, record length (days) = 68.33 start time: 24−Oct −2013 16:00:00 rayleigh criterion = 1.0 x0= 1.05, x trend= 0

80 var(x)= 0.01556 var(xp)= 0.0098715 var(xres)= 0.0056881 percent var predicted/var original= 63.4 %

tidal amplitude and phase with 95% CI estimates tide freq amp amp_err pha pha_err snr ∗MM 0.0015122 0.0097 0.005 337.88 32.91 3.6 ∗MSF 0.0028219 0.0170 0.005 183.69 18.22 11 ∗Q1 0.0372185 0.0069 0.006 158.79 53.57 1.1 ∗O1 0.0387307 0.0630 0.007 172.93 5.97 89 ∗K1 0.0417807 0.1156 0.006 159.15 2.71 3.8e+02 ∗N2 0.0789992 0.0242 0.005 329.33 14.38 23 ∗M2 0.0805114 0.0662 0.005 350.07 4.95 1.7e+02 ∗S2 0.0833333 0.0147 0.006 248.89 20.567

Tidal analysis for Puerto Velero - Year 2014 1.5

1 Original Time series 0.5 Tidal prediction from Analysis Original time series minus Prediction 0 Elevation (m)

-0.5 01-Feb 01-Mar Date

100 Analyzed lines with 95% significance level Significant Constituents K1 O1 M2 95% Significance Level N2 MM S2 MSF Q1 NO1 MU2 10-2 J1 L2 M3 Amplitude (m)

0 0.02 0.04 0.06 0.08 0.1 0.12 frequency (cph)

360 Analyzed Phase angles with 95% CI Significant Constituents 270

180

Greenwich Phase (deg) 0 0.02 0.04 0.06 0.08 0.1 0.12 frequency (cph)

Fig. B.5: Observed tidal components for station "Puerto Velero" between January 1st and March 4th of 2014, extracted with MATLAB T_Tide toolbox

PUERTO VELERO − 2014 ’2014−01−01 01:00:00’ − ’2014−03−04 14:00:00’ number of standard constituents used: 35 Points used: 1492 of 1502 Using nonlinear bootstrapped error estimates Generating prediction with nodal corrections , SNR is 1.000000 percent of var residual after synthesis/var original: 20.83 % −−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−− date : 25−Aug−2015

81 nobs = 1502, ngood = 1492, record length (days) = 62.58 start time: 01−Jan −2014 01:00:00 rayleigh criterion = 1.0 x0= 0.95, x trend= 0 var(x)= 0.011427 var(xp)= 0.0090492 var(xres)= 0.0023803 percent var predicted/var original= 79.2 %

tidal amplitude and phase with 95% CI estimates tide freq amp amp_err pha pha_err snr ∗MM 0.0015122 0.0216 0.004 314.42 9.98 33 ∗MSF 0.0028219 0.0151 0.003 43.08 14.89 20 ∗Q1 0.0372185 0.0121 0.004 165.85 24.27 9.8 ∗O1 0.0387307 0.0618 0.005 174.44 4.00 1.8e+02 ∗NO1 0.0402686 0.0099 0.005 192.95 33.42 3.2 ∗K1 0.0417807 0.0972 0.004 180.17 2.43 7.4e+02 ∗J1 0.0432929 0.0073 0.004 157.14 28.98 3.4 ∗MU2 0.0776895 0.0095 0.004 329.82 22.20 5.5 ∗N2 0.0789992 0.0257 0.003 326.73 8.57 65 ∗M2 0.0805114 0.0726 0.004 346.80 2.69 3.4e+02 ∗L2 0.0820236 0.0047 0.003 22.43 40.89 2.1 ∗S2 0.0833333 0.0196 0.004 266.01 10.38 30 ∗M3 0.1207671 0.0042 0.003 238.10 44.54 1.7

PUERTO VELERO − 2014 ’2014−03−04 15:00:00’ − ’2014−12−31 23:00:00’ number of standard constituents used: 59 Points used: 7257 of 7257 Using nonlinear bootstrapped error estimates Generating prediction with nodal corrections , SNR is 1.000000 percent of var residual after synthesis/var original: 42.94 % −−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−− date : 25−Aug−2015 nobs = 7257, ngood = 7257, record length (days) = 302.38 start time: 04−Mar−2014 15:00:00 rayleigh criterion = 1.0 x0= 0.16, x trend= 0

82 Tidal analysis for Puerto Velero - Year 2014 0.6 Original Time series Tidal prediction from Analysis 0.4 Original time series minus Prediction

0.2

0 Elevation (m)

-0.2 01-Apr 01-Jul 01-Oct Date

100 Analyzed lines with 95% significance level Significant Constituents K1 O1 M2 95% Significance Level P1 MSMMF N2 S2 -2 SSAMM Q1 10 MSF NO1 J1 K2 2N2NU2MKS2 SO1 M3 Amplitude (m)

0 0.02 0.04 0.06 0.08 0.1 0.12 frequency (cph)

360 Analyzed Phase angles with 95% CI Significant Constituents 270

180

Greenwich Phase (deg) 0 0.02 0.04 0.06 0.08 0.1 0.12 frequency (cph)

Fig. B.6: Observed tidal components for station "Puerto Velero" between March 4th and December 31st of 2014, extracted with MATLAB T_Tide toolbox var(x)= 0.015853 var(xp)= 0.0090663 var(xres)= 0.0068078 percent var predicted/var original= 57.2 %

tidal amplitude and phase with 95% CI estimates tide freq amp amp_err pha pha_err snr ∗SSA 0.0002282 0.0123 0.003 126.04 13.14 17 ∗MSM 0.0013098 0.0193 0.003 309.29 8.87 45 ∗MM 0.0015122 0.0117 0.002 261.18 14.03 22 ∗MSF 0.0028219 0.0049 0.003 86.69 36.88 3.4 ∗MF 0.0030501 0.0188 0.003 326.66 7.81 39 ∗Q1 0.0372185 0.0119 0.003 165.55 16.76 12 ∗RHO1 0.0374209 0.0031 0.003 168.03 63.67 1.1 ∗O1 0.0387307 0.0602 0.003 172.74 3.30 3.4e+02 ∗NO1 0.0402686 0.0051 0.003 228.66 34.33 3.3 ∗P1 0.0415526 0.0284 0.003 166.67 5.04 90 ∗K1 0.0417807 0.0998 0.003 165.85 1.54 1.2e+03 ∗J1 0.0432929 0.0054 0.003 159.02 37.52 2.7 ∗OO1 0.0448308 0.0044 0.004 149.33 63.57 1.1 ∗2N2 0.0774871 0.0042 0.003 327.10 39.21 2.8 ∗N2 0.0789992 0.0241 0.002 325.33 6.38 98 ∗NU2 0.0792016 0.0045 0.003 355.97 33.78 2.4 ∗M2 0.0805114 0.0672 0.003 347.85 2.23 5.6e+02 ∗MKS2 0.0807396 0.0036 0.003 103.28 57.70 1.3

83 ∗S2 0.0833333 0.0193 0.003 259.75 9.18 42 ∗K2 0.0835615 0.0053 0.004 253.19 38.81 2.2 ∗M3 0.1207671 0.0024 0.002 240.28 62.12 1.1

84 APPENDIX THREE

MOHID WATERQUALITY MODULE EQUATIONS

MODEL EQUATIONS KEYWORDS

1. PHYTOPLANKTON

1.1 Growth processes

The variation of the state variable P is written:

∂ P = ()µ − − − − ()P P rP eP mP .P GZ 2 .Z1 (1) ∂t

µ P - phytoplankton specific growth rate supported by nitrogen

- endogenous respiration and photorespiration rP

eP - fraction of primary production exuded as DON

m - phytoplankton natural mortality rate P P 86 GZ 2 - loss of phytoplankton due to mesozooplankton grazing

The phytoplankton specific growth rate is:

µ = max Ψ()Ψ ()Ψ ( ) P VP . T P . L P . N P (2)

max VP - maximum specific nitrogen uptake rate at a reference temperature GROWMAXF

Ψ(), Ψ(), and Ψ() represents the specific growth rate dependency on temperature, light and nutrients respectively. T P L P N P

Water Quality Module Manual 1

1.1.1 Nitrogen limiting factor

A Michaelis-Menten function is used for nitrogen limitation:

N inorg Ψ()N = (3) P P + K N N inorg

P K N - half-saturation constant ÑSATCONS

N - useful concentration of inorganic dissolved nitrogen (ammonia + nitrate). inorg

1.1.2 Light limiting factor

 ()  I z  1−  I()z I ()− Ψ()=  s  ()= Kd .z

L .e with I z I .e (4) 87 P 0 I s

I s - optimum light intensity for photosynthesis I - effective solar radiation at the water surface 0 z - vertical position K d - light extinction factor. This factor is obtained according to Parsons et al. (1995),

2 / 3 K = 0,04 + 0,0088.Ch + 0,54.Ch (5) d

Ch - chlorophyll concentration.

Water Quality Module Manual 2

1.2 Respiration

Respiration is divided in dark respiration and in photorespiration. The dark respiration is defined, according to Parker et al.(1980), as:

= P 0,069T re K r e (6)

P - phytoplankton endogenous respiration constant K r FENDREPC T - temperature

The photorespiration, proportional to the gross photosynthetic rate, is:

= P µ rp K p P (7)

P K p - proportionality factor. PHOTORES So, the respiration rate is defined as:

88 P = + r re rp (8)

1.3 Excretion

P e = K .µ .()1− Ψ()L (9) P e P

K P - excretion constant EXCRCONS e

The amount of nitrogen excreted by phytoplankton is given by:

e N = ()e + r P .α (10) P P P

r P - respiration rate α FRATIONC P - the phytoplankton N:C ratio

Water Quality Module Manual 3

1.4 Natural mortality

The natural mortality, following a modified Michaelis-Menten formulation proposed by Rodgers and Salisbury (1981), is:

P µ Tref m = m . P (11) e max P K P + m µ P

Tref mmax - maximum mortality rate at a reference temperature FMORTMAX P K m - mortality semi-saturation constant FMORTCON

Dead phytoplankton concentrations are them converted into nitrogen units using the phytoplankton N:C ratio ()α by: P

N = α mP mP . P (12) 89

2. BACTERIA

The state equation of the bacterial biomass B is:

∂B = ()µ − e − m .B − ()G B .Z1 (13) ∂t B B B Z1

µ B - total bacterial uptake e - excretion rate BARESPCO B NATMORB mB - natural mortality rate B GZ1 - grazing rate of microzooplankton on bacteria

Water Quality Module Manual 4

Total uptake rate of bacteria ()µ is the sum of the specific uptake rate for each one of the nutrient sources (DOMnr, ammonium, B and POM):

µ = µ Ndnr + µ N 2 + µ Np (14) B B B B

The specific uptake rate of bacteria is dependent on resource availability (organic substrate), accordingly to a Michaelis-Menten function, and on temperature. It is written as:

N µ N = V max . x .Ψ()T (15) B B B + K n N x

max VB - maximum specific nutrient uptake rate BMAXUPTA

N x - available substrate K B - half-saturation constant for nutrient uptake BACNCONS n

For ammonium uptake to take place DOMnr and POM concentration must be higher than the bacteria minimum substrate BACMINSUB 90 concentration.

()α The nitrogen uptake is converted in carbon units using the N:C ratio of bacteria B assuming that the uptake of ammonia need BRATIONC carbon in the corresponding rate to keep a constant composition. For the transformation of DOMnr and PON the N:C ratio of dissolved organic matter ()α is used. SRATIONC S

µ Npnr + µ Np µ N 2 µ = B B + B (16) B α α S B

Dead bacteria is also converted into nitrogen units according to:

N = α mB mB . B (17)

Water Quality Module Manual 5

3. ZOOPLANKTON

3.1 Growth processes

3.1.1 Microzooplankton

The variation of the microzooplankton biomass Z is written: 1

∂Z 1 = ()µ − e − m .Z − G Z1 .Z (18) ∂ Z1 Z1 Z1 1 Z 2 2 t

µ Z1 - microzooplankton gross growth rate Z1 GZ 2 - predation rate of mesozooplankton on microzooplankton m - specific mortality rate Z1

eZ1 - specific excretion rate

91 The microzooplankton gross growth rate is defined as:

µ = a .G B (19) Z1 Z Z1

aZ1 - assimilation coefficient of microzooplankton for bacteria CILBACASS B GZ1 - the grazing on bacteria

The parameterization of microzooplankton grazing on bacteria is

B = max Ψ B Ψ() GZ1 gZ1 . Z1. T (20)

max g Z1 - maximum ingestion rate CINGMAX Ψ B Z1 - limitation by available bacteria biomass Ψ()T - limitation by temperature

Water Quality Module Manual 6 This term is dependent on food availability accordingly to a Michaelis-Menten function including accessible food concentration (prey concentration*capture efficiency) and the threshold standing stock below which predation will cease. If the available food is ()Ψ prey = lower than this concentration, limitation of ingestion will reach its maximum pred 0 .

c B .B − s Bmin Ψ B = Z1 Z1 , if c B .B − s Bmin > 0 (21) Z1 K + ()c B .B − s Bmin Z1 Z1 Z1 Z1 Z1

c B - capture efficiency of bacteria Z1 EFFCAPBA Bmin sZ1 - threshold standing stock of bacteria below which predation cease GRAZBACMIN K Z1 - half saturation constant for grazing INGCONSC If the condition is not satisfied then Ψ B = 0 Z1

3.1.2 Mesozooplankton

The time variation of the mesozooplankton biomass Z 2 is: 92

∂Z 2 = ()µ − e − m .Z (22) ∂ Z 2 Z 2 Z 2 2 t

µ Z 2 - mesozooplankton gross growth rate

m - specific mortality Z 2 eZ 2 - excretion rate

µ = g P .G P + g Z1.GZ1 (23) Z 2 Z 2 Z 2 Z 2 Z 2

P ZOPHYASS g Z 2 - assimilation coeficiente of phytoplankton by mesozooplankton Z1 g Z 2 - assimilation coefficient of microzoopalnkton by mesozooplankton ZOCILASS P GZ 2 - grazing of phytoplankton Z1 GZ 2 - predation of microzooplankton

Water Quality Module Manual 7

P = ρ Ψ P Ψ() GZ 2 P .I max . Z 2 . T (24)

Z1 = − ρ − P Ψ Z1 Ψ() GZ 2 (1 P ).(I max GZ 2 ). Z 2 . T (25)

ρ - proportion of phytoplankton in mesozooplankton ingestion P PHYRATING

I max - maximum ingestion rate ZINGMAX Ψ P Z 2 - limitation by phytoplankton concentration Z1 Ψ - limitation by microzooplankton concentration Z 2

These limitation are defined as:

c P .P − s Pmin Ψ P = Z 2 Z 2 (26) Z 2 P + ()P − Pmin K Z 2 cZ 2 .P sZ 2 93

Z1 − Z1min Z1 cZ 2 .P sZ 2 Ψ = (27) Z 2 Z1 + ()Z1 − Z1min K Z 2 cZ 2 .P sZ 2

P - capture efficiency of phytoplankton EFFCAPHI cZ 2 s Pmin - threshold standing stock of phytoplankton below which grazing cease GRAZFITOMIN Z 2 P K Z 2 - half saturation constant for ingestion phytoplankton INGCONSZ Z1 cZ 2 - capture efficiency of microzooplankton EFFCAPCIL s Z1min - threshold standing stock of microzooplankton below which predation cease Z 2 GRAZCILMIN Z1 K Z 2 - half saturation constant for microzooplankton ingestion INGCONSZ

Water Quality Module Manual 8

3.2 Natural mortality

Zooplankton specific mortality rate m is directly related to the concentration of prey F . Bellow a threshold concentration of X X min max prey ( FX ), the mortality is high and constant mX , given that it is assumed that zooplankton mortality is related to starvation. So, GRAZBACMIN mortality is written as: MAXMORTCI ; MAXMORTZ

a m = X + 0 > min mX mX , if FX FX , (28) FX or

max min = , if ≤ (29) mX mX FX FX

a m - shape factor for the mortality curve MORTCICOEF ; MORTZCOEF X MINMORTCI ; MINMORTZ m0 X - minimum mortality rate

94 min max 0 Each group of zooplankton has its own FX , FX , mX , and mX values. FX for microzooplankton corresponds to bacteria concentration and for mesozooplankton to phytoplankton and microzooplankton concentrations. Carbon released in this process is converted in nitrogen using the N:C ratio ()α for microzooplankton and mesozooplankton. X CRATIONC ; ZRATIONC

N = α mX mX . X (30)

3.3 Excretion

The excretion rate is given by Andersen and Nival (1989) as a temperature function: eX

= ()T eX ax .bX (31)

a X - excretion rate at 0ºc CEXCFAC ; ZEXCFAC b - shape factor for the excretion curve X CEXCCONS ; ZEXCCONS T - temperature

Water Quality Module Manual 9

The carbon release is converted into nitrogen units using the N:C ratio ()α . CRATIONC ; ZRATIONC X

N = r α eX eX . X (32)

3.4 Respiration

The respiration rate r is used for the oxygen simulation. It is assumed that oxygen consumption of heterotrophs is a constant X ()ρ X , and that the whole process is temperature dependent. So, the respiration rate used in oxygen differential equations is given CREFRESP ; ZREFRESP by:

r = ρ .Ψ()T (33) X X

Water Quality Module Manual 10 4. NITROGEN DYNAMICS

The simulation of nitrogen dynamics in the WQ model assumes 6 different forms of this nutrient. The dynamics of each one of this forms is therefore addressed. Starting with common rates for most of the forms we have,

Nitrification rate:

(T −20.0) []O K = K ref .T . 2 (34) nit nit nit sat + [] Knit O2

K ref - reference nitrification rate nit NITRIREF

Tnit - nitrification temperature coefficient TNITCOEF T - temperature K sat - nitrification semi-saturation constant nit NITSATCO

Denitrification rate:

sat − K 96 = ref (T 20.0) dnit K dnit K dnit .Tdnit . (35) sat + [] K dnit O2

ref K dnit - reference denitrification rate DENITREF

Tdnit - denitrification temperature coefficient TDENCOEF

sat - denitrification semi-saturation constant K dnit DENSATCO

Particulate organic nitrogen decomposition rate:

− P Np = ref (T 20.0) K dec K dec .Tdec . (36) Kr P + P

ref NOPREF K dec - PON decomposition reference rate T - PON decomposition temperature coefficient NOPCOEF dec

Water Quality Module Manual 11

Refractory dissolved organic nitrogen mineralization rate:

(T −20.0) P K Ndr = K ref .T . (37) min min min Kr P + P

ref K min - reference mineralization rate of DONr NMINR

Tmin - DONr mineralization temperature coefficient TMINR P - phytoplankton nutrient regeneration half-saturation rate Kr FREGSATC

− 4.1 Nitrate ( NO ) 3

∂N 1 = − []()− Φ α µ + − (38) 1 N 2 .P. P .P Knit .N1−2 K dnit .N1 ∂t

Phytoplankton Nitrite 97

Φ The ammonia preference factor N 2 used in the model is described by the formula:

[][]N . N K P .[]N Φ = 1 2 + N 2 (39) N 2 ()P + []()P + []()[][]+ ()P + [] K N N1 . K N N 2 N1 N 2 . K N N1

P K N - half-saturation constant NSATCONS

− 4.2 Nitrite ( NO2 )

∂N 1−2 = − K nit .N 2 K nit .N1−2 (40) ∂t LOMON Ammonia

Water Quality Module Manual 12 + 4.3 Ammonia ( NH ) 4

∂ N 2 N Sol In N N Sol In N Sol In = []()e .ε − ()Φ .µ .α .P +[]()e .α − µ 2 .B + ()()e .ε .Z + e .ε .Z ∂t LPOMOP OOO ONN2 OPOOP O LOMB OB ONBO LZOM1OZ ONO1 LZOM2OZ ONO2 Bacteria Microzooplankton Mesozooplankton Phytoplankton (41)

+ Ndr + ()Np φ − LKOMmin .NONdr K dec . P .N p K nit .N 2 LOMO ONO DONr PON

Φ N - phytoplankton ammonia preference factor 2 ε Sol In FSOLEXCR P - phytoplankton soluble inorganic excretion fraction

ε Sol In Z - zooplankton soluble inorganic excretion fraction ZSOLEXCR φ - available PON for transformation into ammonia P PHDECOMP

4.4 Non-refractory dissolved organic nitrogen (DONnr) 98 ∂N dnr = []()1− ε Sol In .e N .ε Dis Or .P − ()µ Ndnr .B + []e N .1− ε Sol In (.ε Diss Or .Z + []e N ).()1− ε Sol In .ε Diss Or .Z (42) ∂t LOMOP OO PONOP OO LOMB ON LZOM1 OOOZ ONZOOO1 LZOM2 OOOZ ONZOOO 2 Phytoplankton Bacteria Microzooplankton Mesozooplankton

ε Dis Or P - phytoplankton dissolved organic excretion fraction FDISSDON Diss Or ε - zooplankton dissolved organic excretion fraction Z ZDISSDON

4.5 Refractory dissolved organic nitrogen (DONr)

∂N dr = K Np .()1−φ .N − K Ndr .N (43) ∂t LdecOMO ONP O P min dr PON

Water Quality Module Manual 13

4.6 Particulate organic nitrogen (PON)

∂N P = []e N .()()1− ε Sol In .1− ε Dis Or + m N .P − ()µ Np + m N .B +[]m N + ()1− a .G B .α + e N .()()1− ε Sol In .1− ε Diss Or .Z ∂t LPOMOOPOOO ONPOOOOPO LBOMO ONBO LZOM1 OOOZO1 OZO1 OB OOZ1ONOOZ OOOOZ OOO1 Phytoplankton Bacteria Microzooplankton []N ()()− ε Sol In − ε Diss Or + N + δ + ϕ − Np ()−φ − Np φ () LeZOM2 .1OOZOO.O1 ONZ OOOmOZ 2O.Z 2 P N LK decOM.O1 ONP O.N P LK decOMO. PON.ON P Mesozooplankton DONr Ammonia

δ - stoichiometric food web losses, defined by P

δ = ()− P P α + − Z1 Z1 ()α P 1 g Z 2 .GZ 2 . Z 2 1 g Z 2 .GZ 2 . Z1 (45)

ϕ N - non assimilated phytoplankton and microzooplankton

ϕ = µ ()α −α 99 N Z 2 . P Z 2 (46)

Water Quality Module Manual 14

APPENDIX

A1. TEMPERATURE EFFECT

The temperature effect, Ψ(T ) , on the various biological processes addressed by the model follows the concept of Thornton & Lessen (1978):

Ψ = (T ) ∗ (T ) (T ) kA kB (A1) where,

γ − 1 .(T Tmin ) (T ) = k1.e k A γ − (A2) + 1 .(T Tmin ) − 1 k1.(e 1) 100

λ − 2 .(Tmax T ) (T ) = k4.e kB γ − (A3) + 2 .(Tmax T ) − 1 k4.(e 1) with,

k (1− k ) ln 2 1 k (1− k ) γ = 1 2 (A4) 1 T opt − T min min

k (1− k ) ln 3 4 k (1− k ) γ = 4 3 (A5) 2 − opt Tmax Tmax

Water Quality Module Manual 15

opt Tmin - minimum temperature for the optimal growth interval TOPTFMIN ; TOPTZMIN; TOPTBMIN opt Tmax - maximum temperature for the optimal growth interval TOPTFMAX; TOPTZMAX; TOPTBMAX

Tmin - minimum tolerable temperature TFMIN; TZMIN; TBMIN

Tmax - maximum tolerable temperature TFMAX; TZMAX; TBMAX

The remaining constants k1 , k 2 , k3 , and k 4 , are used to control the shape of the temperature effect response curve. Given the lack of knowledge on the temperature effects on the various organisms considered in this study, theses values are assumed to be equal for all kinds of organisms in the model.

A2. State variables in the model 101

Variable Definition Unit P Phytoplankton B Bacteria Z1 Microzooplankton Z2 Mesozooplankton N1 Nitrate N2 Ammonium N1-2 Nitrite Ndnr Dissolved organic matter - labile Ndr Dissolved organic matter - refractory Np Particulate organic matter

Water Quality Module Manual 16 A3. KEYWORDS

A3.1 Non specific rates & constants

NSATCONS NSatConst Nitrogen half-saturation constant, mgN/l

FREGSATC PhytoNutRegenerationSatConst Phytoplankton nutrient regeneration half saturation rate, mgC/l

FRATIONC AlfaPhytoNC Phytoplankton ratio between Nitrogen and Carbon, mgN/mgC Redfield ratio

A3.2 Rates & constants to the PHOSPHORUS simulation

PMINEREF KPhosphorusMineralizationRate Reference Phosphorus mineralization rate, 1/T

PMINCOEF TPhosphorusMineralization Phosphorus mineralization temperature coefficient

FRATIOPC AlfaPhytoPC Redfield ratio phytoplankton ratio between Phosphorus and Carbon, mgP/mgC

ZRATIOPC AlfaZooPC Zooplankton ratio between Phosphorus and Carbon, mgP/mgC 102 A3.3 Rates & constants to the OXYGEN simulation

PHOTOSOC PhotosynthesisOxygenCarbonRatio Photosynthesis Oxygen:Carbon ratio, (M/L^3)/(M/L^3) mgO2 / mgC

ZOCRATIO RatioOxygenCarbonZooRespiration Zooplankton respiration Oxygen:Carbon ratio, mgO2 / mgC

NITONRAT NConsOxyNitRatio Secondary Oxygen production due to Nitrate consumption, (M/L^3)/(M/L^3)

A3.4 Rates & constants to the BOD simulation

BODCOEF BODOxidationCoefficient BOD oxidation coefficient

BODREF BODOxidationReferenceRate Reference BOD oxidation, 1/T

ODOSSAT BODOxygenSSatConstant Oxygen limitation half-saturation constant, 1/T

DENITRON DenitConvOxyNitMass During the Denitrification the organic material is decomposed, we need to convert Oxygen mass to Nitrogen mass, (M/L^3)/(M/L^3)

Water Quality Module Manual 17 A3.5 Rates & constants to the NITROGEN simulation

ZRATIONC AlfaZooNC Zooplankton ratio between Nitrogen and Carbon, mgN/mgC

NMINR KRefrAmmoniaMinRate Reference ammonia mineralization rate of the refractory DON, 1/T

NOPREF KPartDecompRate Reference particulate organic Nitrogen decomposition rate, 1/T

PHDECOMP PhytoAvaibleDecomp

DENITREF KDenitrificationRate Reference denitirfication rate, 1/T

NITRIREF KNitrificationRate Reference nitirfication rate, 1/T

TMINR TRefrAmmoniaMin Nitrogen mineralization temperature coefficient of the refractory DON

NOPCOEF TPartDecomposition Particulate organic Nitrogen decomposition temperature coefficient

TDENCOEF TDenitrification Denitirfication temperature coefficient

TNITCOEF TNitrification Nitrification temperature coefficient

NITSATCO NitrificationSatConst Nitrification semi-saturation constant, mgO2/l

DENSATCO DenitrificationSatConst Denitrification semi-saturation constant, mgO2/l 103 FSOLEXCR PhytoSolublInorgExcreFraction Soluble inorganic fraction of the phytoplankton excretions

FDISSDON PhytoExcreDissOrgFraction Dissolved organic fraction of the phytoplankton excretions

If ( PropCalc%Bacteria) then

PLANK_OC_RAT PlanktonOxygenCarbonRatio

ZSOLEXCR ZooSolublInorgExcreFraction Soluble inorganic fraction of the zooplankton excretions

ZDISSDON ZooExcreDissOrgFraction Dissolved organic fraction of the zooplankton excretions else

NMINENR KNonRefrAmmoniaMinRate Reference ammonia mineralization rate of the non refractory DON, 1/T

TMINNR TNonRefrAmmoniaMin Nitrogen mineralization temperature coefficient of the non refractory DON end if

Water Quality Module Manual 18 A3.6 Rates & constants to the PHYTOPLANKTON simulation

PSATCONS PSatConst Phosphorus half-saturation constant, phosphorus, M/L^3

GROWMAXF GrowMaxPhytoRate Maximum phytoplankton growth rate, 1/T

FMORTMAX PhytoMortMaxRate Phytoplankton maximum mortality, carbon, M/(L^3.T)

TOPTFMIN TOptPhytoMin Minimum temperature of the optimal interval for the phytoplankton growth, oC

TOPTFMAX TOptPhytoMax Maximum temperature of the optimal interval for the phytoplankton growth, oC

TFMIN TPhytoMin Minimum tolerable temperature of the interval for the phytoplankton growth, oC

TFMAX TPhytoMax Maximum tolerable temperature of the interval for the phytoplankton growth, oC

TFCONST1 FK1 Constant to control temperature response curve shape

TFCONST2 FK2 Constant to control temperature response curve shape

TFCONST3 FK3 Constant to control temperature response curve shape

TFCONST4 FK4 Constant to control temperature response curve shape

FMORTCON FMortSatConst Mortality half saturation rate, M/(L^3.T) 104 PHOTORES PhotorespFactor Fraction of actual photosynthesis which is oxidised by photorespiration

FENDREPC PhytoEndogRepConst Phytoplanktonendogenousrespirationconstant1/T

EXCRCONS PhytoExcretionConstant Phyto excretion constant

If (.NOT. PropCalc%Bacteria) then

ASS_EFIC E Assimilation efficiency of the phytoplankton by the zooplankton end if

Water Quality Module Manual 19 A3.7 Rates & constants to the ZOOPLANKTON simulation

TOPTZMIN TOptZooMin Minimum temperature of the optimal interval for the zooplankton growth oC

TOPTZMAX TOptZooMax Maximum temperature of the optimal interval for the zooplankton growth oC

TZMIN TZooMin Minimum tolerable temperature of the interval for the zooplankton growth oC

TZMAX TZooMax Maximum tolerable temperature of the interval for the zooplankton growth oC

TZCONST1 ZK1 Constant to control temperature response curve shape

TZCONST2 ZK2 Constant to control temperature response curve shape

TZCONST3 ZK3 Constant to control temperature response curve shape

TZCONST4 ZK4 Constant to control temperature response curve shape

ZREFRESP ZooReferenceRespirationRate Rate of consumption of Carbon by respiration and non-predatory mortality at the reference temperature, 1/T

GRAZFITOMIN GrazPhytoMin Minimum phytoplankton concentration for the existence of grazing, mgC/l

If ( PropCalc%Bacteria) then 105

GRAZCILMIN GrazCiliateMin Minimum phytoplankton concentration for the existence of grazing, mgC/l

ZEXCFAC ZooExcretionFactor

ZEXCCONS ZooExcretionConst Zoo excretion constant

MORTZCOEF ZooMortalityCoef

MINMORTZ ZooMinMortalityRate

MAXMORTZ ZooMaxMortalityRate 1/day

INGCONSZ ZooIngestionConst 1/2 sat

EFFCAPHI ZooEfficiencyCapturePhyto

EFFCAPCIL ZooEfficiencyCaptureCiliate

ZINGMAX ZooIngestionMax

ZOPHYASS ZooAssimilationPhytoRate

ZOCILASS ZooAssimilationCiliateRate

Water Quality Module Manual 20 PHYRATING PhytoRatioIngestionZoo else

GROWMAXZ GrowMaxZooRate Maximum zooplankton growth rate, 1/T

IVLEVCON IvlevGrazConst Ivlev grazing constant

PREDMOR ZPredMortalityRate Predatory mortality rate, 1/T

A3.8 Rates & constants to the CILIATE simulation

GRAZBACMIN GrazBactMin Minimum phytoplankton concentration for the existence of grazing, mgC/l

CREFRESP CiliateReferenceRespirationRate Rate of consumption of Carbon by respiration and

CEXCFAC CiliateExcretionFactor Excretion constant

CEXCCONS CiliateExcretionConst Excretion constant

MORTCICOEF CiliateMortalityCoef 106

MINMORTCI CiliateMinMortalityRate

MAXMORTCI CiliateMaxMortalityRate

INGCONSC CiliateIngestionConst 1/2 sat

EFFCAPBA CiliateEfficiencyCaptureBacteria

CINGMAX CiliateIngestionMax

CILBACASS CiliateAssimilationBacteriaRate

Water Quality Module Manual 21 A3.9 Rates & constants to the BACTERIA simulation

CRATIONC AlfaCilNC Ciliate ratio between Nitrogen and Carbon, mgN/mgC

BRATIONC AlfaBacteriaNC Bacteria ratio between Nitrogen and Carbon, mgN/mgC

SRATIONC AlfaSubstratNC Organic dissolved substrat ratio between Nitrogen and Carbon, mgN/mgC

NATMORB BacteriaNonGrazingMortalityRate !1/day

BARESPCO BacteriaExcretionRate

BMAXUPTA BacteriaMaxUptake day-1 /24

BMINUPTA BacteriaMinUptake

BACMINSUB BacteriaMinSubstrate

BACNCONS NitrogenSaturationConstBacteria mgN/l

TOPTBMIN TOptBacteriaMin Minimum temperature of the optimal interval for the Bacteria growth, oC

TOPTBMAX TOptBacteriaMax Maximum temperature of the optimal interval for the Bacteria growth, oC

TBMIN TBacteriaMin Minimum tolerable temperature of the interval for the Bacteria growth, oC 107 TBMAX TBacteriaMax Maximum tolerable temperature of the interval for the Bacteria growth, oC

TBCONST1 BK1 Constant to control temperature response curve shape

TBCONST2 BK2 Constant to control temperature response curve shape

TBCONST3 BK3 Constant to control temperature response curve shape

TBCONST4 BK4 Constant to control temperature response curve shape

Water Quality Module Manual 22