Lakes and Groundwater of the Rieti Basin, Central Italy: Hydrochemistry

University of Nevada, Reno

Lakes and Groundwater of the Rieti Basin, Central Italy: Hydrochemistry,

Paleolimnology, and Seismologic Influences

A dissertation submitted in partial fulfillment of the requirements for the degree of

Doctor of Philosophy in Hydrogeology

Carey Claire Archer

Dr. Paula Noble/ Dissertation Advisor

August 2017

THE GRADUATE SCHOOL

We recommend that the dissertation prepared under our supervision by Carey Claire Archer entitled

Lakes and Groundwater of the Rieti Basin, Central Italy: Hydrochemistry,

Paleolimnology, and Seismologic Influences

be accepted in partial fulfillment of the requirements for the degree of

DOCTOR OF PHILOSOPHY

Paula J. Noble, Ph.D., Advisor

David Kreamer, Pd.D., Committee Member

Michael R. Rosen, Ph.D., Committee Member

Simon R. Poulson, Ph.D., Committee Member

Scott Mensing, Ph.D., Graduate School Representative

David Zeh, Ph.D., Dean, Graduate Scool

August, 2017 i

Abstract

A combination of water and sediment core samples were recovered and analyzed to study

the hydrochemistry and paleolimnology of two lakes in the Rieti Basin and adjacent

groundwater springs. Chemical and stable isotopic tracers were used to characterize water

samples and to examine downcore changes in sediment cores. The lakes, Lungo and

Ripasottile (LUN and RIP) have been described as surface outcroppings of the

groundwater table, yet the surface-groundwater interactions have not previously been

investigated in detail. High-discharge springs representing local and regional aquifers

were sampled as a means of comparing and evaluating the chemical data from the lakes,

including Vicenna Riara (VIC), an alluvial aquifer which has hydrochemistry similar to

LUN, and Peschiera (PES), which flows from a major regional carbonate aquifer. Results

from the modern study suggest that LUN is in connection with the alluvial aquifer of the

basin, and that RIP receives substantial input from Santa Susanna Spring (SUS), which is

sourced from a regional carbonate aquifer. SUS has a characteristic chemistry with high

2- - 2+ 2+ SO4 , HCO3 , Mg , and Ca concentrations. This work also shows that in the modern

configuration, RIP has a shorter water retention time, an important consideration for the

paleolimnological portion of this work. The seismic sequence of 2016-2017 provided an

opportunity to study the hydrochemical response of groundwater springs located within a

40-60km region of the epicenters of the main shocks (>6.0 Mw). The springs considered included 3 which had been previously sampled during the modern study; PES, VIC, and ii

SUS. A fourth spring < 5km from the epicenter of the October 26th and 30th 2016 main shocks, Nerea (NER), was added. Both SUS and PES exhibited transient increases in electrical conductivity, alkalinity, and trace metal concentrations immediately after the

Aug. 24th mainshock. The Nerea spring had a similar response to the Oct. 30th mainshock.

The mechanism proposed for the observed increases in elemental concentrations is increased co-seismic pore pressure that cleared faults, pore spaces, and/or long-residence time conduits containing high concentrations of dissolved constituents. Subsequent mainshocks elicited less chemical response after these fluids had already been cleared.

The post-seismic enrichment in the stable isotopic composition of the dissolved inorganic

13 carbon (δ CDIC) may also indicate greater fluid-rock interaction, or alternatively may suggest a second mechanism a play; a potential input of deep-sourced CO2 upwelled along conduits provided by fault movement. All physiochemical and trace element parameters decreased to pre-earthquake values over the weeks-months following the mainshocks. These findings, along with reports of strong ground shaking during mainshocks and aftershocks in the Rieti Basin, led to a re-examination of sediment cores recovered from LUN and RIP within the context of paleoseismicity and the potential for these lakes to record past earthquakes. Event layers that occurred coevally in both lakes were identified according to a compilation of seismic signatures from past studies. After application of the age models and historically documented major (Mw>6.0) earthquakes within 40 km of the lakes, four seismites were proposed with distinct sedimentological iii

and geochemical compositions. The common feature of seismites attributed to the 1298,

1349, and 1703 earthquakes was a homogenite formed either by a rapid influx of groundwater at the sediment-water interface or strong ground shaking causing sediment and pore water mixing and resuspension. All events, including one identified in 1639, contained a geochemical anomaly as well, namely the elements that were identified in

Ch. 1 as indicative of an input of regional groundwater with high sulfate. Complications with earthquake attribution increased during the modern period as human landscape alteration and lake eutrophication may have overprinted earthquake signals. The variable lake level and extent of LUN and RIP through time serve as a backdrop for these and all paleoenvironmental interpretations, so another study was conducted to focus specifically on hydrological regime evolution over the past ~2000 years. The carbon of the inorganic and organic fractions from both sediment cores were studied, emphasizing major transitions at stratigraphic zones. Carbon cycling, evidenced by stable isotopes of carbon in organic matter and carbon and oxygen in carbonate, functioned differently during each historical period and in LUN versus RIP. Though historical records provide some information on lake extent and flooding, proved a useful proxy in discerning water depth, potential sources of inflows and outflows and extent of connection between the two lakes as well as surrounding marsh. iv

Dedication

To Jonathan, and my entire family- thank you for the unending support and guidance. v

Acknowledgements

This work would not have been possible without Paula Noble, Ph.D., who saw the potential in me and was my advocate in every way. Funding support was provided by the

National Science Foundation (award number 1228126) and the Paul Yaniga Scholarship

Fund. I also wish to thank all my Italian colleagues at Tuscia University in Viterbo, the

INGV and La Sapienza in Rome, the ARPA Lazio in Rieti, and Insubria University in

Como, who welcomed me, shared both their data and knowledge with me, and tolerated my mediocre Italian skills. I especially am grateful for the Riserva Naturale di Laghi

Lungo e Ripasottile and their staff, who helped with sample collection and provided housing for me during my visits. I also would like to the staff at LacCore in Minneapolis, and the large Lakes Observatory in Duluth Minnesota for all of their help in core processing, and analyses. At UNR, my progress would not have been possible without the help and support of my fellow PhD student, Kerry Howard, and the Geological Sciences office manager, Marie Russell. Finally, I’d like to thank my dissertation committee for generously lending their thoughts and time to this work. vi

Table of Contents

Abstract …………………………………………………………………………....i

List of Tables ……………………………………………………………………..vii

List of Figures …………………………………………………………………….ix

Introduction ……………………………………………………………………….1

Chapter 1 – Hydrochemical determination of source water contributions to Lake Lungo

and Lake Ripasottile (central Italy)………………………………………………..18

Chapter 2 – Hydrogeochemical response of groundwater springs during central Italy

earthquakes (24 August 2016 and 26-30 October 2016) ………………………….67

Chapter 3 – Lakes as paleoseismic records in a seismically-active, low-relief area (Rieti

Basin, central Italy)………………………………………………………………...122

Chapter 4 – Evidence of palaeohydrological change in the Rieti Basin, Italy, from lake sediment core stable isotope analysis……………………………………………..192

Summary and Recommendations for Future Work……………………………….219 vii

List of Tables CHAPTER 1

Table 1. Geographic and limnologic characteristics of the study sites…………..53

Table 2. Physical and chemical parameters of all waters sampled, including major ion

concentrations in mg L-1. Results of the mixing simulation carried out using

PHREEQC are also included. …………………………………………………….54

Table 3. Stable Isotope Data. Values are expressed in per mil (‰) relative to the

standard indicated in the subscript. ………………………………………………55

Table S1. Rieti historical precipitation data ……………………………Appendix 1

Table S2. Rieti historical temperature data……………………………..Appendix 1

Table S3. Monthly major ion concentrations in Lago Lungo in 2011 collected and

analyzed by ARPA, Lazio………………………………………………Appendix 1

Table S4. Major ion concentrations in Lake Ripasottile in 2010 and 2011 collected

and analyzed by ARPA, Lazio…………………………………………. Appendix 1

CHAPTER 2 Table 1. All physiochemical parameters, Nerea spring…………………………..113

Table 2. All physiochemical parameters, Rieti springs…………………………..114

Table 3. Trace metal concentrations at SUS, PES, VIC, and NER. …………….. 116 viii

Table 4. Pearson correlation matrix for A) NER, B) VIC, C) SUS and D) PES…119

CHAPTER 3 Table 1. Earthquake signals in lake records—an overview. Signals are organized by

their mechanism (physical or chemical)………………………………………….129

Table 2. Past earthquakes by year, magnitude and location…………………… .142

Table 3. Event layer selection and description ………………………………….150

CHAPTER 4

34 13 18 Table 1. Stable isotope results (δ STS, δ Ccarb, δ Ocarb) from discrete core depths in LUN and RIP …………………………………………………………………….215 ix

List of Figures

INTRODUCTION

Figure 1. Study area map showing position of the lakes ……...………………………..1

CHAPTER 1.

Figure 1. Study area map………………………………………………………………..58

Figure 2. Correlograms of lake physical parameters and local meteorological data in a.

LUN and b. RIP………………………………………………………………………….59

Figure 3. Correlograms of major ion concentrations in a. LUN and b. RIP ……………60

Figure 4. Piper diagram of major ion concentration of waters in the Rieti Basin area….61

Figure 5. Oxygen and Hydrogen isotope composition of waters in the Rieti Basin and

Velino River Valley…………………………………………………………………...... 62

Figure 6. Bicarbonate concentration versus carbon isotope composition of dissolved inorganic C in sampled waters…………………………………………………………...63

Figure 7. Sulfur and oxygen isotope composition of sulfate in waters of the study area.64

Figure 8. Conceptual model of groundwater and surface water flow in the northern sector of the Rieti Plain…………………………………………………………………………65 x

CHAPTER 2.

Figure 1. Study Map, including sampling locations, major faults, hydrogeologic units, recharge area boundaries, and main shock epicenters………………………………….105

Figure 2. Time series plots of A) alkalinity, B) Electrical conductivity (EC), C) pH and

D) Temperature from SUS, VIC, PES, and NER…………………………………… ..106

Figure 3. Time series plots of trace elements at SUS, VIC, PES, and NER………….108

Figure 4. δ13C of dissolved inorganic carbon (DIC) in time-series from SUS, VIC, PES,

NER………………………………………………………………………………..…..109

Figure 5. Time series plot of A) δ18O and B) δ34S of sulfate of the Rieti area

springs………………………………………………………………………….………110

Figure 6. Water isotope results of all springs sampled pre- and post- seismic sequence,

plotted with meteoric water lines and rainwater weighted means. ……………………111

Figure 7. Conceptual diagram of pre- and post-seismic groundwater flow conditions at

the macro- and micro- scales within the major carbonate aquifers studied……………112

CHAPTER 3.

Figure 1. Study area map, showing A) active faults, historical earthquake epicenters, and

B) major springs, lake, rivers, and GIS simulated historical lake area extent………….183

Figure 2. Correlation of LUN and RIP using magnetic susceptibility; application of the previously defined LUN age model to RIP…………………………………………….185 xi

Figure 3. A) Section photos of LUN and RIP considered in the investigation, including

XRD results, major earthquake years and age error range bars, and event layers. B) and

C) are detailed views of events 2L and 2R and associated proxies. …………………187

Figure 4. Multi-parameter downcore selected geochemical and isotopic results from A)

LUN and B) RIP………………………………………………………………………189

Figure 3S1. Smear slide photographs of event layers……………………..…Appendix 1

CHAPTER 4.

Figure 1. Downcore plots of organic carbon and carbonate parameters through the major stratigraphic zones in LUN and RIP…………………………………………………216

13 Figure 2. Plots of C:N versus δ Corg in A) LUN and B) RIP. Circle size indicates the

relative weight percent organic carbon………………………………………………217

13 18 Figure 3. Stable isotopes of carbon (δ Ccarb) versus oxygen (δ Ocarb) of carbonate in the

different zones of LUN and RIP, including carbonate bedrock and modern lake water

values ………………………………………………………………………………..218

Figure 4S1 and 4S2. Core photos and selected elemental plots from the laminated

interval in LUN and RIP…………………………………………………..…Appendix 1 1

Figure 1. Study area map, showing location in Italy (top), lakes’ position within the Rieti Basin (middle) and a close-up of lakes Ripasottile and Lungo and surrounding area. Images from Google Earth (2017)

Introduction Lakes are an integration of environmental signals, past and present. The sediment that

accumulates over time and its physical and chemical properties record processes within

the lake and its catchment over time. Paleolimnology utilizes these records of change

with the aim of reconstructing aspects of the system. It takes advantage of chemical and

biological data that are used as proxies in place of direct measurements (Smol 2010).

Compared with other nonmarine depositional environments, lakes have several

advantages; they often are characterized by more constant and continuous sedimentation

rates that yield high-resolution core records and they are particularly sensitive to local

environmental change (Cohen 2003; Birks and Birks 2006). Small lakes, in particular, or

those with surface area <5 km2 respond to catchment-scale perturbations (Perini et al.

2009; Schindler 2009). Lake water itself and associated groundwater also respond to local forcing and provide information on change over a range of time scales. Using chemical indicators as proxies can serve as a link between sediment, lake, and catchment processes, and has the potential to record hydrologic, climate, human land-use, and tectonic changes (Burzigotti et al. 2003; Bertrand 2008; Perini et al. 2009; Ludovisi and

Gaino 2010; Hillman et al 2016).

The Rieti Basin, the study site for this dissertation, is an intermontaine extensional basin located within the Central Apennine region of Lazio, Italy (Cavinato and De Celles

1999). Paleoenvironmental studies based in Italy have the unique benefit of the long

duration of human occupancy and the extensive written records of historical periods.

Archival materials from central Italy and the central Apennine region date to the Roman

Era, spanning more than 2000 years (Ferelli et al. 1990; Leggio and Serva 1991). These accounts have been utilized in studies on paleoclimate and human land-use histories

(Leggio 1995, Calderini et al. 1998; Buntgen et al. 2011, Mensing et al. 2015; 2016).

Studies on paleoenvironment in the Rieti Basin have primarily focused on long-term changes since the Pleistocene (Ferelli et al. 1990; Lorenzetti 1990; Calderoni et al. 1994).

Recent change (past ~2000 years) in the hydrologic regime have yet to be clarified.

Paleolimnological research often uses a multi-proxy approach for testing the dependability of interpretations (Cohen, 2003; Birks and Birks, 2006). Combining a multi-proxy approach sediment investigation with knowledge from available historical documents gives internal constraints on observations. An additional means for comparison is studying two lakes in tandem that occupy the same catchment (e.g.,

Roberts et al. 2016). Lakes Lungo and Ripasottile (LUN and RIP), the two lakes investigated in the work, are less than 2 kilometers apart and are similar in size, catchment area, and surrounding land use (Figure 1). At its inception, this PhD work was framed solely around sediment core analysis, but the research evolved to incorporate modern lake chemical data, groundwater chemical analyses, and a seismological component. By expanding to these areas a more complete portrayal of this dynamic system was achieved.

In Chapter 1, the modern hydrochemical setting of LUN, RIP and local groundwater

springs are described. Paleolimnological studies benefit from a sufficient understanding

of baseline, or modern, conditions, as a means of comparison. The modern groundwater

and surface water interactions in the Rieti Basin have not been studied extensively and

inflows and outflows to LUN and RIP are not so evident. The water level of RIP is

maintained by a hydraulic pumping station at its outflow and a complex network of

irrigation and drainage canals exist around both lakes. Investigation into local groundwater levels at the beginning of field work proved enigmatic due to private ownership of wells and artificial groundwater removal for flood control (Carrara et al.

1994). Instead, chemical and isotopic tracers were used to characterize these lakes, local groundwaters discharging at springs, and their connections. The lakes and rivers of the

Rieti Basin are monitored by the local environmental protection agency, ARPA LAZIO

(Agenzia Regionale Protezione Ambientale del Lazio), primarily for chemical and biological contaminants, though some baseline measurements are recorded routinely. A water chemistry study was conducted on the lakes in 2004 in response to increasing eutrophication, (Franceschini et al. 2004), but was focused on nutrient levels without identification or discussion of source waters. The lakes are part of a nature preserve

(Riserva Naturale di Laghi Lungo e Ripasottile) because they are a critical habitat for wildlife, especially migratory bird species (Sterpi et al. 2013). This enhances water quality concern and the need for an accurate water budget to determine how water, and in

turn nutrients, enter the lakes. Groundwater levels and hydrogeologic structures of the

region were mapped by Martarelli et al. (2008) and the entire central Apennine region by

Capelli et al. (2012). These studies identified general piezometric head contours, spring

discharge, and major aquifers units. Several other studies focused on aquifers of interest, because the Apennine region hosts major karstic aquifers and many serve as drinking

water supplies that are of interest to researchers and policy makers (Civita and Fiorucci

2010; Spadoni et al. 2010; Petitta et al. 2011; Tallini et al 2014). The communication

between local and regional groundwaters and the lakes in the Rieti Basin was not

understood prior to this dissertation, with the hydrochemical investigation included as

Chapter 1.

The central Apennine mountain range is part of a post-collisional orogenic belt in

the Mediterranean-African plate convergence system (Cello et al. 1997; Galadini and

Galli 2000). Within the last year, a segment of the Central Apennine Fault System

(CAFS) ruptured, triggering a seismic sequence punctuated by high magnitude (Mw>

6.0), destructive earthquakes (MMI=10), the largest with magnitude 6.6 (USGS 2017).

The epicenters of these events occurred approximately 30-40 km from the Rieti Basin, and subordinate fault ruptures also occurred along parallel systems within carbonate bedrock bordering the Rieti Basin. These events motivated the immediate (within 2 days of the first mainshock) collection of groundwater samples from springs sampled previously as part of Chapter 1, as the prior chemical characterization of these waters

provided a fortuitous situation to study potential effects post-seismicity. Earthquake

hydrology, the study of the hydrologic response to seismicity, predicts that stress from

the fault rupture itself as well as seismic waves moving through the aquifer medium will

cause changes in hydrogeologic properties (Muir-Wood and King 1993; Liao et al. 2015;

Manga and Wang 2015). Studies on groundwater in carbonate aquifers of the Central

Apennines after recent large earthquakes, specifically the 1997 Colforito Basin and 2009

L’Aquila seismic sequences, found responses in groundwater level, flow rate, spring discharge, and chemistry (Quattrocchi 1999; Carro et al. 2005; Amoruso et al. 2011;

Falcone et al. 2012). The chemical responses were attributed to co-seismic pore pressure increase and greater permeability within the karst flow circuit. The transient increased hydraulic conductivity effectively cleared long mean residence time fractures and/or reservoirs in the host rock where groundwater had more water-rock interaction (Amoruso et al. 2011; Falcone et al. 2012; Galassi et al. 2014). Another mechanism for chemical change post-seismicity was related to aquifers that receive more contribution from mantle-derived CO2. Major earthquakes allow for upwelling of dissolved CO2, where

ruptured faults serve as conduits permitting mixing with regional groundwater

(Quatrocchi 1999; Chiodini et al. 2011; Falcone et al. 2012). Four springs were sampled during the most recent seismic events; three of these are sourced by aquifers with similar size and properties to those studied after the Colforito and L’Aquila events. Chapter 2 describes the physiochemical, stable isotopic composition, and trace metal analyses of

groundwater springs before and several months following the 2016-2017 seismic

sequence.

The Central Apennines have been a site of active extensional tectonics since the

Pliocene Era (Cello et al. 1997). Crustal motion is concentrated on a series of normal

faults in high-topography regions that are capable of producing large (> Mw 6.5)

earthquakes (Galadini and Galli 2000). Italians have kept detailed historical records of

past seismicity that include earthquake effects, perceived intensities, and inferred

epicenters and span more than 1000 years (Galli et al. 2008; Galli and Molin 2014;

Rovida et al. 2016). Paleoseismic studies using trenching evidence across surface faults

have been conducted in the Rieti area, primarily focusing on long-term records including events since the Pleistocene (Michetti et al. 1995). The lake sediment cores collected from LUN and RIP were not initially intended for paleoseismic study, yet the apparent ground shaking in the basin during the 2016-2017 earthquakes as well as the measured response in the regional groundwater chemistry (Chapter 2) prompted re-examination of the cores through the lens of potential earthquake effects. Lake sediment cores have been used as paleoseismic tools in a variety of settings and, compared to sub-marine records, provide an opportunity to identify seismic signatures for larger events in the more recent past (Doig 1986; Monecke et al. 2004; Carrillo et al. 2008; Wilhelm et al. 2016). These seismic signatures in lakes can be from direct and/or indirect earthquake effects. Direct effects include physical disruptions such as seiche waves, turbidity flows, slumping or

other mass-wasting events, while indirect effects are those that are caused by seismically-

induced changes in the lake catchment that are recorded by lake water or sediment

(Bertrand et al. 2008; Boës et al. 2010; Avsar et al. 2014). These indirect signatures

include changes to surface inflows or to groundwater flow as changes in amount or chemistry of groundwater input. The majority of lacustrine paleoseismic studies examine

the direct effects through examination of deformation structures and slump deposits in

high-relief areas where lakes have steep slopes capable of failure (Monecke et al. 2004;

Bertrand et al. 2008). Using core records from shallow or small-volume lakes in paleoseismic investigations, though, has expanded recently, as novel seismites, or earthquake-related sedimentary beds, are described (Boës et al. 2010; Hubert-Ferrari et al. 2012;). The primary physical mechanism for forming earthquake signatures in these lakes is water oscillations or seiche waves from strong ground shaking that cause sediment resuspension and characteristic deposits after resettling (Doig 1986; Carrillo et al. 2008; Avsar et al. 2014). The Rieti Basin lakes are susceptible to strong ground shaking because of their situation within alluvial basin-filling deposits adjacent to normal border faults (Tertulliani 2000). In Chapter 3, four event layers were identified that occur coevally in LUN and RIP and correspond to ages of major past earthquakes. These potentially earthquake-related beds are described and seismic as well as non-seismic explanations for their formation are discussed.

Chapter 4 includes the entire core sequences from LUN and RIP and highlights the

carbon system dynamics and stable isotope records from both cores. The paleoseismic

investigation (Ch. 3) focused on only the upper portion (~400 cm), or most recent

sediment because more accurate historical information is available in this time range. The

entire sequences of LUN and RIP, spanning approximately 2700 and 1300 years,

respectively, have been categorized into zones based on pollen content and gross

stratigraphy (Mensing et al. 2015). This chapter reinvestigates these zones within the

context of hydrological change in and around the lakes. Lake level, mixis, inflows, and

surrounding marsh extent were investigated using the organic and inorganic carbon sediment fractions. The organic C fraction, specifically stable isotopes organic C, organic

C weight percent, and weight percent total nitrogen are proxies used often in combination to determine the source of organic matter (Meyers 2003; Vreca and Muri 2006), paleoproductivity (Routh et al. 2004; Hillman et al. 2016), and past organic matter (OM) recycling and lake level change (Meyers and Ishiwatari 1993; Pueyo et al. 2011). The inorganic C fraction, mainly in the form of calcite in these lakes, was studied using stable

13 18 isotopes of carbon and oxygen (δ Ccarb and δ Ocarb) and weight percent carbonate. These are also common paleolimnological proxies yielding information on source of inorganic

C (McKenzie 1985; Pueyo et al. 2011; Hillman et al. 2016) and hydrologic conditions at the time of calcite precipitation, such as lake water temperature and stratification (Talbot and Kelts 1990; Rosen et al. 1995; Mayer and Schwark 1999). The stable isotope

composition of sulfur (δ34S) was also measured at select depths in both cores and is

included in this chapter because of the close relationship between S and C cycles in lake

sediment (Urban et al. 1999; Holmer and Storkholm 2001; Cohen 2003). Sedimentary

δ34S can yield information on C supply to sediments, sulfate concentrations in lake water,

and OM diagenesis (Nriagu and Soon 1985; Fry et al. 1995; Ding et al. 2016). The proxy

interpretations in Chapter 4 are used to develop a scenario for paleohydrological evolution that may serve as a backdrop for future studies on the Rieti Basin.

Research Objectives:

I. What is the modern hydrochemical regime? Which groundwaters contribute to

the lakes, and can these contributions be identifying chemical distinctions in

the nearby groundwater springs?

II. Do groundwater spring chemical parameters respond to the 2016-2017 central

Apennine seismic sequence? How do chemical and isotopic values of Rieti

Basin springs compare with the Nerea area spring over the post-seismic time

series? Are there patterns in timing or duration of the response that relation to

aquifer location or properties?

III. How can the sediment of small volume lakes in a low-relief area record past

earthquakes? Do intervals in the lakes cores with distinct geochemistry and

sedimentology signify paleoseismicity and do these intervals correspond to

years with major past earthquakes close to the lakes? How might these signals

of paleoseismicity be transformed before, during, or after sedimentation, and

are there alternative mechanisms of formation?

IV. How can changes in the past hydrological regime be reconstructed from stable

isotopic and bulk geochemical data of various carbon sources? The bulk

sedimentological character exhibits clear zonation in both lakes, but what can

be inferred about evolution of the lake system from one zone to the next using

carbon system proxies?

References Agenzia Regionale Protezione Ambientale del Lazio (2017) http://www.arpalazio.gov.it/ Amoruso, A., Crescentini, L., Petitta, M., Rusi, S., & Tallini, M. (2011). Impact of the 6 April 2009 L'Aquila earthquake on groundwater flow in the Gran Sasso carbonate aquifer, Central Italy. Hydrological Processes, 25(11), 1754-1764. Avşar, U., Hubert‐Ferrari, A., Batist, M. D., & Fagel, N. (2014). A 3400-year lacustrine paleoseismic record from the North Anatolian Fault, Turkey: Implications for bimodal recurrence behavior. Geophysical Research Letters, 41(2), 377-384. Bertrand, S., Charlet, F., Chapron, E., Fagel, N., & De Batist, M. (2008). Reconstruction of the Holocene seismotectonic activity of the Southern Andes from seismites recorded in Lago Icalma, Chile, 39 S. Palaeogeography, Palaeoclimatology, Palaeoecology, 259(2), 301-322. Birks, H. H., & Birks, H. J. B. (2006). Multi-proxy studies in palaeolimnology. Vegetation history and Archaeobotany, 15(4), 235-251.

Boës, X., Moran, S. B., King, J., Cağatay, M. N., & Hubert-Ferrari, A. (2010). Records of large earthquakes in lake sediments along the North Anatolian Fault, Turkey. Journal of Paleolimnology, 43(4), 901-920. Büntgen, U., Tegel, W., Nicolussi, K., McCormick, M., Frank, D., Trouet, V., Kaplan, J.O., Herzig, F., Heussner, K.U., Wanner, H. and Luterbacher, J. (2011) 2500 years of European climate variability and human susceptibility. Science, 331(6017), 578-582. Burzigotti, R., Dragoni, W., Evangelisti, C., & Gervasi, L. (2003, December). The role of Lake Trasimeno (central Italy) in the history of hydrology and water management. In IWHA 3rd International Conference, Alexandria, Egypt. Calderini, G., Calderoni, G., Cavinato, G.P., Gliozzi, E., & Paccara, P. (1998) The upper Quaternary sedimentary sequence at Rieti Basin (central Italy): a record of sedimentation response to climatic changes. Palaeogeography, Palaeoclimatology, Palaeoecology 140, 97-111. Calderoni, G., Carrara, C., Ferreli, L., Follieri, M., Gliozzi, E., Magri, D., Narcisi, B., Parotto, M., Sadori, L., & Serva, L., (1994) Palaeoenvironmental, palaeoclimatic and chronological interpretations of a late Quaternary sediment core from Piana di Rieti (central Apennines, Italy). Giornale di Geologia 3 (56/2), 43–72. Capelli G, Mastrorillo L, Mazza R, Petitta M, Baldoni T, Banzato F, Doredana C, Di Salvo C, La Vigna F, Taviani S, Teoli P, (Cartographer) (2012) Carta Idrogeologica del Territorio della Regione Lazio. Carrara, C. Ferreli, L., Guerrieri, L. & Serva, L. (1994). The Rieti Intermountain Basin and St. Francesco d’Assisi. Field Trip Guide From the 32 nd International Geological Congress, Florence, Italy. Vol. 3, D01 to P13 Carrillo, E., Beck, C., Audemard, F., Moreno, E., & Ollarves, R. (2008). Disentangling Late Quaternary climatic and seismo-tectonic controls on Lake Mucubají sedimentation (Mérida Andes, Venezuela).Palaeogeography, Palaeoclimatology, Palaeoecology, 259(2-3), 284–300. Carro, M., De Amicis, M., & Luzi, L. (2005). Hydrogeological changes related to the Umbria–Marche earthquake of 26 September 1997 (Central Italy). Natural hazards, 34(3), 315-339. Cavinato, G. P., & De Celles, P. G. (1999). Extensional basins in the tectonically bimodal central Apennines fold-thrust belt, Italy: response to corner flow above a subducting slab in retrograde motion. Geology, 27(10), 955-958.

Cello, G., Mazzoli, S., Tondi, E., & Turco, E. (1997). Active tectonics in the central Apennines and possible implications for seismic hazard analysis in peninsular Italy. Tectonophysics, 272(1), 43-68. Chiodini G., Caliro S., Cardellini C., Frondini F., Inguaggiato S. & Matteucci F. (2011) Geochemical evidence for and characterization of CO2 rich gas sources in the epicentral area of the Abruzzo 2009 earthquakes. Earth and Planetary Science Letters, 304, 389-398. Civita, M. & Fiorucci, A. (2010) The recharge-discharge process of the Peschiera spring system (central Italy). Aqua Mundi 02019,161-178 Cohen A. S. (2003) Paleolimnology: The History and Evolution of Lake Systems. Oxford University Press, 528 pp. Ding, X., Li, D., Zheng, L., Bao, H., Chen, H. F., & Kao, S. J. (2016). Sulfur Geochemistry of a Lacustrine Record from Taiwan Reveals Enhanced Marine Aerosol Input during the Early Holocene. Scientific reports, 6:38989. Doig, R. (1986). A method for determining the frequency of large-magnitude earthquakes using lake sediments. Canadian Journal of Earth Sciences, 23(7), 930-937. Falcone, R.A., Carucci, V., Falgiani, A., Manetta, M., Parisse, B., Petitta, M., Rusi, S., Spizzico, M. and Tallini, M. (2012) Changes on groundwater flow and hydrochemistry of the Gran Sasso carbonate aquifer after 2009 L’Aquila earthquake. Italian Journal of Geosciences, 131(3), 459-474. Ferreli, L., Parotto, M., Serva, L., (1990) Evoluzione del reticolo idrographico nella Piana di Rieti degli ultimi 4000 anno. Memorie - Società Geologica Italiana 45, 90-910. Franceschini S., Damiani G., Venanzi D., Fabiani S., Schirone B., Mancini L., (2004) Environmental quality of the protected area of Lungo and Ripasottile lakes (Rieti, Italy). Rapporti ISTISAN 04/39. Fry, B., Giblin, A., Dornblaser, M. & Peterson, B. (1995). Stable Sulfur Isotopic Compositions of Chromium-Reducible Sulfur in Lake Sediments Geochemical Transformations of Sedimentary Sulfur, pp 397-410. Galadini, F., & Galli, P. (2000). Active tectonics in the central Apennines (Italy)–input data for seismic hazard assessment. Natural Hazards, 22(3), 225-268. Galassi, D. M., Lombardo, P., Fiasca, B., Di Cioccio, A., Di Lorenzo, T., Petitta, M., & Di Carlo, P. (2014). Earthquakes trigger the loss of groundwater biodiversity. Scientific reports, 4, 6273.

Galli, P., Galadini, F., & Pantosti, D. (2008). Twenty years of paleoseismology in Italy. Earth-Science Reviews, 88(1), 89-117. Galli, P. A., & Molin, D. (2014). Beyond the damage threshold: the historic earthquakes of Rome. Bulletin of Earthquake Engineering, 12(3), 1277-1306. Gliozzi, E & Mazzini, I. (1998) Palaeoenvironmental analysis of Early Pleistocene brackish marshes in the Rieti and Tiberino intrapenninic basins (Latium and Umbria, Italy) using ostracods. Palaeogeography, Palaeoclimatology, Palaeoecology 140, 325–333. Hillman, A., Abbott, M., Yu, J., Steinman, B., & Bain, D. (2016). The isotopic response of Lake Chenghai, SW China, to hydrologic modification from human activity. The Holocene, 26(6), 906–916 Hubert-Ferrari, A., Avşar, U., El Ouahabi, M., Lepoint, G., Martinez, P., & Fagel, N. (2012). Paleoseismic record obtained by coring a sag-pond along the North Anatolian Fault (Turkey). Annals of Geophysics, 55(5), 929-953. Leggio, T. (1995) Trasformazioni del paesaggio dei monti Sabini dall'eta romana al medioevo. In: Leggio, T., Marini, M. (Eds.), Il paesaggio della conca reatina. Problemi ed esperienze di una ricerca multidisciplinare, pp. 51-70. Rieti. Leggio, T., Serva, L. (1991) La bonifica della Piana di Rieti dall'eta romana al Medioevo:influenze sui mutamenti del paesaggio. Not. dell'Enea 25-26, 61-70. Liao, X., Wang, C. Y., & Liu, C. P. (2015). Disruption of groundwater systems by earthquakes. Geophysical Research Letters, 42(22), 9758-9763. Lorenzetti, R., (1990) Lacus Velinus. Per la salubrita dell'aere et per l'abundantia. In: La bonifica dell'agro reatino dall'antico Lacus Velinus alla riorganizzazione del territorio, Milano. Losher, A.J., Kelts, K.R. (1990) Organic sulfur fixation in freshwater lake sediments and the implication for C/S ratios. Terra Nova 1, 253–261. Ludovisi, A., & Gaino, E. (2010). Meteorological and water quality changes in Lake Trasimeno (Umbria, Italy) during the last fifty years. Journal of Limnology,69(1), 174–188 Martarelli L., Petitta M., Scalise A., Silvi A., (2008) [A Cartografia idrogeologica sperimentale della Piana Reatina (Lazio)] [Article in Italian]. In: Mem. Descr. Carta Geol. d’It. LXXXI. (2008): 137-156.

Mayer, B., & Schwark, L. (1999). A 15,000-year stable isotope record from sediments of Lake Steisslingen, Southwest Germany. Chemical Geology, 161(1), 315-337. McKenzie, J. A. (1985). Carbon isotopes and productivity in the lacustrine and marine environment. Chemical Processes in Lakes, John Wiley and Sons, New York New York. 1985. p 99-118. Mensing, S. A., Tunno, I., Sagnotti, L., Florindo, F., Noble, P., Archer, C., Zimmerman, S., Pavón-Carrasco, F.J., Cifani, G., Passigli, S., Piovesan, G. (2015). 2700 years of Mediterranean environmental change in central Italy: a synthesis of sedimentary and cultural records to interpret past impacts of climate on society. Quaternary Science Reviews, 116, 72-94. Mensing, S., Tunno, I., Cifani, G., Passigli, S., Noble, P., Archer, C., & Piovesan, G. (2016). Human and climatically induced environmental change in the Mediterranean during the Medieval Climate Anomaly and Little Ice Age: A case from central Italy. Anthropocene, 15, 49-59. Meyers, PA, & Ishiwatari, R. (1993). Lacustrine organic geochemistry—an overview of indicators of organic matter sources and diagenesis in lake sediments. Organic geochemistry 20(7) 867-900. Meyers, P. A. (2003) Applications of organic geochemistry to paleolimnological reconstructions: A summary of examples from the Laurentian Great Lakes. Organic Geochemistry 34: 261–289 Michetti, A. M., Brunamonte, F., Serva, L., & Whitney, R. A. (1995). Seismic hazard assessment from paleoseismological evidence in the Rieti Region (Central Italy). Perspectives in Paleoseismology”, Association of Engineering Geologists Bulletin, Special Publication, (6), 63-82. Monecke, K., Anselmetti, F., Becker, A., Sturm, M., & Giardini, D. (2004) The record of historic earthquakes in lake sediments of Central Switzerland. Tectonophysics 394(1-2), 21–40. Muir‐Wood, R., & King, G. C. (1993). Hydrological signatures of earthquake strain. Journal of Geophysical Research: Solid Earth, 98(B12), 22035-22068. Nriagu, J. O., & Soon, Y. K. (1985). Distribution and isotopic composition of sulfur in lake sediments of northern Ontario. Geochimica et Cosmochimica Acta, 49(3), 823-834.

Perini, M, Camin, F, Corradini, F, & Obertegger, U. (2009). Use of δ18O in the interpretation of hydrological dynamics in lakes. Journal of limnology 68(2) 174- 182. Petitta, M., Primavera, P., Tuccimei, P., & Aravena, R. (2011). Interaction between deep and shallow groundwater systems in areas affected by Quaternary tectonics (Central Italy): a geochemical and isotope approach. Environmental earth sciences, 63(1), 11-30. Pueyo, J. J., Sáez, A., Giralt, S., Valero-Garcés, B. L., Moreno, A., Bao, R., Schwalb, A., Herrera, C., Klosowska, B. and Taberner, C. (2011). Carbonate and organic matter sedimentation and isotopic signatures in Lake Chungará, Chilean Altiplano, during the last 12.3 kyr. Palaeogeography, Palaeoclimatology, Palaeoecology, 307(1), 339-355. Roberts, N., Allcock, S.L., Arnaud, F., Dean, J.R., Eastwood, W.J., Jones, M.D., Leng, M.J., Metcalfe, S.E., Malet, E., Woodbridge, J. and Yiğitbaşıoğlu, H. (2016) A tale of two lakes: a multi‐proxy comparison of Lateglacial and Holocene environmental change in Cappadocia, Turkey. Journal of Quaternary Science, 31(4), 348-362. Rosen, M. R., Turner, J. V., Coshell, L., & Gailitis, V. (1995). The effects of water temperature, stratification, and biological activity on the stable isotopic composition and timing of carbonate precipitation in a hypersaline lake. Geochimica et cosmochimica acta, 59(5), 979-990. Routh, J., Meyers, P., Gustafsson, Ö., Baskaran, M., Hallberg, R., & Schöldström, A. (2004). Sedimentary geochemical record of human–induced environmental changes in the Lake Brunnsviken watershed, Sweden. Limnology and Oceanography,49(5), 1560–1569 Rovida, A., M. Locati, R. Camassi, B. Lolli, and P. Gasperini (Editors) (2016). CPTI15, the 2015 Version of the Parametric Catalogue of Italian Earthquakes, Istituto Nazionale di Geofisica e Vulcanologia, Schindler, D.W. (2009) Lakes as sentinels and integrators for the effects of climate change on watersheds, airsheds and landscapes. Limnology and Oceanography 54, 2349–2358. Smol, J. P. (2010). The power of the past: using sediments to track the effects of multiple stressors on lake ecosystems. Freshwater Biology, 55(s1), 43-59.

Spadoni, M., Brilli, M., Giustini, F., & Petitta, M. (2010). Using GIS for modelling the impact of current climate trend on the recharge area of the S. Susanna spring (central Apennines, Italy). Hydrological processes, 24(1), 50-64. Sterpi, L., Sterpi, M., Pastorelli, S., Cento, M., & Sarrocco, S. (2013). Nidificazione di Falco di palude Circus aeruginosus nella Riserva Naturale Regionale dei Laghi Lungo e Ripasottile (Lazio, Italia Centrale). Alula, 20(1-2), 149-151. Talbot, M. R., & Kelts, K. (1990). Paleolimnological Signatures from Carbon and Oxygen Isotopic Ratios in Carbonates, from Organic Carbon-Rich Lacustrine Sediments: Chapter 6, 99-112. Tallini, M., Falcone, R. A., Carucci, V., Falgiani, A., Parisse, B., & Petitta, M. (2014). Isotope hydrology and geochemical modeling: new insights into the recharge processes and water–rock interactions of a fissured carbonate aquifer (Gran Sasso, central Italy). Environmental earth sciences, 72(12), 4957-4971. Tertulliani, A. (2000). Qualitative effects of local geology on damage pattern. Bulletin of the Seismological Society of America, 90(6), 1543-1548. Urban, N. R., Ernst, K., & Bernasconi, S. (1999). Addition of sulfur to organic matter during early diagenesis of lake sediments. Geochimica et cosmochimica acta, 63(6), 837-853. Vreca, P., & Muri, G. (2006). Changes in accumulation of organic matter and stable carbon and nitrogen isotopes in sediments of two Slovenian mountain lakes (Lake Ledvica and Lake Planina), induced by eutrophication changes. Limnology and Oceanography, 51(1, part2), 781–790. Wilhelm, B., Nomade, J., Crouzet, C., Litty, C., Sabatier, P., Belle, S., Rolland, Y., Revel, M., Courboulex, F., Arnaud, F. and Anselmetti, F. (2016) Quantified sensitivity of small lake sediments to record historic earthquakes: Implications for paleoseismology. Journal of Geophysical Research: Earth Surface, 121(1), 2–16.

CHAPTER 1 Hydrochemical determination of source water contributions to Lake Lungo and Lake Ripasottile (central Italy)

Claire Archer, Paula Noble, David Kreamer, Vincenzo Piscopo, Marco Petitta, Michael R. Rosen, Simon R. Poulson, Gianluca Piovesan, Scott Mensing

ABSTRACT

Lake Lungo and Lake Ripasottile are two shallow (4-5 m) lakes located in the Rieti Basin,

central Italy, that have been described previously as surface outcroppings of the

groundwater table. In this work, the two lakes as well as springs and rivers that represent

their potential source waters are characterized physio-chemically and isotopically, using a

combination of environmental tracers. Temperature and pH were measured and water

samples were analyzed for alkalinity, major ion concentration, and stable isotope (δ2H,

δ18O, δ13C of dissolved inorganic carbon, and δ34S and δ18O of sulfate) composition.

Chemical data were also investigated in terms of local meteorological data (air temperature,

precipitation) to determine the sensitivity of lake parameters to changes in the surrounding

environment. Groundwater represented by samples taken from Santa Susanna Spring was

2- 2+ shown to be distinct with SO4 and Mg content of 270 and 29 mg/L, respectively, and heavy sulfate isotopic composition (δ34S=15.2‰ and δ18O=10‰). Outflow from the Santa

Susanna Spring enters Lake Ripasottile via a canal and both spring and lake water exhibits the same chemical distinctions and comparatively low seasonal variability. Major ion

concentrations in Lake Lungo are similar to the Vicenna Riara Spring and are interpreted

13 to represent the groundwater locally recharged within the plain. The δ CDIC exhibit the same groupings as the other chemical parameters, providing supporting evidence of the

13 2 source relationships. Lake Lungo exhibited exceptional ranges of δ CDIC (±5‰) and δ H,

δ18O (±5 ‰ and ±7 ‰, respectively), attributed to sensitivity to seasonal changes. The

hydrochemistry results, particularly major ion data, highlight how the two lakes, though

geographically and morphologically similar, represent distinct hydrochemical facies.

These data also show a different response in each lake to temperature and precipitation

patterns in the basin that may be attributed to lake water retention time. The sensitivity of

each lake to meteorological patterns can be used to understand the potential effects from

long-term climate variability.

INTRODUCTION

The low water volume and lack of buffering potential in some small lakes (surface area

<0.5 km2) can make them particularly sensitive to changes in local environmental conditions and a potential resource for paleohydrologic reconstruction (Wetzel, 2001;

Cohen, 2003). However, the contribution of groundwater to the overall water budget of small lakes is often overlooked or underrepresented (Wetzel, 2001; Perini et al., 2009;

Rosenberry and Lewandowski, 2015). Lake Lungo and Lake Ripasottile, located in the

Rieti Plain (central Italy), are small water bodies that have experienced recent

eutrophication (Franceschini et al., 2004). These lakes are of particular interest because of

their location within a nature reserve, Riserva Naturale dei Laghi Lungo e Ripasottile

(hereafter referred to as Riserva), which was established in 1985 (Franceschini et al. 2004).

Previous studies on the hydrology of the Rieti Plain suggest a combination of inputs from

groundwater and drainage from manmade irrigation channels (Franceschini et al. 2004;

Martarelli et al. 2008). The unknown magnitude and seasonality of these surficial inputs

create a challenging scenario for quantifying the lakes’ water budgets. Major ions and

environmental isotopes as tracers can provide a means for building a conceptual model of

water contributions and connectivity. This methodology has been previously used to study

neighboring groundwater systems in the intermontane basins of the central Apennines

(Petitta et al., 2011, Carucci et al., 2012; Sappa et al. 2012; Tassi et al, 2012; Tallini et al.,

2014), as well as groundwater-surface water interactions in other regions (Gourcy and

Brenot 2011; Sacks et al., 2014).

The aim of this study is to gain a better understanding of hydrologic connectivity between the two lakes in the Riserva, to the nearby surface water, as well as to local and regional aquifers. The definition of inputs into each lake will serve two major purposes.

First, it will allow for more informed conservation management practices, particularly in response to a changing climate. To this end, a hydrochemical approach using major ion concentrations and stable isotopes was chosen in this system where physical measurements of surface inflow and outflows are a less informative method. This method can show temporal variations when no physical monitoring equipment is in place, as in the case of

these lakes. It also can define the commonly overlooked groundwater influence on lake

water budget without directly studying subsurface hydrology. The second purpose of this

study is to use the modern characterization of inputs to make informed interpretations of

paleolimnology of this system and regional paleoclimate going back three millennia, as

part of an ongoing study of lake cores from the Riserva (Mensing et al., 2015). Before the

interpretation of geochemical, paleoecological and isotopic signals in the cores, it is

necessary to determine what factors in the lake and catchment area drive modern changes

in these parameters.

The past few decades in the Mediterranean region, and particularly central Italy, were

marked by decreases in both total precipitation and number of wet days (Brunetti et al.

2012; Carucci et al. 2012). Regional karstic aquifers in this region, hosted in the carbonate ridges, have already experienced decreases in storage and associated spring discharge (Di

Matteo and Dragoni, 2006; Fiorillo et al., 2015). The Santa Susanna Spring, for instance,

discharging the basal flow of the regional carbonate aquifer located to the northeast of the

study area (Fig. 1), has experienced a 30% reduction in recharging water over the past

several decades (Spadoni et al. 2010). The warming trend observed in Italy over the last

approximately 40 years, especially the significant increases in summer temperatures

(Fioravanti et al. 2015), combined with agricultural demand for water in the basin

(Massarutto, 1999) will also likely impact water supply to these lakes. Changes in

precipitation and air temperature associated with local impacts of predicted global climate

change are the primary parameters of concern as they directly and indirectly affect water supply. Without access to groundwater data, this study was limited to postulating the impacts of meteorological trends and looking for evidence of a limnological response. The relationship between these trends and the within-lake conditions was assessed by examining correlations of lake chemical and physical measurements with local meteorological data for the same period. This lake-sensitivity analysis may also help predict how these protected and significant water bodies will react to future shifts in temperature or precipitation. Informed management of these lakes, for wildlife conservation, water quality and water supply, is dependent on an overall better understanding of hydrology in the Rieti Basin.

Regional geologic, hydrogeological, and climatic setting

The Rieti Basin is an intermontane depression located in the central Apennine thrust system

(Fig. 1). Recently uplifted marine sediments, comprised primarily of limestone, characterize the geology of this region. During the Miocene, compressional tectonics forced pelagic carbonate sequences upwards above sea level (Cosentino et al., 2010).

Extensional tectonics dominate in the region since the Upper Pliocene - Lower Pleistocene, forming basins such as Rieti that subsequently aggraded with clastic lacustrine and alluvial deposits (Cavinato and De Cellis, 1999). The Rieti Basin remains seismically active and deep normal faults mark the boundaries between the Meso-Cenozoic carbonate ridges and

the recent continental deposits on the northern, western and eastern edges of the plain. The

recent plain-filling alluvial unit ranges in thickness from 80 m on the eastern side of the

plain to over 400 m on the western side (Martarelli et al., 2008).

Carbonate ridges hosting regional aquifers (Boni et al., 1986) characterize the

hydrogeology of the central Apennines. Discharge occurs at their contact with less

permeable units and yields springs with high flow (up to 18 m3 s-1). Normal faults at the

border of the plain separating these two units also serve as a conduit for groundwater to

travel to the surface and discharge at springs. The hydrogeology of the Rieti Plain, studied

and mapped by Martarelli et al. (2008) and Capelli et al. (2012), is characteristic of the

Apennine tectonic basin flow regime, where the regional flow system is hosted by the surrounding carbonate bedrock and a shallower layered aquifer exists within the more recent alluvium deposits filling the plain (k). These plain-filling deposits are low permeability Pleistocene- Holocene fluvial and lacustrine sediment interbedded with gravelly colluvium and occasional peat layers. The makeup of this unit creates a scenario where local recharge is relatively limited if compared with that of the surrounding carbonate aquifers and groundwater flow mainly occurs in the coarser grained units or at the contacts between the beds (Martarelli et al. 2008; Capelli et al., 2012). At the

northeastern edge, the Santa Susanna Spring (hereafter referred to as SUS) (Tab. 1)

represents the highest discharge spring of those in direct contact with the plain as well as

the base-flow of the regional aquifer (Zuppi and Bortolami, 1982). This regional aquifer is

hosted in the Terminillo and Reatini mountain carbonate aquifer units to the east and the

spring emerges at the intersection of two normal faults bounding this unit (Martarelli et al.,

2008; Spadoni et al., 2010). Mesozoic evaporites comprise a portion of this carbonate unit, underlain at the base by Triassic dolomites (Capelli et al., 2012).

The Vicenna Riara Spring (hereafter referred to as VIC) (Tab. 1) is distinct among springs included in this study because of its location near the center of the plain, about 0.8 km SSE of Lake Lungo. Water flows to this spring along a buried fault and through preferential flow-paths within the Pleistocene conglomeratic alluvial deposits that lie in the central- eastern part of the plain (Martarelli et al., 2008).

The Velino River (Fig. 1) represents the major portion of surface water flowing into the

plain (35 m3 s-1, Tab.1; the Turano and the Salto Rivers enter the plain at 2 m3 s-1 and 1 m3

s-1, respectively, and converge with the Velino River). Velino’s waters are sourced

upstream to the southeast from the regional carbonate aquifer. Here, at its headwaters, the

river acts as a linear spring, discharging groundwater from the regional aquifer (Petitta,

2009). Downstream of this, and within the Rieti Plain, the river is mostly a losing stream,

with its water contributing to the alluvial aquifer of the plain.

Located about 8 km upstream from the Rieti Plain in the Velino River Valley, the San

Vittorino Plain holds several springs that discharge in the same manner as SUS, where

large discharges occur at the contact between carbonate units and plain-fill deposits. Two

of these springs were sampled in this study, the Peschiera and the Terme di Cotilia springs

(hereafter referred to as PES and COT, respectively) (Fig. 1). These localities were included because of their large discharge (Tab. 1) and potential contribution to the waters of the Rieti Plain. Previous hydrochemical investigations were also conducted on these waters, allowing the limited number of samplings from this study to fit into a greater context. Local research has also focused on Lake Paterno, a sinkhole lake located in the S.

Vittorino Plain with a similar geologic setting to Lake Lungo and Ripasottile (Tassi et al.

2012). Many of the previous studies on groundwater in this area focused in part on how tectonic discontinuities allow deep, mineralized fluids to migrate up and mix with regional aquifer water, as in the case of COT, and to mix with lake water, as in the case of Lake

Paterno (Zuppi and Bortolami, 1982; Petitta et al., 2011; Tassi et al., 2012). PES, with the highest discharge in the study area (i.e., 18 m3 s-1), represents a major drinking water supply to Rome. The groundwater contribution to the spring is mainly from the aquifer hosted within the uplifted carbonate shelf deposits to the ESE, in bedrock consisting of Triassic-

Paleocene limestone located in the Giano-Nuria mountains (Petitta, 2009; Civita and

Fiorucci, 2010).

The surface water hydrology of the Rieti Plain is complicated by the extensive human alteration of the landscape and waterways (Franceschini et al., 2004; Martarelli et al.,

2008). Prior to human modification, the plain was occupied by one large lake, Lacus

Velinus, until approximately the Roman Era, when a combination of human land- reclamation efforts and climate change decreased the area and water level of the lake. The

area was modified to its current configuration in the 1930’s, when canal construction and

drainage efforts removed excess surface water (Calderoni et al., 1994; Martarelli et al.,

2008). Presently, a network of canals and ditches drains water from nearby springs and agricultural land into wetlands surrounding both Lake Lungo and Ripasottile (hereafter referred to as LUN and RIP). RIP also receives inflowing water from an artificial channel from the nearby SUS. This water flows from the spring then is diverted through a fish farm before flowing into the RIP. The magnitude of SUS as a water source and its impact on

RIP’s water balance was not previously studied. Another artificial channel connects RIP to the wetlands on the west end of LUN. The level of the lakes is maintained by a pumping station located at the northwest corner of RIP where effluent is directed into the Velino

River. This prevents the lake depths from increasing and maintains a scenario where the shallow lakes’ water columns continue to mix throughout the year without a prolonged period of stratification.

The Rieti Basin experiences climate with typical Mediterranean precipitation patterns, characterized by the majority of precipitation occurring in autumn and winter

(Supplementary Tab. 1), with relatively arid summers (Combourieu-Nebout et al., 2015).

Within Italy, the Rieti Plain lies in the climate region categorized as the warm temperate mountainous zone based on the continentality (seasonal shifts) of annual temperatures, the mean annual precipitation, and the potential evapotranspiration (Costantini et al., 2013).

The mean monthly temperature in this area varies between 4°C in January and 21°C in July

(Leone, 2004). Studies based on long-term records prior to the 21st century found the

annual mean precipitation at 1117 mm (Leone, 2004), though the mean annual precipitation

from 2003-2015 collected from gauges in Rieti was 1013 mm (Supplementary Tab. 1).

METHODS

Data compilation

Monthly data collected by the regional environmental protection agency, ARPA Lazio, from 2010-2015 regarding lake temperature, pH, alkalinity, conductivity, and major ion concentrations were compiled. Monthly mean temperature data and total precipitation data over the sampling period were extracted and tabulated from annual records published on

the website of the Ufficio Idrografico e Mareografico, Lazio (Supplementary Tabs. 1-4).

Another dataset of lake physio-chemical parameters collected by ARPA Lazio between

2010 and 2015 was also included in the present study with data collected between 2010-

2015. These data, although not continuous, serve as a high-resolution record of near

monthly lake temperature, alkalinity, pH, and conductivity over several years. Major ion

concentration data for the two lakes was available monthly for one year (2011=.

Sample collection

Field surveys and sample collection for this study were conducted five times over a 15- month period during the following months: June 2014, July 2014, February 2015, May

2015 and September 2015. At each lake locality (1 and 2 in Fig. 1), samples were collected

from the center of the lake at the median depth in the water column using a Van Dorn

sampler (Priscu and Dodds, 1988). Physical parameters (temperature, pH, electrical

conductivity) were measured on-site using a Yellow Springs Instrument (YSI®) 85 multi-

parameter sensor and an Oakton® pH meter. Water samples were filtered on-shore using a

0.45 µm Millipore filter, then alkalinity was measured by titrating with 1.6N sulfuric acid to a pH endpoint of 4.5. Replicate analysis of samples for alkalinity showed that the standard deviation from the mean was within 5%. The remaining filtered samples were stored in high-density polyethylene (HDPE) bottles with no head space, then one aliquot was placed in a 10-mL glass vial for stable isotopic analysis of oxygen and hydrogen in water. During subsequent storage and transport of samples, a temperature of 4.5°C was maintained.

The Velino River (hereafter referred to as VRV; 3 in Fig. 1) was sampled 1 meter from shore, at the depth halfway between the channel surface and bottom. At each spring locality

(4. 5, 6 and 7 in Fig. 1), samples were collected as close as possible to the discharge point.

All subsequent procedures for sample collection and physical parameter measurement were identical to those described above for the lake localities.

Lake surveys

During the May 2015’ field season, surveys of LUN and RIP were conducted to quantify

lake bathymetry and to explore the potential for groundwater flow or seepage at the lake

perimeter. A handheld sonar system (Hawkeye H22PX) was used to determine water depth

along transects, and a GPS (global positioning system) unit was used to record position.

The depth data from these transects were used to calculate lake volume by creating a

triangulated irregular network (TIN) in ArcGIS 10.4 (ESRI®). The YSI sensor was also

used to survey temperature at 24 points along the LUN perimeter (<1 m water depth) and

48 points at RIP perimeter. Volumetric flow rate in the channel connecting the two lakes

and in the artificial channel originating at SUS and entering RIP was determined by

measuring channel cross sectional area then multiplying by average flow velocity.

Analyses

Major ion concentrations of all samples were determined using ICP-MS and ICP-OES at the Western Environmental Testing Laboratory in Sparks, NV, USA. Samples were prepared by trace metals digestion using U.S. EPA method 200.2 (Martin et al., 1994).

Chloride and sulfate were analyzed using EPA method 300.0 (Pfaff, 1993) and all other ions using EPA method 200.7 (Maxfield and Mindak, 1985). The accuracy was controlled by using internal laboratory standards prior to every sample run, after every tenth sample, and immediately following a sample run. Replicate analysis of one field duplicate, one

analytical duplicate sample and two laboratory standards per sample collection date

showed that the precision was within 5%.

Waters were prepared for isotopic analysis by precipitation of dissolved sulfate as BaSO4

34 following the method of Carmody et al. (1998). BaSO4 precipitates were analyzed for δ S

using V2O5 as a combustion aid, and followed the methods of Giesemann et al. (1994) and

18 Grassineau et al. (2001). BaSO4 precipitates were analyzed for δ O following the method of Kornexl et al. (1999). The analytical error, estimated by replicate analysis, was ±0.2‰ and ±0.4 ‰ for δ34S and δ18O of dissolved sulfate, respectively. Water samples were

prepared for isotopic analysis of dissolved inorganic carbon by precipitation as SrCO3 after the method of Usdowski et al. (1979), and then analyzed using the method of Harris et al.

18 (1997), with analytical error within ±0.2‰. Waters were analyzed for δ O using the CO2

2 - H2O equilibration method of Epstein and Mayeda (1953), and for δ H using the method

of Morrison et al. (2001). The analytical error of these measurements was ±0.1‰ and

±1.0‰ for δ18O and δ2H, respectively. All stable isotope analyses were carried out at the

University of Nevada, Reno. All values are reported using delta notation (δ‰), and the

standards used were V-SMOW for oxygen and hydrogen, V-PDB for carbon and V-CDT

for sulfur.

Descriptive statistics of the local meteorological data (mean temperature and total

precipitation; Supplementary Tabs. 1 and 2) and records of lake physical and chemical

parameters were calculated initially to confirm a normal distribution of the data.

Correlation between these data sets was then investigated using the Pearson product- moment correlation. This analysis, and all r-values, were calculated using the Excel Data

Analysis tool pack and values are displayed in correlograms. The significance of these correlations was determined using the table of critical values for Pearson’s r. The significance level of correlations is indicated by the degree of shading in the correlograms.

Saturation indices for calcite, anhydrite and gypsum were calculated using the program

PHREEQC for Windows, version 2.16.02 (Parkhurst and Appelo, 1999). PHREEQC was also used to simulate mixing between select waters.

RESULTS

Physical-chemical parameters

Physical-chemical data collected during the sampling period (Tab. 2) were compiled with meteorological and physical-chemical data for the two lakes that were collected over a longer time period by ARPA, Lazio. Major ion concentration data collected and analyzed by ARPA were also available monthly for 2011. The data collected during this study (2014,

2015) fall within the range of values observed for previous years (2010-2015). The local meteorological data (mean monthly temperature, total monthly precipitation;

Supplementary Tabs. 1 and 2) were used to test the sensitivity of the two lakes to changes in local climate factors over short (monthly) time scales.

Lake surveys yielded updated lake bathymetric data that were used to calculate lake

volume. The maximum depth in LUN was 4.5 m and in RIP was 4.2 m. These results

contrasted with previous studies on the lakes’ physiography that found a maximum depth

in LUN of 7.25 m and in RIP of 7.5 m (Riccardi, 2006). Physical surveys of the perimeter

of LUN yielded no measurable flowing water. These surveys also found that for the

measured locations around the perimeter, LUN temperature variations were ±0.5°C. In RIP

the temperature was 8°C colder (10°C versus 18°C) at the entrance of the SUS- diverted

channel inflow, but 18±0.7°C along the rest of the perimeter. Physical surveys conducted

during July 2014 and May 2015 indicated a flow velocity of close to zero in the channel

between the two lakes. The volumetric flow rate of the diverted channel entering RIP was

approximately 1 m3 s-1 during May 2015 and September 2015. Lake surface area and

volume in LUN was calculated at 413,000 m2 and 870,000 m3, respectively (Tab. 1). In

RIP, surface area is 626,000 m2 and volume is 1,000,000 m3.

Correlation matrices (Figs. 2 and 3) display how the physical and chemical parameters may

be influenced by meteorological data (temperature, precipitation) at lakes LUN and RIP.

The stronger r-values (those approaching 1 or -1), indicated by the degree of shading in the box, were observed between LUN water temperature and average monthly air temperatures

(0.92). Other notably high correlations (r>0.75, r<-0.75) were found between LUN nitrate concentration and sulfate, calcium, and air temperature, as well as RIP sulfate and potassium concentration. RIP shows less correlation between air and lake water

temperature (r= 0.64) than LUN, but a potentially negative correlation between air

temperature and alkalinity (r= -0.7). In LUN, temperature compared to nitrate, sulfate, and

calcium concentrations exhibited high values, while in RIP the only ion that may correlate

with temperature was nitrate.

Results of the PHREEQC geochemical modeling of the saturation index of the lakes and

spring waters with respect to calcite (SIcalcigte) shows that the lakes (RIP, LUN) had

consistently positive values throughout the sampling period (Tab. 2). The other sites were

variable throughout the year, ranging from -0.44 at COT (September 2015) to +1.28 at RIP

(February 2015). All samples were undersaturated with respect to gypsum and anhydrite

(SIgypsum, SIanhydrite <-1).

The major ion data collected and analyzed during 2014-2015 (Tab. 2, Fig. 4) were used in

the delineation of two main hydrochemical facies in the study region: Ca-HCO3- type water

- - and Ca-HCO3 -SO4. SUS and RIP represent the Ca-HCO3 -SO4 facies, whereas VIC and

- LUN represent the Ca-HCO3 facies. VIC and LUN are also distinct in their slightly

elevated Na+ and K+ relative to the other waters. Additionally, a third facies can be

2+ categorized as the high-TDS Ca -HCO3- type exemplified by COT, with average TDS

values almost four times that of the other samples.

Stable isotopes

The δ2H and δ18O composition of the waters sampled ranged from -58 to -37‰ and -9.0 to

-5.3‰, respectively (Fig. 5 and Tab. 3). Samples generally fall along the central Italian

Meteoric Water Line (cIMWL) calculated by Longinelli and Selmo (2003), with only a few notable exceptions. The major groupings of sites by their position along the local meteoric water line are: i) the samples of three waters with origins upstream from the basin in the Velino River Valley (VRV, PES and COT); ii) SUS and RIP; iii) LUN and VIC.

LUN shows the largest variability of any of the localities’ samples. Values of precipitation from rain gauges were also included (locations given in Fig. 1, from Spadoni et al., 2010;

Tab. 3) at two different elevations within the study area; the first within the Rieti Plain at

378 m asl and the second along the carbonate ridge to the east at 1375 m asl.

13 - 2 The δ CDIC versus HCO3 concentration shows similar groupings to that of the δ H and

δ18O data (Tab. 3 and Fig. 6). Samples collected from VRV and PES group together with

13 - more enriched δ CDIC and high concentrations of HCO3 . RIP and SUS again plot together,

as do LUN and VIC during February and July 2014. COT plotted separate from all other

13 sites, with comparatively enriched δ CDIC values of +2.9 to +4.7‰. Little variability was

13 seen in δ CDIC during the different sampling periods at all localities, with the exception of

LUN. The range in values from LUN was -8.9 to -13.9‰, with the most depleted values in

July 2014 and an enrichment observed during May and September 2015. The amount of

- DIC in each water, represented by the HCO3 concentration, did not change substantially

over the sampling period.

34 18 δ Ssulfate and δ Osulfate ranged from -9.9‰ to +17.2‰, and from -0.7‰ to +10.8‰,

respectively. The sites that exhibited the most temporal variability were VIC, LUN, and

34 VRV (Tab. 3). The range of δ Ssulfate values was 25‰, 3.3‰, and 5.4‰, respectively in

VIC, LUN and VRV. The comparison of values in this study with typical ranges of sulfate

34 18 stable isotope values (Fig. 7) are displayed using the mean values of δ Ssufate and δ Osulfate

at each locality for all the sampled months.

DISCUSSION

Hydrologic provenance

In the central Apennines, recharge elevation is the primary control of oxygen and hydrogen

stable isotopic values of groundwater (Zuppi and Bortolami, 1982; Longinelli and Selmo,

2003, Tallini et al., 2014). The waters sampled located in the Velino River Valley, i.e.,

PES, COT and VRV, have δ18O and δ2H values that support a high-elevation recharge area.

The recharge area for PES is the Velino-Nuria Mountains to the southeast, extending from

approximately 415 m to 2000 m asl (Petitta, 2009; Civita and Fiorucci, 2010).

13 The relatively enriched δ CDIC values of PES, VRV and COT can be attributed to water- rock interactions between groundwater and carbonate shelf deposits and localized dolomites making up the hydrogeological unit hosting the aquifer (Petitta, 2009; Petitta et al., 2011). The VRV is predominantly a gaining stream in the upstream Velino Valley adjacent to PES and COT, where discharge originates from linear springs (Petitta, 2009).

The river was sampled, however, downstream in a location where it feeds groundwater of

the plain (Martarelli et al., 2008). COT, discharging on the northern edge of the plain, receives groundwater from both the regional carbonate aquifer and gas-rich waters uprising along tectonic discontinuities (Zuppi and Bortolami, 1982; Petitta, 2009). The input of mineralized water from a deep source with extensive water-rock interactions give to this

13 spring its distinct high-dissolved ion signature, low pH, and significantly enriched δ CDIC

values (Tab. 3). Unlike Lake Paterno, as shown by Tassi et al. (2012), however, the lakes

of the Rieti Plain do not receive significant contributions from this mineralized spring

water. This could be because of the difference in physical proximity to COT and related

waters or because of the difference in lake depth (Lake Paterno maximum depth is >50 m)

prevents communication of the Rieti Plain lakes with the deep aquifer system.

The recharge occurring in the carbonate unit of the Reatini Mountains to the east of the

plain that supplies SUS extends over a broader area and wider range of elevations

compared with the springs in the Velino River Valley (Fig. 1, Spadoni et al., 2010),

producing the slightly more positive values of SUS compared to these upstream spring

samples. The two major springs sampled within the plain, VIC and SUS receive water from

two hydrochemically differentiable sources. Evidence for this is seen in the divergence in

major ion concentrations (Fig. 4), the stable isotope values (Figs. 5-7), and the difference

18 2 in flow rate (Tab. 1). The δ O and δ H values of VIC water are heavier than samples from

the other springs, indicating a predominant local recharge component with contributions

also from infiltration through the alluvial-conglomerate fan unit at the eastern edge of the

2- plain (Fig. 8). The VIC hydrochemistry showed a lack of elevated SO4 concentrations and

18 34 13 absence of the δ Osulfate, δ Ssulfate, and δ CDIC signature that characterizes water from the regional carbonate aquifer. These data indicate that there is not a significant contribution from the regional carbonate aquifer to this spring. Our data indicate that the low- permeability lacustrine and fluvial deposits at the lowest elevations of the plain prevent notable water exchange between this unit and the regional carbonate aquifer unit.

18 34 The δ Osulfate and δ Ssulfate values of SUS are similar to those of COT, indicating the

possibility of a related sulfate source (Fig. 7). A likely common source is groundwater from

the same hydrogeological unit, the Mt. Terminillo thrust belt (Petitta, 2009; Spadoni et al.,

2010), where Triassic marine evaporites present in the Mesozoic sedimentary sequence are

18 34 the dominant source of sulfate (Petitta et al., 2011). The mean δ Osulfate and δ Ssulfate of

PES, VRV, RIP and LUN also fall within the field of typical values for waters influenced

by dissolution of marine Mesozoic evaporites (Fig. 7). The water sampled that falls outside

this range, VIC, was in the range of values typical of terrestrial evaporites or sulfate formed

by oxidation of reduced sulfur compounds in soil (Krouse and Mayer, 2000).

The analysis of waters from the two lakes, LUN and RIP, provides information on the

relationship between surface and groundwater. Previous work noted the connection

between the lakes and groundwater (Franceschini et al., 2004; Martarelli et al., 2008), but

they do not specifically examine the mechanism or the provenance. The results of this study

found that chemical and physical parameters of RIP group closely to SUS, especially

2- 2+ conductivity, alkalinity, and dissolved SO4 and Mg concentrations. The similar grouping of SUS and RIP also can be seen in stable isotope values (δ2H, δ18O δ13C and δ34S, Figs.

5-7), demonstrating the magnitude of water flowing from the artificial channel diverting

SUS water into RIP. The concentrations of the conservative ions Na+, K+, and Cl- (Tab. 2), however, are slightly higher in RIP than SUS and the δ18O and δ2H are heavier, suggesting dilution by another minor component. The channel connecting LUN to RIP was not flowing or contributing significantly during the sampling period, exhibited by the lakes’ markedly different chemistries in most parameters. The high water table in this part of the plain, varying between 1 and 4 m below the ground surface (Martarelli et al., 2008) links both lakes to the shallow aquifer within the plain (Fig. 8). The relative importance of this groundwater through the low-permeability sediments can be estimated by examining the chemical data and modeling the mixing of sampled waters using PHREEQC. The results show that the chemistry of RIP can be produced by mixing 80% SUS and 20% LUN water

(Tab. 2). LUN was used as a representative of the shallow alluvial aquifer because this water was not sampled and because LUN receives the majority of its water from local recharge and the shallow plain aquifer. LUN water chemistry was similar to the alluvial spring, VIC, in most parameters but the major ion concentrations are slightly greater than the spring. This suggests lake water evaporation and/or a minor contribution to the plain aquifer of the SUS-type groundwater. Field surveys of the LUN perimeter temperature and

flow conducted during this study found an absence of relatively colder water or anomalies.

This was evidence for the lack of a point source of surface flow or groundwater

contribution to the lake. The piezometric data collected by Martarelli et al. (2008) showed

that the groundwater table decreases from SUS in the direction of LUN, so a percentage of this groundwater may contribute to the plain aquifer but the low permeability of the lacustrine unit surrounding LUN likely prevents a large conveyance.

Effect of season on waters sampled

LUN is an interesting case because two of the samples (May and September 2015) have

13 δ CDIC values that group with SUS and RIP, where the other samples (June and July 2014,

February 2015) plot separately (Fig. 6). A scenario where groundwater represented by SUS

is feeding into LUN during this period does not agree with the stable isotopic composition of δ18O and δ2H in LUN (Fig. 5). The dissimilar major ion data also does not support this

2- (Fig. 4), as samples from LUN do not exhibit elevated SO4 , another chemical tracer of

groundwater discharging at SUS. Lake metabolism and primary productivity also play a

role in changing the isotope signature of the DIC pool (Bade et al., 2004; Myrbo and

Shapley, 2006). The saturation indices with respect to calcite (SIcalcite) in LUN show

saturation (0< SIcalcite <1) during all months sampled. A large change in lake productivity

would consume CO2 and drive lake carbonate equilibrium towards lower saturation with

respect to calcite (Profft and Stutter, 1993; Wetzel, 2001). The SIcalcite values of LUN in

May and September, however, do not display a shift towards lower values (Tab. 2). The

seasonal shift of LUN isotopic values, then, is likely related to a change in relative

contributions of watershed, or surface water, DIC to LUN, instead of a major change in primary productivity.

In 2015, the months sampled (May and September) fall into a period of above average temperature and below-average precipitation for this region (Supplementary Tabs. 1 and

2). The previous summer (June and July 2014) and prior months experienced above-

average precipitation. Temperature during the summer of 2015 (May-September) was also

above average every month. This suggests that the most noticeable difference between the

two years for the lakes was the water supply and degree of potential evaporation and

18 2 evapotranspiration. The δ O and δ H values support this hypothesis. The May and

September 2015 LUN samples plot below the cIMWL (slope = 7.15) on a line with

markedly lower slope (3.9). This value falls within the range (3.5 to 6) of surface water

evaporation lines (Gourcy and Brenot, 2011; Gat, 1981). The June and July 2014 samples

18 2 plot close to or above the line (Fig. 5). The similarity in δ O and δ H values between LUN

and VIC during the high-precipitation year suggests that both receive significant

contributions from local recharge within the plain. RIP did not display the same

evaporation signal, with all samples clustered around the meteoric water line.

The carbon isotope values respond to seasonal changes through several mechanisms. With

apparently less contribution from the alluvial groundwater and more time for the lake water

13 to exchange with atmospheric CO2 (atmospheric δ CDIC values = -7‰, Clark and Fritz,

1997), these low-precipitation samples (May and September 2015) are more enriched and

approach -7‰. Li et al. (2011) found a similar response in lakes in southwestern China,

although this region in Asia experiences opposite timing of wet and dry seasons than in

central Italy. The July 2014 sample could also be anomalously depleted for warm-season

13 δ CDIC in LUN because above-average precipitation brought increased runoff and a greater

13 flux of soil DIC (δ CDIC values ~23%, Clark and Fritz, 1997) to the lake.

18 34 The greatest seasonal changes in δ Osulfate and δ Ssulfate were observed in the samples LUN and VRV as well as the spring located within the plain, VIC. All other samples showed variability from season to season of less than 1-2 ‰. The concentration of sulfate in all the samples, however, did not change substantially among seasons (Tab. 2). This could indicate that the influence of anaerobic bacteria seasonally metabolizing sulfate is more important in some samples. Otherwise, a change in flux water containing soil particles where these microbes proliferate causes the isotope values to change during wetter periods

(Jedrysek, 2005). The influence of sulfate in fertilizers applied to crops in this area should not be disregarded (Vitòria et al., 2004), nor should the potential influence of atmospheric deposition of sulfate (Krouse and Mayer, 2000). The strong correlation in LUN between nitrate and sulfate concentrations (Fig. 3a) suggests that sulfate inputs may be related to fertilizer application or runoff of soil into the lake.

Sensitivity to meteorological parameters

The effect of different weather conditions, and potentially past and future climate, on lake

physio-chemical parameters was investigated using correlation analysis (Figs. 2 and 3) and

also by identifying seasonal shifts in some stable isotope values (Figs. 5 and 6).

The high coefficient values for monthly mean air temperature and water temperature in

LUN (0.92) versus lower in RIP (0.64) can be related to the hypothesized longer residence

time of water in LUN. If this is the case, then other parameters, such as total dissolved

solids (TDS) or accumulation of contaminants, will also be higher in LUN compared to

RIP and may also be more sensitive to outside forcing. The negative correlation

coefficients between precipitation and air temperature in both lakes is likely a consequence

of the Mediterranean climate in this region of Italy, where the majority of rain and snow

fall during the cold months. Alkalinity is greater in both lakes during colder periods, likely

- 2- because of increased delivery of dissolved ions, particularly HCO3 and CO3 , during cold and associated wet months.

The major ion correlations with meteorological parameters (Figs. 2 and 3) also offer insight to lake hydrology during different periods of the year. Nitrate showed strong negative correlation coefficients with temperature in both lakes, suggesting that either wet-season

(winter) runoff from surrounding agriculture delivers most of the nitrate to the lakes, or that algal assimilation of nitrate is most intense in the warm months, leaving lower detectable concentrations during warm periods (Wetzel, 2001). These potential explanations, if further investigated, can be utilized in the management and remediation of

the eutrophication of these lakes. Although marshy vegetation surrounds much of both

lakes, intercepting the irrigation canal-water before it flows in, this system may be less

2- effective at preventing nitrate from entering during the rainy months. In LUN, SO4 and

Ca2+ also show significant negative correlation coefficients with respect to temperature.

These ions may also be delivered to the lake by a greater flux of water during the wet months.

While there is not sufficient data available to quantify the water budgets for these two lakes, the correlation data and water stable isotope data suggest a substantial difference in one or more of the terms of the budget between them. A simplified water budget in lakes is given by the equation (modified from Rosenberry et al., 2015):

/ = ± ±

(eq. 1) ∆𝑉𝑉 ∆𝑡𝑡 𝑃𝑃 − 𝐸𝐸 𝑆𝑆𝑆𝑆 𝐺𝐺𝐺𝐺 where ΔV/Δt is the change in storage over time, or renewal time, P is precipitation, E is evaporation, SF is surface flow (either into or out of the lake) and GF is groundwater flow

(also either in or out of the lake). The mean annual evaporation can be estimated using the equation derived by Dragoni and Valigi (1994) for lakes located in central Italy:

. . = 19.007 3 063 0 486 𝑚𝑚 𝑚𝑚 𝑚𝑚 (eq. 2) 𝐸𝐸 𝑖𝑖 𝑇𝑇 where Em is the monthly evaporation (in mm), im is the Thornthwaite insolation monthly

◦ index (42 N latitude) and Tm is the monthly average air temperature. The annual mean

value of 1059 mm yr-1 was calculated using monthly temperature over the period of 2003-

2015. Mean annual precipitation, calculated by adding monthly totals over each 12-month period, over the same time period is 1010 mm yr-1. As warmer and drier conditions prevail, a concurrent increase in E and decrease in P may lead to a further deficit in P compared to

E (Fioravanti et al., 2015; Brunetti et al., 2012). The groundwater flow in and out was also assumed equal and opposite solely for the purpose of this estimation. Removing these terms, the ‘theoretical renewal time’, or the volume of the lake divided by the volume of its outflow (Ambrosetti et al., 2002), was calculated. The volumetric flow rate through the artificial channel from SUS to RIP, measured during May 2015 and September 2015 at 1 m3 s-1, is the only active surface flow observed in the two-lake system.

Although data for the rate of pumping water out of RIP was not obtained, a daily record of lake surface level from March- December 2014 shows that the level was maintained within

0.2 m, so the volume of outflow was estimated as the volume of this inflow channel water.

In this scenario, the rate of water cycling through RIP would cause average renewal time of 11.7 days. This value is low compared with other studies on lake renewal time but it is reflective of the more than two orders of magnitude difference in volume between RIP and these lakes (Ambrosetti et al. 2002; Varekamp, 2003). Without measurable or quantifiable water flowing in or out of LUN, the same estimation was not possible for this lake but a conceptual comparison is now feasible. It is unlikely, based on physical surveys, that significant through-flow for LUN at the surface or even as an unmeasured groundwater

component, persists throughout the entire year with near constant discharge as in RIP. This

difference in theoretical renewal time yields an explanation for the dissimilarities in

chemical parameters observed. This calculated rapid renewal rate in RIP also provides an

explanation for the lack of sensitivity of major ions in RIP to air temperature or

2 18 13 precipitation (Fig. 3) and lack of seasonal variation in δ H, δ O, and δ CDIC values (Figs.

5 and 6). In these parameters, the lake chemistry resembles the spring contributing inflow

(SUS), instead of reflecting within-lake processes, such as primary productivity or

evaporation. The difference in theoretical renewal time of the lakes also has potentially significant implications for how contaminants, agricultural runoff, and nutrients persist and cycle.

CONCLUSIONS

This hydrochemical investigation provides information on the provenance of source water for Lake Lungo and Lake Ripasottile and the response of these lakes to seasonal variations in temperature and water supply. Using chemical, isotopic, physical and meteorological data we could show that there is limited connectivity between these two lakes and that they

respond independently to local weather conditions. Despite the human landscape

modifications of LUN and RIP and their geographical proximity, it was found that nearly

all lake chemical parameters were more similar to their primary prospective groundwater

influences than to each other. These findings are evidence for hydrochemical investigations

as viable approaches in limnological studies when the availability of physical

measurements of flow is limited.

The recharge areas and elevations of the springs sampled (SUS and VIC) are separate and distinct, evidenced by the difference in the δ2H and δ18O of the waters. These observations

supported the characterization of these springs as representative of the distinctive source

waters to the lakes. RIP, similar to SUS in almost all chemical and isotopic parameters, is

sourced primarily from the carbonate aquifer and higher elevation recharge, whereas LUN,

along with VIC, is sourced mainly from the alluvial aquifer and local recharge. The

conveyance of carbonate aquifer water to RIP, represented by SUS water, is mainly

transferred via flow in an artificial channel that discharges into the lake. This creates a

faster flow-through regime than in LUN. Seasonal changes in LUN stable isotopes of water

also indicate a longer residence time of water in this system compared to RIP. The variable

seasonal contribution of groundwater and local recharge to LUN is evident in its

13 hydrochemistry, especially demonstrated by LUN δ CDIC values.

In the context of Italy’s past and future climate, the physio-chemical response of these lakes to changing seasons demonstrates how they may respond to shifting climate trends. If water supply to the lakes decreases in the future because of continuing patterns in lower wet- season precipitation and higher temperatures, there may be significant effects on lake water chemistry and trophic status. To determine the specific effects at RIP, a more frequent sampling campaign may be employed that reflects the fast water renewal rate calculated in

this work. LUN showed more sensitivity to changes in seasonal weather parameters, particularly illustrated by the correlation between nitrate concentration and the local air temperature. This should be considered for management of the protected area, especially within the context of increasing agricultural demands on groundwater in the basin and how this demand may grow with a changing climate.

ACKNOWLEDGMENTS

We would like to thank the Riserva Naturale dei Laghi Lungo e Ripasottile and their employees, especially Paolo Bellezza and Maurizio Sterpi, who helped significantly with sample collection and by providing housing at the field site. The Agenzia Regionale per la

Protezione Ambientale della Regione Lazio was also fundamental in this work through their data contribution. Professors Maurizio and Mario Barbieri and Sapienza University of Rome generously contributed with lab space and time for a portion of the sample processing. A portion of funding for this work was provided by the National Science

Foundation grant (GSS-1228126) awarded to Mensing and Noble, and also by the Paul

Yaniga Memorial Trust.

REFERENCES Ambrosetti W, Barbanti L, Sala N, 2002. Residence time and physical processes in lakes. J. Limnol. 62(Suppl. 1):1-15. Bade D, Carpenter S, Cole J, 2004. Controls of δ13C‐DIC in lakes: Geochemistry, lake metabolism, and morphometry. Limnol. Oceanogr. 49:1160-1172.

Bluszcz P, Lucke A, Ohlendorf C, Zolitschka B, 2009. Seasonal dynamics of stable isotopes and element ratios in authigenic calcites during their precipitation and dissolution, Sacrower See (northeastern Germany). J. Limnol. 68:257-273 Boni C, Bono P, Capelli G, 1986. [Schema idrogeologico dell’Italia central].[Article in Italian]. Mem. Soc. Geol. Ital. 35:991-1012. Brunetti M, Caloiero T, Coscarelli R, Gullà G, Nanni T, Simolo C, 2012. Precipitation variability and change in the Calabria region (Italy) from a high resolution daily dataset. Int. J. Climatol. 32:57-73. Capelli G, Mastrorillo L, Mazza R, Petitta M, Baldoni T, Banzato F, Doredana C, Di Salvo C, La Vigna F, Taviani S, Teoli P, 2012. [Carta Idrogeologica del Territorio della Regione Lazio].[Map in Italian]. Regione Lazio, Rome. Carmody RW, Plummer LN, Busenberg E, Coplen TB, 1998. Methods for collection of dissolved sulfate and sulfide and analysis of their sulfur isotopic composition. U.S. Geological Survey, Open-File Report 97-234: 91 pp. Carucci V, Petitta M, Aravena R, 2012. Interaction between shallow and deep aquifers in the Tivoli Plain (Central Italy) enhanced by groundwater extraction: A multi- isotope approach and geochemical modeling. Appl. Geochem. 27:266-280. Cavinato GP, De Cellis PG, 1999. Extensional basins in the tectonically bimodal central Apennines fold‐thrust belt, Italy: response to corner flow above a subduction slab in retrograde motion. Geology 27:955-958. Civita M, Fiorucci A, 2010. The recharge-discharge process of the Peschiera spring system (central Italy). Aqua Mundi 02019:61-178 Clark I, Fritz P, 1997. Environmental isotopes in hydrogeology. CRC Press/Lewis Publishers, Boca Raton: 328 pp. Cohen A, 2003. Paleolimnology: the history and evolution of lake systems. Oxford University Press, USA. Combourieu-Nebout N, Bertini A, Russo-Ermolli E, Peyron O, Klotz S, Montade V, Fauquette S, Allen J, Fusco F, Goring S, Huntley B, Joannin S, Lebreton V, Magri D, Martinetto E, Orain R, Sadori L, 2015. Climate changes in the central Mediterranean and Italian vegetation dynamics since the Pliocene. Rev. Palaeobot. Palynol. 218:127-147. Costantini E, Dazzi C, Costantini E, Fantappié M, L’Abate G. 2013. Climate and pedoclimate of Italy, p. 19-37. In: E Costantini and C Dazzi (eds.) The soils of Italy. Springer, Dordrecht. Cosentino D, Cipollari P, Marsili P, Scrocca D, 2010. Geology of the central Apennines: a regional review. J. Virtual Expl. 36:12. Di Matteo L, Dragoni W, 2006. Climate change and water resources in limestone and mountain areas: the case of Firenzuola Lake (Umbria, Italy), p. 83-33. In: N.

Goldscheider, J. Mudry, L. Savoy and F. Zwahlen (eds.), Proceedings of the 8th Conf.on Limestone Hydrogeology. Dragoni W, Valigi D, 1994. [Contribution to the knowledge of evaporation in lakes of Central Italy].[Article in Italian]. Geologica Romana 30:151-158. Epstein S, Mayeda T, 1953. Variation of O 18 content of waters from natural sources. Geochim. Cosmochim. 4:213-224. Fioravanti G, Piervitali E, Desiato F, 2015. Recent changes of temperature extremes over Italy: an index-based analysis. Theor. Appl. Climatol. 123:473-486. Fiorillo F, Petitta M, Preziosi E, Rusi S, Esposito L, Tallini M, 2015. Long-term trend and fluctuations of karst spring discharge in a Mediterranean area (central-southern Italy). Environ. Earth Sci. 74:153-172. Franceschini S, Damiani G, Venanzi D, Fabiani S, Schirone B, Mancini L, 2004. Environmental quality of the protected area of Lungo and Ripasottile lakes (Rieti, Italy). Rapporti ISTISAN 04/39. Giesemann A, Jäger HJ, Norman AL, Krouse HR, Brand WA, 1994. Online sulfur-isotope determination using an elemental analyzer coupled to a mass spectrometer. Anal. Chem. 66:2816-2819. Gourcy L, Brenot A. 2011. Multiple environmental tracers for a better understanding of water flux in a wetland area (La Bassée, France). Appl. Geochem. 26:2147-2158. Grassineau NV, Mattey DP, Lowry D. 2001. Sulfur isotope analysis of sulfide and sulfate minerals by continuous flow-isotope ratio mass spectrometry. Anal. Chem. 73:220- 225 Harris D, Porter LK, Paul EA, 1997. Continuous Flow Isotope Ratio Mass Spectrometry of Carbon Dioxide Trapped as Strontium Carbonate. Commun. Soil Sci. Plan. 28:747-757. Jedrysek M. 2005. S–O–C isotopic picture of sulphate-methane-carbonate system in freshwater lakes from Poland. A review. Environ. Chem. Lett. 3:100-112. Kornexl BE, Gehre M, Höffling R, Werner RA, 1999. On-line d18O measurement of organic and inorganic substances. Rapid Commun. Mass. Sp. 13:1685-1693. Krouse HR, Mayer B, 2000. Sulphur and oxygen isotopes in sulphate, p. 195-231. In: P.G. Cook and A.L. Herczeg (eds.), Environmental tracers in subsurface hydrology. Springer, New York. LaBaugh J, Rosenberry D, 2008. Introduction and Characteristics of flow, p. 1-38. In: D.O. Rosenberry and J.W. LaBaugh (eds.), Field techniques for estimating water fluxes between surface water and ground water: U.S. Geological Survey Techniques and Methods 4-D2. Leone A, 2004. [Ambiente e territorio agroforestale: line guida per la pianificazione sostenibile e gli studi di impatto ambientale].[Book in Italian]. Franco Angeli, Milan.

Li SL, Liu CQ, Tao FX, Lang YC, Han GL, 2005. Carbon biogeochemistry of ground water, Guiyang, Southwest China. Groundwater 43:494-499. Longinelli A, Selmo E, 2003. Isotopic composition of precipitation in Italy: a first overall map. J. Hydrol. 270:75-88. Massarutto A, 1999. Agriculture, water resources and water policies in Italy. Working paper 33-99. Fondazione ENI E. Mattei, Milan. Martarelli L, Petitta M, Scalise A, Silvi A, 2008. [Cartografia idrogeologica sperimentale della Piana Reatina (Lazio)].[Article in Italian]. Mem. Descr. Carta Geol. It. 81:137-156. Martin T, Creed J, Long L, 1994. Method 200.2: sample preparation procedure for spectrochemical determination of total recoverable elements. In: C.K. Smoley (ed.) Methods for the Determination of Metals in Environmental Samples. Environmental Monitoring Systems Laboratory, U.S. Environmental Protection Agency, Cincinnati, OH, USA. Maxfield R, Mindak B, 1985. EPA Method Study 27, Method 200.7: Trace Metals by ICP. US Environmental Protection Agency, Office of Research and Development, Environmental Monitoring and Support Laboratory. Mayer B, 2005. Assessing sources and transformations of sulphate and nitrate in the hydrosphere using isotope techniques, p. 67-89. In: P.K. Aggarwal, J.R. Gat and F.O. Froehlich (eds.), Isotopes in the water cycle: past present future of a developing science. Springer, Dordrecht. Mensing SA, Tunno I, Sagnotti L, Florindo F, Noble P, Archer C, Zimmerman S, Pavón- Carrasco FJ, Cifani G, Passigli S, Piovesan G, 2015. 2700 years of Mediterranean environmental change in central Italy: a synthesis of sedimentary and cultural records to interpret past impacts of climate on society. Quaternary Sci. Rev. 116:72- 94. Morrison J, Brockwell T, Merren T, Fourel F, Phillips AM, 2001. On-line high-precision stable hydrogen isotopic analyses on nanoliter water samples. Anal. Chem. 73:3570-3575. Myrbo A, Shapley M, 2006. Seasonal water-column dynamics of dissolved inorganic 13 carbon stable isotopic compositions (δ CDIC) in small hardwater lakes in Minnesota and Montana. Geochim. Cosmochim. Ac. 70:2699-2714. Parkhurst D, Appelo C, 1999. User's guide to PHREEQC (version 2)- A computer program for speciation, batch-reaction, one-dimensional transport, and inverse geochemical calculations: U.S. Geological Survey Water-Resources Investigations Report 99- 4259: 312 pp. Perini M, Camin F, Corradini F, Obertegger U, 2009. Use of δ18O in the interpretation of hydrological dynamics in lakes. J. Limnol. 68:174-182.

Petitta M, 2009. Hydrogeology of the middle valley of the Velino River and of the S. Vittorino Plain (Rieti, Central Italy). Ital. J. Eng. Geol. Environ. 1-157:181. Petitta M, Primavera P, Tuccimei P, Aravena R, 2011. Interaction between deep and shallow groundwater systems in areas affected by Quaternary tectonics (Central Italy): a geochemical and isotope approach. Environ. Earth Sci. 63:11-30. Pfaff J, 1993. Method 300.0 Determination of inorganic anions by ion chromatography. US Environmental Protection Agency, Office of Research and Development, Environmental Monitoring Systems Laboratory. Priscu J, Dodds W, 1988. An inexpensive device for sampling large volumes of lake water from discrete depths. Freshwater Biol. 20:113-115. Riccardi R, 2006. [Studi geografici sui laghi Lungo, Ripasottile e Ventina]. Quaderni della Riserva Naturale dei laghi Lungo e Ripasottile 1-47. Rosenberry D, Lewandowski J, 2015. Groundwater‐the disregarded component in lake water and nutrient budgets. Part 1: effects of groundwater on hydrology. Hydrol. Process. 29:2895-2921. Sacks L, Lee T, Swancar A, 2014. The suitability of a simplified isotope-balance approach to quantify transient groundwater-lake interactions over a decade with climatic extremes. J. Hydrol. 519:3042-3053. Sappa G, Barbieri, M, Ergul S, Ferranti F, 2012. Hydrogeological conceptual model of groundwater from carbonate aquifers using environmental isotopes (δ18O, δ2H) and chemical tracers: a case study in southern Latium Region, Central Italy. J. Water. Resour. Prot. 4:695-716. Spadoni M, Brilli M, Giustini F, Petitta M, 2010. Using GIS for modelling the impact of current climate trend on the recharge area of the S. Susanna spring (central Apennines, Italy). Hydrol. Process. 24:50-64. Tallini M, Falcone R, Carucci V, Falgiani A, Parisse B, Petitta M, 2014. Isotope hydrology and geochemical modeling: new insights into the recharge processes and water- rock interactions of a fissured carbonate aquifer (Gran Sasso, central Italy). Environ. Earth Sci. 72:4957-497. Tassi F, Cabassi J, Rouwet D, Paslozzi R, Marcelli M, Quartararo M, Capecchiacci F, Nocentini M, Vaselli O, 2012. Water and dissolved gas geochemistry of the monomictic Paterno sinkhole (Central Italy). J. Limnol. 71:245-260. Usdowski E, Hoefs J, Menschel G, 1979. Relationship between 13C and 18O fractionation and changes in major element composition in a recent calcite-depositing spring - A model of chemical variations with inorganic CaCO3 precipitation. Earth Planet. Sci. Lett. 42:267-276. Varecamp J, 2003. Lake contamination models for evolution towards steady state. J. Limnol. 62(Suppl.1):67-72.

Vitòria L, Otero N, Soler A, Canals A. 2004. Fertilizer characterization: isotopic data (N, S, O, C, and Sr). Environ. Sci. Technol. 38:3254-3262. Wetzel R, 2001. Limnology: lake and river ecosystems. 3. Academic Press, San Diego: 1006 pp. Zuppi GM, Bortolami GC, 1982. Hydrogeology: a privileged field for environmental stable isotopes applications. Some Italian examples. Rend. Soc. Ital. Mineral. Petrol. 38:1197-1212.

Tab. 1. Geographic and limnologic characteristics of the study sites. Sample Coordinates Elevation Surface Lake Mean code (m asl) area or volume depth discharge (m3) (m) Surface Lago Lungo LUN 42.4723° N 370 0.41 km2 870,000 3 waters 12.8484° E Lago di RIP 42.4741° N 367 0.63 km2 1,000,000 3 Ripasottile 12.8170° E Velino VRV 42.4799° N 374 35 m3 s-1 0.5 River 12.8085° E Groundwater Vicenna VIC 42.4679° N 375 0.07 m3 s-1 springs Riara 12.8602° E Santa SUS 42.5020° N 388 5.0 m3 s-1 Susanna 12.8519° E Peschiera PES 42.3655° N 413 18 m3 s-1 13.0043° E Terme di COT 42.3764° N, 414 0.25 m3 s-1 Cotilia 12.9981° W

Tab. 2. Physical and chemical parameters of all waters sampled, including major ion concentrations in mg L-1. Results of the mixing simulation carried out using PHREEQC are also included. - 2- 2 2+ + 2 Temperature Conductivity pH Cl SO4 HCO3- Na Mg K Ca SIcalcit SIanhyd SIgyps + + (°C) (µS/cm) (from e alkalinity) SUS 2015-September 10.4 855 7.3 3.6 270 275 2.9 30 0.55 140 0.18 -1.5 -1.0 2015-May 10.3 798 7.1 3 260 235 2.8 29 0.68 130 -0.1 -1.5 -1.0 2015-February 10.7 851 7.6 3.7 280 296 2.6 30 0.56 120 0.53 -1.5 -1.0 2014-July 10.9 835 7.4 3.7 270 266 2 27 0 110 0.13 -1.5 -1.0 VIC 2015-September 12.2 520 7.4 6.7 7 304 5.3 2 1.2 95 0.29 -3.0 -2.6 2015-May 13.2 506 7 6.6 6.8 299 5.2 2.1 1.7 94 0 -3.0 -2.6 2015-February 13 527 7.4 4.8 10 357 3 2.1 0.78 100 0.38 -2.9 -2.4 2014-July 13 534 7.3 5.5 6.3 292 4.2 2 1.2 96 0.16 -3.0 -2.6 PES 2015-September 12.4 597 7.2 4 6.8 411 4.4 18 1.1 100 0.22 -3.1 -2.6 2015-May 11.4 594.8 7 3.4 8.9 409 3.1 19 0.84 100 -0.44 -3.0 -2.5 2015-February 10 653 7 7 6.8 540 4.9 20 1.2 100 0.09 -3.1 -2.6 2014-July 12.8 617 7.2 4.1 8.4 404 2.6 19 0.8 94 0.12 -3.0 -2.6 COT 2015-September 14.6 2024 5.6 22 130 1534 18 56 3.7 340 -0.44 -1.6 -1.2 2015-May 14.2 1925 5.7 17 130 1183 17 55 3.8 340 0.37 -1.6 -1.1 2015-February 12.4 2040 6 23 150 1721 16 58 3.3 350 -0.03 -1.5 -1.1 RIP 2015-September 16.2 792 8 5.5 220 201 3.5 28.5 0.86 130 0.8 -1.5 -1.1 2015-May 18 837 7.5 3.5 210 232 3.2 26 0.87 120 0.08 -1.5 -1.2

2015-February 10.1 755 8.4 5.4 190 333 3.6 25 0.86 120 1.28 -1.6 -1.2 2014-July 17.8 760 8.2 5.5 210 239 2.7 23 0 100 0.96 -1.6 -1.2 LUN 2015-September 19.6 517 7.9 7.4 11.4 240 7.6 7.1 1.8 105 0.83 -2.7 -2.4 2015-May 22.8 419.5 7.6 8.7 9.3 220 8.6 5 1 58 0.34 -3.0 -2.6 2015-February 10.3 666 7.7 8.7 21 384 7.1 6.4 1.5 110 0.69 -2.6 -2.8 2014-July 20.1 587 8 6.8 9.9 307 5.6 4.9 1 84 0.94 -2.8 -2.5 VRV 2015-September 12.8 789 7 9.7 32 546 5.4 24 1.2 108 0.15 -2.4 -2.0 2015-May 12.4 798 6.9 5.8 27 534 5.4 20 1.3 20 -0.07 -2.4 -2.0 2015-February 9.2 613 8.7 9.2 20 549 6.2 14 1.3 91 1.66 -2.7 -2.2 2014-July 12.7 769 7.5 9.4 27 470 6.2 20 1.2 110 0.5 -2.5 -2.0 MIX 80:20% 4.3 218 275 2.7 23 0.2 105 0.28 -1.6 -1.3 SUS:LUN (PHREEQC simulation)

Tab. 3. Stable isotope data. Values are expressed in per mil (‰) relative to the standard indicated in the subscript. 13 34 18 18 2 δ CPDB δ SVCDT δ O-sulfate δ OVSMOW δ HVSMOW SUS 2015-September -8.8 15.5 9.6 -8.4 -53 2015-May -8.9 14.9 10.0 -8.6 -54 2015-February -9.0 15.4 9.3 -8.4 -52 2014-June -9.0 -8.6 -53 2014- July -10.0 15.3 10.7 -8.6 -53 VIC

2015-September -13.5 15.2 6.0 -7.1 -45 2015-May -12.0 8.6 4.2 -7.4 -46 2015-February -13.5 8.8 5.2 -7.1 -43 2014-June -13.4 -6.9 -43 2014- July -13.8 -9.9 -0.7 -6.9 -43 PES 2015-September -4.8 12.1 9.0 -9.0 -57 2015-May -4.2 12.1 8.3 -9.0 -57 2015-February -3.9 12.2 9.0 -8.9 -55 2014- July -6.1 10.2 8.0 -8.9 -56 COT 2015-September 2.9 16.9 10.6 -8.9 -57 2015-May 4.7 17.2 10.2 -9.0 -58 2015-February 4.1 17.0 9.9 -9.0 -56 RIP 2015-September -7.7 15.8 9.3 -7.7 -49 2015-May -8.8 15.3 9.5 -8.2 -52 2015-February -9.2 15.1 9.4 -8.2 -50 2014-June -8.7 -7.7 -48 2014- July -10.7 15.3 10.8 -7.7 -48

LUN 2015-September -8.9 14.4 7.4 -5.3 -37 2015-May -9.5 12.2 7.7 -6.0 -41 2015-February -12.0 12.3 9.3 -7.0 -44 2014-June -11.2 -6.3 -40 2014- July -13.9 11.1 6.8 -6.3 -40

VRV 2015-September -3.1 14.5 10.0 -8.7 -56 2015-May -4.1 9.1 9.2 -8.7 -56 2015-February -6.6 9.2 8.5 -8.1 -50 2014- July -4.7 13.0 8.5 -8.7 -55 Rain Gauges - Spadoni et al. (2010) Elevation 378.0 -6.0 -36 1375.0 -8.2 -53

Fig. 1. Map of study area. Black dots indicate sampling localities. Elevations contours show 150 meter intervals. Numbers correspond to names given in Tab. 1: 1, Lake Lungo (LUN); 2, Lake Ripasottile (RIP); 3, Velino River (VRV); 4, Vicenna Riara Spring (VIC); 5, Santa Susanna Spring (SUS); 6, Peschiara Spring (PES); 7, Cotilia di Terme Spring (COT).

Fig. 2. Correlograms of lake physical parameters and local meteorological data in a. LUN and b. RIP, during each season over the years 2010-2015. Values of the correlation coefficient (r) are displayed, while shading denotes the level of significance (p). Black denotes the most significant linear correlation and white denotes not a significant correlation; the null hypothesis cannot be rejected.

Fig. 3. Correlograms of major ion concentrations in a. LUN and b. RIP for the only year where monthly data was available, 2011. Also included are mean monthly local air temperature and total monthly precipitation during 2011. Shading pattern is the same as in

Fig. 2. Major ion concentrations for this time-period were collected and analyzed by ARPA, Lazio. Meteorological data (Temp., Precip.), retrieved from http://www.idrografico.roma.it/, are the monthly mean air temperature or total - precipitation from the month when the major ion sample was taken. Phosphate (PO4 ) concentrations were not available for RIP, so are not included in (b.).

Fig. 4. Piper diagram of major ion concentration of waters in the Rieti Basin area. Values are the mean of the five seasons sampled. For individual season values, refer to Tab. 2.

Fig. 5. Oxygen and Hydrogen isotope composition of waters in the Rieti Basin and Velino River Valley. MMWL, cIMWL (Longinelli and Selmo, 2003), and GMWL, are the Mediterranean Meteoric Water Line, central Italian Meteoric Water Line, and the Global Meteoric Water Line, respectively, and from top to bottom. Rainwater samples are from locations within the Rieti Basin and surrounding mountains from Spadoni et al. (2010). Month is indicated for LUN samples that fall below the meteoric water line.

Fig. 6. Bicarbonate concentration versus carbon isotope composition of dissolved inorganic C in sampled waters. Major groupings, I, II, and III, are denoted by the dashed circles and described in the text. LUN sample months that are outside the groupings are labeled.

Fig. 7. Sulfur and oxygen isotope composition of sulfate in waters of the study area. Plotted values are the means of the four seasons sampled. Fields indicate typical ranges in values of sulfate from terrestrial sources, atmospheric sulfate, soil sulfate, and from marine evaporites (modified after Mayer, 2005; 67-81, fig. 1; with permission).

Fig. 8. Conceptual model of groundwater and surface water flow in the northern sector of the Rieti Plain. Drawing is not made to scale but as relative orientation. A) Map-view of the plain. B) Cross-section view. 1, Plain-filling fluvial and lacustrine deposits (Holocene); 2, lenses of peat and peaty clays (Late Pleistocene- Holocene); 3, conglomeratic alluvial deposits and braided gravels (Pliocene- Pleistocene); 4, carbonate bedrock (Mesozoic) and regional aquifer; 5, lakes (overhead and cross-sectional view); 6, Velino River; 7, manmade waterways or canals; 8, water table; 9, major spring; 10, Eastern boundary normal fault; 11, buried normal fault; 12, water flow in surface water courses - arrow size is relative

to flow; 13, groundwater flow in low-permeability units; 14, aquifer recharge - size of arrow corresponds to potential infiltration; 15, groundwater flow paths through major aquifers.

CHAPTER 2 Hydrogeochemical response of groundwater springs during central Italy earthquakes (24 August 2016 and 26-30 October 2016) Abstract

The seismic sequence of 2016-2017 in central Italy affected aquifer properties, causing

notable hydrochemical variations in groundwater springs. The sequence consisted of four

mainshocks and several hundred aftershocks over five months that ruptured normal fault

segments within carbonate domains of the central Apennines. Major karstic aquifers are

held within the faulted area that supply drinking water to major metropolitan areas. We

compare hydrochemical parameters before and after the earthquakes at springs associated

with three major carbonate aquifers, Nerea (NER), Santa Susanna (SUS), and Peschiera

(PES) springs, and one shallow alluvial aquifer, Vicenna Riara (VIC) spring. Physio-

chemical parameters were measured in spring water samples (electrical conductivity

(EC), alkalinity, pH, temperature), and samples were then analyzed for trace element

concentrations (Al, Cr, Mn, Fe, Co, Ni, Cu, Rb, Sr, U) and stable isotope composition

2 18 34 18 13 (δ H and δ O of water, δ S and δ O of sulfate, and δ C of dissolved inorganic carbon).

The EC and alkalinity was elevated relative to pre-earthquake values at PES and SUS

after the first mainshock (Aug. 24th) and another peak was observed after the next

mainshocks on Oct. 26th and 30th, though not as prominent. NER also exhibited peaks in

EC and alkalinity following the Oct. 26th and 30th earthquakes but not after the Jan. 18th

mainshock. Trace element concentrations of all springs generally showed the same post-

mainshock peaks as the physiochemical parameters, though SUS exhibited less response after the Oct. mainshocks than the August mainshock. Trace element concentrations

declined and most trace elements returned to pre-earthquake values in all four spring waters by the end of sampling in April 2017. The source of the solutes is proposed to be water stored in smaller fractures or abandoned karstic flowpaths longer mean residence time groundwater with more rock-water interaction. These fluids would then be expelled

into the main flow paths as a result of transient increases in pore pressure and hydraulic

conductivity from co-seismic aquifer stress. The pattern of post-mainshock peaks show a

progressive decline in trace element concentration, EC and alkalinity from August

through January main shocks. The weak response in January may be explained by

progressive depletion of high solute fluids as earlier shocks may have flushed out the

18 2 higher salinity reservoirs. The δ OH2O and δ HH2O showed no significant change from pre-

earthquake values, indicating no major change flowpaths or mixing from alternative

aquifer sources. The δ34S and δ18O of sulfate do not show any discernible signals that can

18 be tied to the seismicity, except for a brief enrichment of PES δ Osulfate after the Aug.

13 shock, possibly related to aeration of groundwater. The δ CDIC at PES becomes enriched

outside the range of pre-earthquake values following the Aug 24th event through the Oct

30th event, while SUS enrichment occurred after the October shocks. This enrichment

may be a sign of input from deeply-sourced dissolved CO2 after dilation of specific fault

conduits. Differences in carbon isotopic responses in NER, VIC, PES and SUS are

attributed to aquifer heterogeneity including different degrees of faulting and proximity

to mainshock epicenters. Correlation between changes in trace element concentration and

precipitation data was poor, ruling out seasonal effects on observed variations. Future

work involves the installation of temperature, EC, and water level sensing equipment in

wells and continued monitoring of spring physio-chemical parameters.

INTRODUCTION

Co-seismic hydrological and chemical response at groundwater springs following

strong earthquakes is a significant concern worldwide, especially where groundwater is an important resource (Pasvanoglu et al., 2004; Skelton et al. 2008; Reddy et al. 2011;

Skelton et al., 2014). The Apennines, a mountain belt in central Italy characterized by regional carbonate groundwater systems interacting with active normal faults capable of

producing Mw 6.5 to 7.0 seismic events, are particularly suceptible (Cello et al. 1997,

Valensise et al. 2003). These aquifers also provide water supply to major metropolitan

areas in the region. On August 24, 2016, the first main shock (Mw= 6.0) of the 2016-17 central Italy Seismic Sequence struck the central Apennines in the area where Latium joins Umbria, Marche and Abruzzi and was followed one hour later by a Mw 5.4 shock.

The epicenter of the event was located at the segment boundary between the Mt. Vettore and Mt. Laga faults. On October 26, 2016 and on October 30, 2016, three other big shocks (Mw 5.5, Mw 5.9 and Mw 6.5, respectively) ruptured again the Vettore Fault and its NW extension. On January 18th 2017, four shocks with magnitude greater than 5.0 struck a different fault segment to the south of the previous events, with a mainshock of

Mw 5.7. The seismic sequence totaled over 26,000 earthquakes, including aftershocks, activating two major fault segments and spanning over 40 km (Chiaraluce et al. 2017).

The October 30th shock was the largest event to hit the central Apennines in 26 years (Chiaraluce et al. 2017). Sampling began two days afterwards the first mainshock on

August 24th 2016. Sampling localities included springs discharging aquifers in the Rieti area, including the Peschiera spring, which feeds the aqueduct of Rome, and the Santa

Susanna spring, which discharges from a similar large aquifer. These springs, sampled previously in 2014 and 2015, provide comparison for pre-earthquake physiochemical and stable isotopic parameters.

The goal of this study is to evaluate the strong earthquake sequence effects on the hydrochemistry of groundwater from various regional aquifers through the analysis of pre-

and post-seismic water samples. The aquifers are subjected to the direct effects of shaking and tectonic displacement of large hydrogeologic structures, which in turn affects the flow paths and flow dynamics on different time scales.The comparison between the responses of springs situated different distances from a sequence of large-magnitude events, as well as the progressive responses of these springs throughout the seismic sequence, is especially significant for understanding the resilience of groundwater systems in an active tectonic zone.

Tectonic Setting

The ruptured faults are located within a relay zone between two major overlapping NNW–SSE trending normal faults in the Central Apennines. The Apennines fold-thrust belt is part of the accretionary wedge caused by the roll back of the Adriatic subduction towards the east (Cavinato and De Celles 1999). The Quaternary–Neogene normal faults, formed by the subsequent west-to-east migration of the regional extensional regime, govern the intra-montane basin evolution and its filling through continental clastic deposits and also are capable of producing large devastating earthquakes (M 6.0-7.3) (Roberts and Michetti 2004). The ruptured fault segments are focused on either side of the Olevano–Antrodoco–Sibillini (OAS) thrust system and occurred in between two similar seismic sequences that occurred in the central Apennine

in the past 20 years: the 1997 Colfiorito sequence to the north, and the 2009 L’Aquila

events to the south (Chiaraluce et al. 2017).

BACKGROUND: EARTHQUAKE HYDROLOGY

The pre-, co- and post-seismic changes observed in groundwater can be attributed

to several factors originating from strain and/or rupture along faults or fractures changing

the properties of the aquifer containing that groundwater (Reddy et al. 2011, Woith et al.

2013). The strain or stress provided by an earthquake can be divided into static and

dynamic components: the former due to the movement or fault offset and the latter due to

the strain from seismic waves on the aquifer (Manga and Wang 2015). Groundwater response at varying distances from the epicenter has been widely documented (Claesson et al. 2004; Hartmann and Levy 2006; Falcone et al. 2012, La Vigna et al. 2012; Skelton et al. 2014) as changes in groundwater level, spring discharge or groundwater physical

and chemical parameters. The distance from epicenter to groundwater, termed near-field

(within the distance of the ruptured fault length), intermediate field (1-10 ruptured fault lengths) and far-field (greater than 10 ruptured fault lengths) by previous studies (Wang and Manga 2010, La Vigna et al. 2012), is important as dynamic and static stresses caused by the earthquake decay differently with distance, and categorization may help predict earthquake hydrologic effects. The response, duration, and propagation distance of that response depends on the aquifer hydrogeology, pore pressure spread, and the

earthquake strength (Liu et al. 2010; Manga and Wang 2015). The duration of the

response, as short- mid- and long-term effects, are important to identify as permanent

aquifer deformation has implications on large aquifers that are major drinking water

sources (Malakootian and Nouri 2010). In fractured carbonate aquifers like those in the

Apennines, transient responses are common as new microfractures are formed, existing

fractures are cleared and new flow paths form but eventually return to their original state

(Amoruso et al. 2011, Galassi et al. 2014).

This study reports on the groundwater chemistry response to the seismic sequence

of August 2016 - January 2017 from springs located within the intermediate field, 30-40 km away from all of the epicentral areas, and in one instance a spring within 5 km of the

epicentral area (near-field) for the October shocks. The collection of pre-earthquake data

from the springs included in this study allows for evaluation of post earthquake effects

and a rare comparison between near and intermediate-field responses to major (M>6)

earthquake events. Pre-, co- and post-seismic chemical data was generated in this study

from Nerea spring, the near-field spring for the October shocks, from daily water samples

collected and stored at the Nerea S.p.A. bottling plant. A pre- earthquake study of the

hydrochemistry in the Rieti area included characterization of three of the springs in this

study during four seasons the previous two years (Archer et al. 2016). Stable isotope data

(δ18O and δ2H of water, δ13C of dissolved inorganic carbon, δ18O and δ34S of sulfate) collected from the Rieti area springs during the two years prior to the seismic sequence

provides a means to study how flowpath and groundwater-rock interaction (Petitta et al.

2011) may be affected by the major earthquakes.

Major aquifers in the central Apennines are hosted within fractured Meso-

Cenozoic carbonate massifs that overly an evaporitic basal structures (Quattrocchi, 1999).

Recharge occurs mainly in the high-elevations and aquifers are characterized by two major flowpaths: one shallower with high flow rate through major discontinuities and the other with low to medium flow rate where seepage or uprising of deep mineralized fluids occurs along fractures (Amoruso et al. 2010, Petitta et al. 2011). This dual- or multiple flow- structure allows for differences in residence times of water within the same aquifer.

The major faults and tectonic structure of these aquifers can serve two purposes: allowing the upwelling of deep mineralized fluids and mixing of groundwater with different chemistries (Petitta et al. 2011), or compartmentalizing the aquifer, acting as lower permeability aquicludes (Amoruso et al. 2011). The mineralized waters containing dissolved CO2 sourced from the mantle are found in a reservoir within the Mesozoic

basement rock unit that is originally sourced from deeper within the crust or mantle

(Chiodini et al. 2004, Petitta et al. 2011). The low-permeability portion of these aquifers

contains water with higher concentrations of dissolved ions, including trace elements

(Morgantini et al. 2009).

HYDROGEOLOGICAL SETTING

Nerea Area

The hydrogeological unit comprising the Nerea spring recharge area (Figure 1) is a major carbonate aquifer within the Umbria-Marche carbonate ridge that is characteristic of the Central-Apennine (Mastorillo and Petitta 2014). In these karstic large-area

aquifers, faults and fractures within the limestone control flow (Carro et al. 2005). The

aqueduct feeding the Nerea S.p.A bottling plant was sampled during this study as part of

the bottling plant commercial license. The aqueduct flows from the Uccelletto spring,

with a discharge measured at 0.57 m3/s sourcing the Nera-Ussita basal aquifer (Tarragoni,

2006).

Rieti Area

Deep normal faults at the border of the Rieti plain separate the marine Meso-

Cenozoic carbonate ridges from Plio-Quaternary continental deposits. This structural

style is characteristic of central Apennine extensional-basins, where the regional flow

system is hosted by the surrounding carbonate bedrock and many springs discharge at the

contacts with lower-permeability basin deposits (Martarelli et al. 2008; Capelli et al.,

2012). At the northeastern edge of the Rieti Plain, Santa Susanna Spring (SUS) is the highest discharge spring (5.5 m3/s) of those in direct contact with the plain, is supplied by

base-flow of the regional aquifer (Zuppi and Bortolami, 1982), and emerges at the

intersection of two normal faults (Spadoni et al. 2010). This regional aquifer is hosted in

the Terminillo and Reatini mountain carbonates to the east with infiltration of 545 mm/yr

and total recharge area over 300 km2 (Spadoni et al. 2010). The aquifer host rock consists

of imbricately thrust carbonates with evaporitic gypsum and anhydrite units in the

subsurface (Martarelli et al., 2008; Spadoni et al., 2010, Capelli et al., 2012) The aquifer

geology is reflected in the SUS water chemistry, which has the highest EC and the

highest sulfate concentration of all the springs sampled in this study (Archer et al., 2016).

Locally, interbedded Triassic dolomitic limestone outcrops are observed (Martarelli et

al., 2008). The average residence time of water in this aquifer is on the order of 15-20

years, calculated using values of aquifer properties, recharge area and Euclidian distance

to the spring (Spadoni et al. 2010). Vicenna Riara Spring (VIC) is distinct among springs

included in this study because of its location near the center of the plain. Water flows to

this spring along a buried fault and through preferential flow-paths within the Pleistocene

conglomeratic alluvial deposits that lie in the central-eastern part of the plain, and this

spring is characterized by a relatively low discharge of 0.07 m3/s (Martarelli et al., 2008).

Peschiera spring (PES) is located about 8 km to the east of the Rieti Plain in the

Velino River Valley (Figure 1). PES has high discharge (i.e., 18 m3 s-1) and represents a

major drinking water supply to Rome. The groundwater contribution to the spring is

mainly from the aquifer hosted within the uplifted carbonate shelf deposits to the ESE in

bedrock consisting of Triassic – Paleocene limestone located in the Giano-Nuria mountains (Petitta, 2009; Civita and Fiorucci, 2010). The mean residence time of water is

~5 years in this aquifer, though two flow paths have been identified with significantly

different mean residence times (~5 versus 25-30 years) that contribute to water

discharging at the spring. Recharge area for PES is delineated in Figure 1, and the mean

infiltrating volume of water is calculated by Civita and Fiorucci (2010) to be 17.87 m3/s.

METHODS

FIELD METHODS

The physio-chemical parameters of pH, temperature and electrical conductivity

(EC) were evaluated on site using specific field probes: a HANNA instruments (USA) HI

9025 pH-meter equipped with sensors for pH and temperature and a HANNA

Instruments HI 9033 conductivity probe for electrical conductivity. Hardness and alkalinity were directly evaluated on site using MERCK (Germany) titration kits doing three replicates for every sample site. All field sampling and analysis procedures were made wearing nitrile gloves to avoid any kind of sampling contamination, and all sample bottles were rinsed three times with the environmental sample water prior to collection.

Sample storing, shipping

All LPDE bottles used to collect samples were pre-washed using NALGENE

L900 (USA) soap. Bottles for cation analysis were also washed with 2% HNO3, bottles

for isotope analysis were washed with NALGENE L900 and two times with 2% HNO3,

and bottles for trace elements were washed with NALGENE L900 and two times with 2%

HNO3. Sample bottles were rinsed three times with the sample to be collected before being filled. Samples for laboratory analysis were filtered on site with a 0.45 micrometers sterile Millex-GS millipore MCE membrane.

Inductively coupled plasma mass spectrometry (ICP-MS)

Samples for trace element analysis were acidified with 2% ultrapure HNO3,

previously obtained by sub-boiling distillation of 65% acid using Milestone (USA)

duoPUR. The samples were then analyzed using a Thermo scientific (USA) Icap Q

instrument. To certified the quality of analysis and observe possible instrumental bias, an

internal standard with 10 ppb of indium (In) was spiked in all the samples. The eecovery

of In spikes were all within ten percent. Elements chosen for the analysis were Al, Cr,

Mn, Co, Cu, Ni, Fe, Pb, U, Rb, and Sr. Errors estimated by replicate analysis were lower

than ± 0.01 ppb for Cr, Mn, Pb, U, Rb and Sr. Error for Al and Fe was ± 0.2 ppb, for Ni was ±0.03 ppb, and for Cu was ± 0.5 ppb. Other elements (Se, As, Cd) were analyzed to see possible release after the earthquake, but the concentrations results were lower than our blank values and showed no discernable change. A sample of the acid used to preserve samples in the laboratory was also analyzed. Although the acid blanks do contain low concentrations of the trace elements analyzed, they are below the

concentrations of the lowest environmental samples. Concentrations presented in this

paper have not been corrected to exclude this low-level contamination (see Table 3 for all results). The pre-earthquake ICP-MS analyses for PES, SUS and VIC were performed using archived waters collected during the previous study of Archer et al. (2016). These

waters were sampled using the same procedure and were preserved with nitric acid. All

ICP-MS analyses were performed at the Università degli Studi Dell’Insubria

Dipartimento di Scienza e Alta Tecnologia, Como Italy, as a collaborative effort with

UNR.

ISOTOPE METHODS

Waters were prepared for isotopic analysis by precipitation of dissolved sulfate as

BaSO4 following the method of Carmody et al. (1998). BaSO4 precipitates were analyzed

34 for δ S using V2O5 as a combustion aid, and followed the methods of Giesemann et al.

18 (1994) and Grassineau et al. (2001). BaSO4 precipitates were analyzed for δ O following the method of Kornexl et al. (1999). The analytical error (1-sigma), estimated by replicate analysis, was ±0.2‰ and ±0.4 ‰ for δ34S and δ18O of dissolved sulfate,

respectively. Water samples were prepared for isotopic analysis of dissolved inorganic

carbon by precipitation as SrCO3 after the method of Usdowski et al. (1979), and then

analyzed using the method of Harris et al. (1997), with analytical error within ±0.2‰.

18 Waters were analyzed for δ O using the CO2 - H2O equilibration method of Epstein and

Mayeda (1953), and for δ2H using the method of Morrison et al. (2001). The analytical

error of these measurements was ±0.1‰ and ±1.0 ‰ for δ18O and δ2H, respectively. All

stable isotope analyses were carried out at the University of Nevada, Reno Stable Isotope

lab. All values are reported using delta notation (δ‰), and the standards used were V-

SMOW for oxygen and hydrogen, V-PDB for carbon and V-CDT for sulfur.

METEOROLOGICAL AND SEISMOLOGICAL DATA COMPILATION

Daily precipitation and temperature measured at the Peschiera spring (PES), Mt.

Terminillo (SUS and VIC) and Leonessa (NER) was retrieved from http://www.idrografico.roma.it/ (2017). The monthly total precipitation amounts were calculated for the sampled months and are included in Tables 4a.-4d.

Dates, location, and magnitudes of earthquakes in the seismic sequence were mapped from the INGV (Istituto Nazionale di Geofisica e Vulcanologia) Centro

Nazionale Terremoti database, http://cnt.rm.ingv.it/en/ .

STATISTICAL ANALYSES

The means and standard deviation of the pre-earthquake data were calculated to demonstrate a range of natural variability. The potential correlations among trace element

concentrations were analyzed using the Pearson correlation coefficient and r2 values.

Compiled monthly precipitation data was also included in these tests to determine if

meteorological parameters had any control on concentrations of dissolved constituents.

The correlation matrices were generated using the total monthly precipitation of the

month prior to the samplings date. All sampling dates were then tested for correlation

with all trace elements and the prior months’ precipitation.

RESULTS

PHYSIO-CHEMICAL MEASUREMENTS

Figure 2a., b., c., and d. show the alkalinity, EC, pH and temperature as time

series plots for the four springs sampled. The pre-earthquake fields for SUS, VIC and

PES (shaded boxes in figures 2, 3 and 4) are the range of values from sampling conducted in 2014-2015 (Archer et al. 2016). The reference values supplied by the Nerea

bottling plant are indicated by the dashed horizontal line and are essentially pre-

earthquake averages for each parameter. Alkalinity (Figure 2a) is elevated at SUS, VIC

and PES after the Aug. 24th mainshock then decreases over the next two-week period.

SUS and VIC alkalinity returned to pre-earthquake values by Oct. 26th, becoming

elevated again in January 2017. PES values remain elevated until returning to pre-

earthquake range by May 17th 2017. NER alkalinity increased sharply after October 30,

peaking on 8 November.

EC values (Figure 2b) follow the same general pattern, where all four springs

were elevated at the onset of sampling after Aug. 24th, into October. After the Oct. 30th

mainshock, SUS and PES increase again above the range of pre-earthquake values. NER exhibits an abrupt increase following the Oct. 26th and/or Oct. 30th earthquakes and

remains elevated until a decrease and return to below reference values by 18 November.

Both EC and alkalinity show only minor variability during January and February 2017

sampling.

The pH of PES was also within the range of pre-earthquake values for all dates in

2016, but was slightly higher than the pre-seismic range in the Feruary and April 2017 measurements. SUS was slightly elevated after the October event, then decreased and was below the pre-seismic range in February 2017. VIC pH decreases outside of the range of previous values for one sampling, 30 August, then was elevated following the

January 18 mainshock (Figure 2c., Table 2). The temperature measured at PES and SUS

(Figure 2d., Table 2) did not deviate outside the range of previous values during the post- earthquake time series. The exception was one measurement of the Peschiera springs on

30 August was abnormally high at 15.9°C. The temperature of VIC in all measurements during the post-earthquake time series was 0.4 – 1.0°C higher than the pre-earthquake range.

CHEMICAL MEASUREMENTS- TRACE ELEMENTS.

Figure 3 shows concentrations of trace elements (Al, Cu, Pb, Sr, Rb, Mn) in time

series of all sampled springs. PES, SUS, and VIC sampling began on Aug. 27th 2016, three days after first earthquake, while the first sampling date at NER was Sep. 29th. All trace element concentrations for the time series and limits of detection for analysis (LOD) are listed in Table 3, including others (Cr, Co, U, Fe, Ni) not displayed in Figure 3 but had similar trends.The Figure 3 subset are considered representative of the trace element response. Compared with the concentrations in Sep. 2015, all Rieti area spring trace element concentrations in Table 3, except Rb in SUS and Sr in VIC, had higher concentrations than pre-earthquake values initially following the Aug. 24th earthquake.

The concentrations of Al, Mn, Pb, Co, Fe and Ni in the Rieti springs were strongly correlated, with Pearson correlation coefficients >0.9 (Tables 4a.-c.). The trace elements

showed two main peaks in concentration in PES concurrent with the Aug. and Oct.

mainshocks, with a smaller peak above pre-earthquake values at the Jan. mainshock. At

SUS, the peak concentrations occurred after the Aug. mainshock, then these elements

gradually decreased in concentration over the sampling period. Rb and U were highly

correlated at all springs and exhibited a slight peak in concentration following the Aug.

mainshock (Rieti springs), with a gradual decrease over the rest of the time series. PES

values of these elements peaked during the Aug. sampling then again at the Oct. 31

sampling. Sr concentrations are high after the Aug. 24th event but do not show major

change after the Oct. 30th event, then increase again at the end of the sampling (Feb., Apr.

2017) in PES and SUS. The Sr concentration at NER did not show significant variability.

NER Al, Cu, Mn, Pb and Ni concentrations also exhibited minor peaks during mid-

November as well as mid-January, prior to the Jan. 18th mainshock.

Tables 4a., b., c., and d., show the results of the Pearson correlation tests for trace

element concentrations and precipitation amounts. Strong correlations (>0.5) were found

among many elemental concentrations over the time series, but precipitation was poorly correlated (<0.5) with all element concentrations for all springs except for Fe in PES, which was negatively correlated (-0.775).

STABLE ISOTOPES

13 The δ CDIC values (Figure 4) of SUS become 1-2‰ enriched above pre-

earthquake samples during the Aug. and Sep. sampling. The Oct. 31 and Nov. 10 samples

then showed enrichments up to 6‰ greater than before the earthquakes. Values of

13 δ CDIC for PES also showed enrichment above the range of previous values following the

Aug. 24th earthquake, then remained enriched during the post-seismic time series. The

13 st most enriched value of δ CDIC at PES occurred on Oct. 31 . Values at VIC were slightly

heavier than pre-earthquake values enriched in the November sampling dates following

the Oct. mainshocks. The three pre-earthquake samples at NER were all from 2016, with

13 δ CDIC values between -10 and -12‰. The post-earthquake sample values were also

within this range with the exception of those within the days following the Oct.

13 mainshocks, where δ CDIC values become up to 5‰ more negative.

18 34 The δ Osulfate and δ Ssulfate are plotted in time series in Figure 5a. and b. The

18 δ Osulfate of SUS showed no discernible response to seismicity, but for PES the value of

18 δ Osulfate was significantly enriched for two samples after the first mainshock (Aug. 27,

18 28). Values of δ Osulfate of VIC also showed enrichment in two samples Aug. 28 and

34 Aug. 30, while δ Ssulfate did not vary outside of the pre-earthquake range. The co- and

34 post-seismic δ Ssulfate of all three localities is within the variability of our measured pre-

seismic values (Figure 5b).

Values of δ18O and δ2H are plotted with the global meteoric water line,

Mediterranean meteoric water line, and the Central Italian meteoric water line (Longinelli

and Selmo, 2003) in Figure 6. The values of VIC, SUS, PES and NER did not change

significantly post-earthquake as compared to pre-earthquake value ranges. The pre-

earthquake values were collected in all seasons and did not exhibit significant seasonal

variability during the years sampled.

DISCUSSION

NATURAL VARIABILITY

For this investigation, groundwater flow and recharge rates should be considered

so that variations in groundwater chemistry can be ruled out as seasonal or dependent on

meteorological conditions. Previous work on the recharge and discharge processes of

SUS and PES estimated aquifer residence times on the order of 15-25 years. These

calculations, however, do not reflect the dual-flow nature of these aquifers, where basal

spring discharge contains a mixture of water from the fast (on the order of days), and slow (on the order of years to tens of years) flow paths (Amoruso et al. 2011; Petitta et al.

2011). The Mediterranean precipitation patterns characterizing this region predict that the

majority of aquifer recharge occurs in the wetter winter months (Costantini et al. 2013).

To test that this seasonality was not also influencing the observed hydrochemical trends

these springs, the fastest possible recharge to discharge flow path (i.e. days) was

considered in testing for trace element concentration correlation with each other and with

precipitation amounts during the month prior to sampling (Tables 4a., b., c., and d.). The

precipitation between samplings did not correlate with measured trace element

concentrations, indicating that water recharging between an earthquake event and

sampling did not influence these measurements. The Fe concentration in PES was

negatively correlated with precipitation and could indicate a relationship between this

element and rainfall amounts, though it is unlikely because this is the only spring and

element where this association exists. Data collected from these springs in 2014-2015

(pre-earthquake values) included major ion chemistry (Archer et al. 2016). These ion

concentrations did not vary significantly among seasons. This lack of increase at SUS or

PES in the fall or winter, or following low-precipitation months, supports a large storage capacity of these carbonate aquifers.

NEREA SPRING

The response in EC and alkalinity at NER was transient but also substantial, with changes of more than 20% from reference values, which serves as a pre-earthquake

“baseline” (Table 1, Figures 2a. and b.). Baseline conditions returned within three weeks of the major earthquake event in all three parameters, suggesting stability was regained.

An interesting feature of these responses was the abrupt increase on October 26, when a magnitude 5.5 and 6.1 occurred in the same day. The epicenter of the 6.1 event

(42.956°N 13.067°E) was just as close to NER as the stronger events on Aug. 24 and

Oct. 30. These parameters recovered slightly, then peaked again over the week of Nov. 3

– Nov. 8. Based on the hydrogeological description of this aquifer, the fast flow allows for a rapid response to seismicity. The recovery from peak values is the main difference between the M 6.1 Oct. 26 event and the M 6.6 Oct. 30 event. The stronger event on Oct.

30 was associated with changes in chemistry that lasted for ~10 days, while the less- strong event caused an EC and alkalinity elevation for ~2 days. The proposed mechanism for these co-seismic increases in concentrations is transient increased aquifer pore pressure clearing congested fractures (fault “cleaning”, Falcone et al. 2012) or pore

spaces (Galassi et al. 2014). This allows for the release of higher salinity water that had

undergone more water-rock interaction (Hartmann and Levy 2006; Liu et al. 2010).

Figure 7, a conceptual diagram of this process, depicts this mechanism at the micro- and

macro-scales before and immediately after the earthquake mainshocks. Interestingly,

NER alkalinity and EC did not exhibit significant changes during or after the Jan. 18th,

2017 seismic sequence. The January earthquake, though, occurred 30-40 km to the south,

so the distance of the fault system that ruptured could possibly explain the absent or

muted response. The magnitude of the mainshock for this event was also lower (Mw 5.5),

so a threshold for earthquake strength and chemical response may also be a contributing

factor. Alkalinity and EC were not analyzed through January 2017 at the Rieti area

springs so it is unknown if the lack of response to the Jan.18 event was characteristic to

NER.

The NER trace element concentrations show prominent peaks occurred on Nov.

2nd and Nov. 15th, with a minor peak on Jan. 11th. Without concentrations prior to the entire seismic sequence, the context of these peaks is not well understood, but the increased concentrations are significant relative to the average concentrations of the time series. The first peak appears to occur in response to the Oct. 26th and/or Oct. 30th

mainshocks, while the Nov. 15th peak could be due to response delay or travel time to the

spring or could be due to a separate shock with lower magnitude, as frequent >4.0 Mw

aftershocks occurred throughout November, including a 4.1 Mw earthquake on Nov. 14th

with an epicenter <1 km from the Nerea spring (Chiaraluce et al. 2017). The coherent behavior of the elements and their apparent response during the seismic period suggests that an increased input of groundwater with higher trace element concentration occurred due to the same mechanism described above for addition of groundwater with higher salinity.

RIETI SPRINGS

Elevated electrical conductivity (EC) and alkalinity in SUS and PES is apparent after the first mainshock (Table 2, Figures 2a and b). The similarities observed in these two parameters are expected, as the bicarbonate ion makes up the majority of alkalinity in these waters and the concentration of this ion is also a component of EC. The precise timing of onset of the increase was not captured because the first sampling date was 3 days after the main earthquake event, but over the four weeks following the event, the return to pre-earthquake values occurs gradually. After the second main shock (October

30th), the EC gradually increases above the pre-earthquake range, though apparently delayed and with a smaller increase relative to the Aug. 24th response. The same mechanism described above for NER is proposed, where cleared solutes from longer- residence time pore spaces and microfractures input higher salinity fluids to the main flow circuit (Figure 7). This is supported by the preponderance of the increase in dissolved parameters after the first major earthquake event and not during or after

subsequent (larger) events, as once these spaces were flushed there were fewer ions left in the pore spaces that had time to undergo significant water-rock interaction to source the same signal.

The initial elevated concentrations in most measured trace element ions in the

Rieti area springs during the seismic sequence compared to 2015 suggests the same mechanism occurring as in NER. Most trace elements at SUS (Al, Mn, Fe, Pb, Cr, Ni,

Co) showed the greatest increase after the first major earthquake (Aug. 24th), then less or no increase after the Oct. 26th, 30th 2016, and Jan 18, 2017 events (Figure 3). This could be because the first event effectively cleared, or flushed out, the pore spaces or fractures where longer mean residence time groundwater collected (Figure 7; Pasvanoglu et al.

2004; Galassi et al. 2014). Subsequent mainshocks would have had the same effect, but the groundwater driven out would have had days instead of multiple years of water-rock interaction. PES concentrations of Al, Mn, Fe, Cu, Pb, Co, and Ni exhibited two prominent peaks: after the Aug. 24th mainshock and the Oct. 26th and/or Oct.30th mainshocks. It is possible that aquifer heterogeneity and the different fault segments that ruptured could have allowed for different reservoirs of higher salinity groundwater in the

PES aquifer to be expelled following each event. More information on the aquifer’s internal structure is needed to confirm this hypothesis, however. The concentrations of

Rb, Sr, and U did not show the same progressive decrease over the time series as the other elements, but remained elevated above 2015 values for the entire sampling.

Variability in response among measured elements and among measured springs could

also be due to the properties of the elements (redox potential, sorption; Appelo and

Postma 2004). For instance, Al concentration is affected by the distribution of clays in aquifers (Morgantini et al. 2009), so the high concentrations of Al in VIC, where recharge and discharge occur in a thick alluvial unit (Martarelli et al. 2008), can be attributed to more available clays within the pore matrix that can be released from co- seismic shaking. Though Sr concentrations are grouped with other trace elements, its concentration in these waters is relatively high, particularly in SUS (~5000-8000 ppb, or

~5-8 mg/L), as this spring discharges groundwater with the greatest proportion of water flowing through the deeper Triassic carbonate and evaporite units where groundwater has more interaction time with Sr-bearing minerals. The concentration of Sr post-earthquake is different for other measured trace elements, potentially because its association with carbonate rocks. The peaks after the Aug. mainshock in SUS and PES could be from the same fracture and/or pore space clearing mechanism described above, then the subsequent declines may be affected by the precipitation of SrCO3 and/or SrSO4 that can

occur at pH values >7.5 (De Vos et al. 2006). The elevations in pH in SUS and PES post-

mainshock may have contributed to Sr precipitation and lower concentrations after the

initial flushing of this ion. Though reactions along flow paths may explain the profiles of

dissimilar trace element trends, the post-mainshock peaks in concentration of most trace

elements at all springs are attributed to dilation of pore spaces releasing slower flow-path

groundwater, allowing it to mix in greater proportions with the fast-moving circuit.

18 2 The values of δ OH2O and δ HH2O in Central Apennines groundwaters are

primarily controlled by recharge elevation (Longinelli and Selmo 2003; Petitta et al.

2011; Tallini et al. 2015). Hydrochemical responses have been observed in different

18 settings that include post-seismic changes in δ OH2O values. These changes were

attributed to aquifer breaching causing a change in aquifer structure or mixing of

different groundwater components (Claesson et al. 2004; Reddy et al. 2011; Skelton et al.

2014). The lack of a change in water isotope values of PES, SUS and VIC post-

earthquake (Figure 6) supports the assumption that increases observed in groundwater

chemical constituents are a result of processes occurring within the aquifer instead of

addition of water from another aquifer. The residence time in these aquifers is sufficiently

long so that aquifer flowpath changes would not be evident in the δ18O and δ2H values

until several years after the seismic sequence, if at all, especially because both SUS and

PES are basal springs and represent an integration of flow within the aquifers (Civita and

Fiorucci 2010, Spadoni et al. 2010). This intra-annual stability is also displayed by the

lack of variation in δ18O and δ2H during the pre-earthquake sampling period (2014-2015,

Figure 6).

The primary sources of dissolved inorganic carbon in these aquifers is dissolution

of carbonate rock along flow paths, organically-derived soil CO2 dissolved during infiltration, and/or CO2 dissolved in mineralized water with a deep-flow circuit and

longer mean residence time (Chiodini et al. 2000, Petitta et al. 2011, Raco et al. 2013).

The δ13C of the carbonate platform comprising Mt. Terminillo and the Reatini Mountains

ranges from ~ +2‰ to +3‰, while in the north, near Nerea, carbonate δ13C range from

~+2‰ to +3.5‰ (Morettini et al. 2002). The contribution of carbonate dissolution to δ13C

13 of DIC in this region is 2.21‰ ± 0.66‰. The range of δ CDIC in central Apennine

groundwaters containing mantle-derived CO2, calculated using a carbon mass-balance together with isotopic and hydrogeological data, is -5‰ to -1‰ (Chiodini et al. 2000;

2004). The post-mainshock δ13C values measured in this study (-5‰ to -3‰, Figure 4)

fall within this range of previously calculated groundwater with contribution of mantle-

derived CO2. Alternatively, or together with the addition of deep-aquifer CO2, increased

contribution of groundwater from longer residence time reservoirs, flow paths, or matrix

porosity with a greater degree of water-rock interaction and marine carbonate isotopic

13 signature could have caused this shift to heavier δ CDIC. The lack of a consistant post-

seismic pH trend at NER, SUS, and VIC (Figure 2c.) and the rapid onset of the carbon

13 isotope enrichment suggest multiple drivers of the observed post-seismic δ CDIC

increase. The enhanced co-seismic hydraulic conductivity in the aquifer may be

responsible for upwelling of CO2-rich waters (“seismic pumping”) that were contained in

deep crustal reservoirs (Quattrocchi 1999; Miller et al. 2004), or, as suggested by

Chiodini et al. (2004) the high pore-pressure from mantle-derived CO2 in these deep

reservoirs may have instigated the fault rupture of these earthquakes and released CO2.

The deep-flow circuit of these aquifers have a similar recharged in the same area and

18 2 have similar δ OH2O and δ HH2O (Petitta et al. 2011), so the addition of more deep-cirsuit

groundwater would not be expected to cause water isotope change in spring waters. The

enrichment of δ13C of DIC in PES and SUS following the main shocks on Aug. 24 and

Oct. 30 was most likely induced by the addition of this deep-sourced CO2 that traveled

along fault conduits, but a pulse of isotopically-heavy DIC from dissolution of marine carbonates cannot be ruled out as an additional contribution.

The stability of δ34S of sulfate during and after the seismic sequence suggests that

there was no change in source of sulfur to aqueous sulfate (figure 5b.). The main source

of S to the waters, dissolution of Triassic evaporite deposits within the aquifer, has an

34 18 isotopic signature with δ S values ranging from +10 to +20‰ and δ Osulfate ranging

34 from +5 to +14‰ (Cortecci et al. 2002). Values of δ Ssulfate retains the signature of this

source regardless of sulfate concentration, as microbial sulfate reduction has little effect

18 on these waters (Petitta et al. 2011, Archer et al. 2016). The δ Osulfate of PES was

enriched in samples following the first mainshock (Figure 5). Most of the oxygen in

aqueous sulfate comes from oxygen dissolved from the Triassic evaporites, but up to 25

percent can be from atmospheric oxygen dissolved in groundwater (δ18O = +23.8‰)

(Drever 2005). There was a significant increase in dissolved oxygen in groundwater of

the Gran Sasso aquifer following the 2009 L’Aquila earthquake attributed to faster flow

rate (Galassi et al. 2014). It is possible that the immediate enrichment in PES could be from faster flow rate and/or new flow paths along earthquake-cleared fractures dissolved

a greater proportion of atmospheric oxygen in groundwater, but this would imply that

34 oxidation of sulfide occurred. This should also change δ Ssulfate values, but without any

18 change observed the variability in δ Osulfate of PES remains enigmatic.

VIC sources a small alluvial aquifer compared to SUS, PES, and NER,

characterized by local recharge that may be influenced by the soil zone processes in the

Rieti Plain. VIC also exhibited greater seasonally variable chemical and isotopic values

before the earthquake activity (Archer et al. 2016) than PES and SUS, so attributing any

observed post-earthquake changes to seismicity is difficult. Responses in VIC EC,

alkalinity, and most trace element concentrations, though, correlate well with the regional

carbonate aquifers, suggesting that this aquifer may be also susceptible to changes in pore

13 pressure from seismicity. The δ CDIC at VIC, however, did not change significantly from

pre-earthquake values, which is expected at a spring sourcing an aquifer that lacks

extensive faulting and connection to a deeper flow circuit.

INTERPRETATION OF HETEROGENEITY

Comparing trace element concentration trends in figure 3, the Rieti Springs (SUS,

PES) generally show a decreasing trend from August 2016 to April 2017, while NER

exhibits several distinct peaks. This apparent difference in duration of chemical signals

may be a result of differences in accommodation of pore pressure in response to fault

movement or differences in the complexities of karst aquifer structure and micro-fracture networks. The two mainshock epicenters were ~15km apart (Figure 1), so the Rieti

springs were near-field for the Aug. 24th and January 18th mainshocks but intermediate-

field for the Oct. 26th and 30th shocks. The Aug. 24th epicenter was also located along the

Olevano-Antrodoco-Sibillini thrust front, which trends NNE-SSW and connects with the

SE edge of the Rieti Basin. This structure served as a dividing line for fault ruptures, especially after the Mw 6.5 event on Oct. 30th, when all seismic activity was confined to

normal faults to the north of this line (Chiaraluce et al. 2017). This may also explain the

disparity between SUS and PES in hydrochemical response to the Oct. 26th / Oct. 30th events, where peaks in trace elements were more prominent in PES. Also, compared to

NER, where EC and alkalinity show an abrupt and significant increase following Oct.

30th, PES and SUS appear to have a delayed response (1-2 weeks later). The reason for timing differences is complex, as either flow rate from aquifer source to spring from fault could be responsible, or a different fault rupture in the complex seismic sequence could

be responsible.

Trace elements of all springs, NER alkalinity and EC and Rieti springs EC

experienced peak levels that persist for days- weeks following earthquakes, returning to

13 pre-earthquake or baseline values within the study period. The δ CDIC and alkalinity

PES, however, remain elevated for several months (Aug. 27th- Nov. 28th 2016). The

persistence of the enrichments in these parameters could be a function of the additional

mechanism for enrichment by the input of deep-sourced dissolved CO2 where faulting created new conduits for upwelling of water from the deep aquifer flow circuit. Or, an alternative explanation for the heterogeneity in duration of parametric responses could be earthquake-induced flow path changes where changes in the aquifer pressure feld and hydraulic conductivity caused groundwater flow through previously abandoned circuits where the initial flow-through represented the initial “flushing out” of erodible constituents. Similar to this mechanism, a co-seismic transient water table rise, smilar to that observed with other major earthquakes in this region (Falcone et al. 2012; Petitta et al. 2015), could introduce groundwater and flow through dry channels previously above the water table and/or epikarst. The epikarst zone, or the uppermost zone of some carbonate bedrock that is extensively fissured and dissolved (Kilmchouk 1995), may provide the additional transient trace elements in solution that decrease in concentration within months as the water table decreases to normal levels.

CONCLUSIONS

A response in physio-chemical parameters and stable isotope values to the 2016-

2017 seismic sequence has been documented by the major carbonate aquifer springs

SUS, PES, and NER. This study found that both near-field and intermediate-field springs were affected by the fault movement and/or ground shaking induced aquifer pore pressure change, and that these effects were transient in nature during this earthquake sequence. The three large aquifers sourced by these springs are similar in hydrogeologic structure but different in proximity to the ruptured faults. The responses, though notable for both near and intermediate field, were different in duration and onset for the three major springs. These complexities confirm a pore pressure response of aquifers to seismic strain and show the probable role of aquifer hydrogeological structure when considering earthquake effects. The enrichment of δ13C of DIC in PES and SUS

following the 2016 main shocks was likely influeced by input of deep-sourced CO2

initiated by movement on faults that serve as conduits, in addition or instead of the

mechanism than the other chemical enrichments observed where groundwater with more

water-rock interaction time is mixed in greater contributions as a post-seismic pulse. The

18 2 lack of change in δ OH2O and δ HH2O after the seismic series verifies that groundwater

chemical changes are the result of co-seismic shaking and chemical evolution and/or flow

path changes within the aquifer instead of mixing with another aquifer or aquifer

breaching. These data provide information on short-term and mid-term responses, however, and the long-term response is dependent on continued monitoring. With

continued analysis and quantification of earthquake effects, the propagation of enriched

13 δ CDIC and the impacts of increased mixing with deep fluids may be understood. The

effects of seismicity on these aquifers, particularly PES and NER that are major drinking

water resources, will be a vital outcome of this work.

REFERENCES Amoruso, A., Crescentini, L., Petitta, M., Rusi, S., & Tallini, M. (2011). Impact of the 6 April 2009 L'Aquila earthquake on groundwater flow in the Gran Sasso carbonate aquifer, Central Italy. Hydrological Processes, 25(11), 1754-1764. Appelo, A. J. & Postma, D. (2004) Geochemistry, Groundwater and Pollution. Second Edition. CRC Press. Archer, C., Noble, P., Kreamer, D., Piscopo, V., Petitta, M., Rosen, M.R., Poulson, S.R., Piovesan, G. and Mensing, S. (2016) Hydrochemical determination of source water contributions to Lake Lungo and Lake Ripasottile (central Italy). Journal of Limnology. Capelli, G., Mastrorillo L., Mazza R., Petitta M., Baldoni T., Banzato F, Doredana C., Di Salvo C., La Vigna F., Taviani S., & Teoli P.(2012). [Carta Idrogeologica del Territorio della Regione Lazio].[Map in Italian]. Regione Lazio, Rome. Carmody, R.W., Plummer L.N., Busenberg E. & Coplen T.B. (1998). Methods for collection of dissolved sulfate and sulfide and analysis of their sulfur isotopic composition. U.S. Geological Survey, Open-File Report 97-234: 91 pp. Carro, M., De Amicis M., & Luzi, L. (2005) Hydrogeological changes related to the Umbria–Marche earthquake of 26 September 1997 (Central Italy). Natural Hazards 34: 315–339. Cavinato, G. P., & De Celles, P. G. (1999). Extensional basins in the tectonically bimodal central Apennines fold-thrust belt, Italy: response to corner flow above a subducting slab in retrograde motion. Geology, 27(10), 955-958. Cello, G., Mazzoli, S., Tondi, E. & Turco, E. (1997) Active tectonics in the central Apennines and possible implications for seismic hazard analysis in peninsular Italy. Tectonophysics 272(1), 43–68.

100

“Central Italy.” 42.463263 N 12.815350 E. Google Earth, Dec. 30 2016. June 5 2017 Chiaraluce, L., Stefano, R., Tinti, E., Scognamiglio, L., Michele, M., Casarotti, E., Cattaneo, M., Gori P., Chiarabba C., Monachesi G., Lombardi A., Valoroso L., Latorre D., Marzorati S. (2017) The 2016 Central Italy Seismic Sequence: A First Look at the Mainshocks, Aftershocks, and Source Models. Seismological Research Letters 88(3), 757–77. Chiodini, G., Frondini, F., Cardellini, C., Parello, F. & Peruzzi, L. (2000) Rate of diffuse carbon dioxide Earth degassing estimated from carbon balance of regional aquifers: The case of central Apennine, Italy. Journal of Geophysical Research: Solid Earth 105(B4), 8423–8434. Chiodini, G., Cardellini, C., Amato, A., Boschi, E., Caliro, S., Frondini, F. & Ventura, G. (2004) Carbon dioxide Earth degassing and seismogenesis in central and southern Italy. Geophysical Research Letters 31(7) Claesson, L., Skelton, A., Graham, C., Dietl, C., Mörth, M., Torssander, P., & Kockum, I. (2004). Hydrogeochemical changes before and after a major earthquake. Geology, 32(8), 641-644. Cortecci, G., Dinelli, E., Bencini, A., Adorni-Braccesi, A. & Ruffa, G. (2002) Natural and anthropogenic SO4 sources in the Arno river catchment, northern Tuscany, Italy: a chemical and isotopic reconnaissance. Applied Geochemistry 17(2), 79– 92. Costantini, E., Dazzi C, Costantini E., Fantappié M., & L’Abate G. (2013). Climate and Pedoclimate of Italy, p. 19-37. In: E Costantini and C Dazzi (eds.) The Soils of Italy. De Vos W., Tarvainen T. (Chief-editors), Salminen R., Reeder S., De Vivo B., Demetriades A., Pirc S., Batista M.J., Marsina K., Ottesen R.-T., O'Connor P.J.1, Bidovec M., Lima A.4, Siewers U., Smith B., Taylor H, Shaw R., Salpeteur I.1, Gregorauskiene V., Halamic J., Slaninka I., Lax K., Gravesen P., Birke M., Breward N., Ander E.L., Jordan G., Duris M., Klein P., Locutura J., Bel-lan A., Pasieczna A., Lis J., Mazreku A., Gilucis A., Heitzmann P., Klaver G., Petersell V. (2006) Geochemical atlas of Europe. Part 2–interpretation of geochemical maps, additional tables, figures, maps. and Related Publications. Geological Survey of Finland. Espoo, Finland. Geol, 98(2), 69-82. Drever JI. (2005) Surface and Ground Water, Weathering, and Soils: Treatise on Geochemistry, Second Edition, Elsevier. 5.

101

Epstein S., & Mayeda T., (1953). Variation of O 18 content of waters from natural sources. Geochimica et Cosmochimica 4:213-224. Falcone, R.A., Carucci, V., Falgiani, A., Manetta, M., Parisse, B., Petitta, M., Rusi, S., Spizzico, M. and Tallini, M., (2012). Changes on groundwater flow and hydrochemistry of the Gran Sasso carbonate aquifer after 2009 L’Aquila earthquake. Italian Journal of Geosciences, 131(3), pp.459-474. Galassi, D.M., Lombardo, P., Fiasca, B., Di Cioccio, A., Di Lorenzo, T., Petitta, M. and Di Carlo, P., (2014) Earthquakes trigger the loss of groundwater biodiversity. Scientific reports, 4, p.6273. Giesemann, A., Jäger, H. J., Norman, A. L., Krouse, H. R., & Brand, W. A. (1994). Online sulfur-isotope determination using an elemental analyzer coupled to a mass spectrometer. Analytical Chemistry, 66(18), 2816-2819. Grassineau, N. V., Mattey, D. P., & Lowry, D. (2001). Sulfur isotope analysis of sulfide and sulfate minerals by continuous flow-isotope ratio mass spectrometry. Analytical Chemistry, 73(2), 220-225. Harris, D., Porter, L. K., & Paul, E. A. (1997). Continuous flow isotope ratio mass spectrometry of carbon dioxide trapped as strontium carbonate. Communications in Soil Science & Plant Analysis, 28(9-10), 747-757. Hartmann, J. & Levy, J. (2006). The influence of seismotectonics on precursory changes in groundwater composition for the 1995 Kobe earthquake, Japan. Hydrogeology Journal 14(7), 1307–1318. Klimchouk, A.B. (1995). The nature and principal characteristics of epikarst. In: Proceedings of 12th international Congress of speleology, La Chaux-de-Fonds, 306 pp. Kornexl, B. E., Gehre, M., Höfling, R., & Werner, R. A. (1999). On-line d18O measurement of organic and inorganic substances. Rapid Communications in Mass Spectrometry, 13(16), 1685-1693. Liu, C.-P., Wang, C.-H. & Hwang, L.-S. (2010). Temporal variation of seepage water chemistry before and after the Hengchun Ms 7.2 earthquake in south Taiwan. Geoderma 155(1-2), 107–114. L'Ufficio Idrografico e Mareografico, Regione Lazio. (2017, June) Dati Pluviometri. Retrieved from http://www.idrografico.roma.it/annali/ (June 6 2017).

102

Longinelli, A., & Selmo, E. (2003) Isotopic composition of precipitation in Italy: a first overall map. Journal of Hydrology 270(1-2): 75–88. Malakootian, M., & Nouri, J. (2010) Chemical variation of ground water affected by the earthquake in Bam Region. International Journal of Environmental Research 4(3):443–454 Manga M., & Wang C.-Y (2015) Earthquake Hydrology. In: Gerald Schubert (editor-in- chief) Treatise on Geophysics, 2nd edition, Vol 4. Oxford: Elsevier. 305-328. Martarelli L., Petitta M., Scalise A., & Silvi A., (2008) [Cartografia idrogeologica sperimentale della Piana Reatina (Lazio)].[Article in Italian]. Memorie descrittive della Carta Geologica d'Italia 81:137-156 Mastrorillo, L., & Petitta, M. (2014). Hydrogeological conceptual model of the upper Chienti River Basin aquifers (Umbria-Marche Apennines). Italian Journal of Geosciences, 133(3), 396-408. Miller, S. A., Collettini, C., Chiaraluce, L., & Cocco, M. (2004). Aftershocks driven by a high-pressure CO2 source at depth. Nature, 427(6976), 724. Morettini, E., Santantonio, M., Bartolini, A., Cecca, F., Baumgartner, P. & Hunziker, J. (2002) Carbon isotope stratigraphy and carbonate production during the Early– Middle Jurassic: examples from the Umbria–Marche–Sabina Apennines (central Italy). Palaeogeography Palaeoclimatology Palaeoecology 184(3-4), 251–273. Morgantini, N, Frondini F, Cardellini C, (2009) Natural trace elements baselines and dissolved loads in groundwater from carbonate aquifers of central Italy, Physics and Chemistry of the Earth, Parts A/B/C, 34 (8–9), 520-529 Morrison, J., Brockwell, T., Merren, T., Fourel, F., & Phillips, A. M. (2001). On-line high-precision stable hydrogen isotopic analyses on nanoliter water samples. Analytical Chemistry, 73(15), 3570-3575. Pasvanoglu, S., Canik B., & Rosen M.R., (2004) Hydrogeology and possible effects of the M w 7.4 Marmara Earthquake (17 August 1999) on the spring waters in the Orhangazi- Bursa area, Turkey. Journal Geological Society of India. 63 313-322. Petitta, M., Primavera, P., Tuccimei, P. & Aravena, R. (2011) Interaction between deep and shallow groundwater systems in areas affected by Quaternary tectonics (Central Italy): a geochemical and isotope approach. Environmental earth sciences, 63(1), pp.11-30.

103

Petitta, M., Caschetto, M., Galassi, D. M., & Aravena, R. (2015). Dual-flow in karst aquifers toward a steady discharge spring (Presciano, Central Italy): influences on a subsurface groundwater dependent ecosystem and on changes related to post- earthquake hydrodynamics. Environmental Earth Sciences, 73(6), 2609-2625. Quattrocchi F (1999) In search of evidence of deep fluid discharge and pore pressure evolution in the crust to explain the seismicity style of the Umbria-Marche 1997- 1998 seismic sequence (Central Italy). Raco B, Dotsika E, Feroni A, et al. (2013). Stable isotope composition of Italian bottled waters. Journal of Geochemical Exploration 124: 203–211. Reddy, D., Nagabhushanam, P. & Sukhija, B. (2011). Earthquake (M 5.1) induced hydrogeochemical and δ18O changes: validation of aquifer breaching—mixing model in Koyna, India. Geophysics Journal International 184(1)., 359–370. Skelton, A., Claesson, L., Chakrapani, G., Mahanta, C., Routh, J., Mörth, M. & Khanna, P. (2008) Coupling Between Seismic Activity and Hydrogeochemistry at the Shillong Plateau, Northeastern India. Pure Applied Geophysics 165(1), 45–61. Skelton, A., Andrén, M., Kristmannsdóttir, H., Stockmann, G., Mörth, C.M., Sveinbjörnsdóttir, Á., Jónsson, S., Sturkell, E., Guðrúnardóttir, H.R., Hjartarson, H. & Siegmund, H., (2014). Changes in groundwater chemistry before two consecutive earthquakes in Iceland. Nature Geoscience, 7(10), pp.752-756. Spadoni M, Brilli M, Giustini F, et al. (2010) Using GIS for modelling the impact of current climate trend on the recharge area of the S. Susanna spring (central Apennines, Italy). Hydrological Processes 24(1): 50–64. Tallini M, Falcone R, Carucci V, et al. (2014) Isotope hydrology and geochemical modeling: new insights into the recharge processes and water–rock interactions of a fissured carbonate aquifer (Gran Sasso, central Italy). Environmental Earth Sciences 72(12): 4957–4971. Usdowski, E., Hoefs, J., & Menschel, G. (1979). Relationship between13C and18O fractionation and changes in major element composition in a recent calcite- depositing spring—A model of chemical variations with inorganic CaCO3 precipitation. Earth and Planetary Science Letters, 42(2), 267-276. Valensise, G, Amato, A, Montone, P & Pantosti, D. (2003) Earthquakes in Italy: Past, present and future. Episodes 26:3 245-249

104

Vigna, L. F., Carucci, V & Mariani, I. (2012). Intermediate-field hydrogeological response induced by L’Aquila earthquake: the Acque Albule hydrothermal system (Central Italy). Italian Journal of Geosciences. (Boll. Soc. Geol. It.) 131 (3): 475- 485 Woith, H., Wang, R., Maiwald, U., Pekdeger, A. & Zschau, J. (2013). On the origin of geochemical anomalies in groundwaters induced by the Adana 1998 earthquake. Chemical Geology 339, 177–186 Zuppi, G. M., & Bortolami, G. C. (1982). Hydrogeology: a privileged field for environmental stable isotopes applications. Some Italian examples. SIMP, 38(3), 1197-1212.

105

Figure 1. Study map. Red lines indicate active faults. White stars indicate earthquake mainshocks, from north to south: Oct. 26th 2016 (42.9087, 13.1288), Oct. 30th 2016 (42.8322, 13.1107), Aug. 24th 2016 (42.6983, 13.2335), Jan. 18th 2017. Orange and yellow stars indicate aftershocks with magnitude greater than 4.0. Orange denotes depths <10 km, Yellow denotes depths >10 km. The blue flags indicate locations of springs sampled, from north to south: NER, SUS, VIC, PES. The yellow outline indicates the SUS recharge area (Spadoni et al. 2010), the white outline indicates PES (Civita and Fiorucci 2010), and the green outline denotes NER recharge area (Tarragoni 2006). Data from: INGV, ISIDe working group (2016) version 1.0, DOI: 10.13127/ISIDe

106

107

Figure 2. Time series plots of A) alkalinity (mg/L as CaCO3), B) Electrical conductivity (EC), C) pH and D) Temperature (ͦC) from SUS, VIC, PES, and NER. The vertical black lines indicate earthquake mainshocks (Aug 24th, Oct. 26th, Oct 30th, Jan. 18th), horizontal shaded fields indicate the pre-earthquake ranges for SUS, VIC, and PES, with the color of shading corresponding to the spring sampled (grey=SUS, blue=VIC, purple=PES and orange=NER). Pre-earthquake values for NER were in the form of a reference value provided by the Nerea S.p.a. bottling plant, and these values are plotted as the 2014-2015 sampling values.

108

109

Figure 3. Time series plots of trace metals at SUS, VIC, PES, and NER. Vertical black lines indicate earthquake mainshocks (Aug 24th, Oct. 26th, Oct 30th, Jan. 18th), horizontal gray fields indicate the pre-earthquake range of values for Al, Cu, Pb, and Mn, where all springs had the same range. The plots of Sr and Rb have dashed horizontal lines to indicate the pre-earthquake values at each individual spring. The pre-earthquake concentration of Sr at SUS (1770 ppb) is not visible on the graph using the current y-axis scale.

Figure 4. δ13C (‰) of dissolved inorganic carbon (DIC) of the Rieti area springs. The vertical black lines indicate earthquake mainshocks (Aug 24th, Oct. 26th, Oct 30th, Jan. 18th), horizontal shaded fields indicate the pre-earthquake ranges of values for SUS, VIC, and PES, with the color of shading corresponding to the spring sampled (purple=PES grey=SUS, blue=VIC, and orange=NER).

110

Figure 5. A) δ18O of sulfate and B) δ34S of sulfate of the Rieti area springs in time series. Shaded fields indicate pre-earthquake ranges of values, with colors corresponding to each spring.

111

Figure 6. Stable isotopes of water from springs pre- and post- earthquake compared to global, central Italian, and Mediterranean meteoric water lines (GMWL, cIMWL, and MMWL; Longinelli and Selmo 2003) and rain-gauge measurements at 2 elevations in the central Apennines (375, 1375 m.a.s.l).

112

Figure 7. Conceptual Diagram of groundwater flow in the regional carbonate aquifers associated with the SUS, PES, and NER springs. A) shows pre-seismic condition at the macro-scale (frame represents ~10m wide, where x>>z), where highest flow rate occurs through major karst conduits, slow flow through fractures in the carbonate bedrock and quasi-stagnant to stagnant flow in the micropores of the rock matrix. B) The macro-scale aquifer condition immediately (days-weeks) following the first earthquake mainshock, C) shows the micro-scale (frame represents ~10cm) pre-seismic condition and D) shows the co-seismic and immediately post-seismic condition. In the post-seismic condition there may be less long mean residence time water (light blue)remaining in the fracture network

113

and some may be replace by the shorter mean residence time water (dark blue) of the fast-flow circuit. Table 1. All parameters, Nerea spring

DATE EC µS/cm Hardness Alkalinity (mg/L (mmol/L) CaCO3) 27-Sep-16 314 1.6 155 12-Oct-16 268 1.4 145 18-Oct-16 267 1.5 140 26-Oct-16 272 1.7 140 26-Oct-16 278 1.6 148 28-Oct-16 271 1.6 130 2-Nov-16 288 1.5 143 3-Nov-16 312 1.6 145 4-Nov-16 290 1.4 150 7-Nov-16 286 1.4 190 8-Nov-16 285 1.7 190 9-Nov-16 287 1.6 170 10-Nov-16 285 1.8 180 10-Nov-16 288 1.7 160 15-Nov-16 287 1.6 160 16-Nov-16 287 1.6 160 17-Nov-16 286 1.6 165 18-Nov-16 254 1.7 165 23-Nov-16 247 1.5 153 29-Nov-16 250 1.5 148 6-Dec-16 248 1.5 153 9-Jan-17 243 1.5 150 11-Jan-17 230 1.5 143 12-Jan-17 247 1.5 153 25-Jan-17 247 1.5 140 1-Feb-17 232 1.5 143 1-Feb-17 256 1.5 150 8-Feb-17 260 1.5 153 15-Feb-17 239 1.5 143 22-Feb-17 242 1.5 150

114

27-Feb-17 243 1.5 153 1-Mar-17 248 1.5 148

Table 2. All parameters, Rieti springs Data pre-2016 were analized as part of Archer et al. (2016; Chapter 1).

- - DIC V - CDT VSMOW H2O -

H2O

O O C H 18 18 2 13 d δ SAMPLE DATE (d/m/y) T°C pH ECµS/cm Alkalinity (mg/L CaCO3) δ sulfate (‰) sulfate δ34S (‰) δ Vicenna 3-Jul-14 13.6 7.34 534 240 -6.9 -43 -0.7 -9.9 -13.8 Riara 15-Feb- 13 7.4 527 - -7.1 -43 5.2 8.8 -13.5 (VIC) 15 22-May- 13.2 7 506 245 -7.4 -46 4.2 8.6 -12 15 20-Sep- 12.2 7.4 523 250 -7.1 -45 6 15.2 -13.5 15 27-Aug- - 7.06 503 303 -7.1 -45 6.4 7.4 -12.7 16 28-Aug- - 7.11 608 305 -7.1 -45 7.9 7.5 -12 16 30-Aug- 14.2 6.37 529 260 -7.2 -45 6.2 8.1 -12.4 16 6-Sep-16 13.8 7.42 529 258 -7.2 -45 5.7 8.4 -11.4 22-Sep- 13.8 7.4 497 258 16 31-Oct- 14.1 7.49 500 245 -7 -44 5.4 9 -12 16 2-Nov-16 14.1 7.54 520 243 5.6 10-Nov- 250 -7.1 -45 8.2 -11.9 16 28-Nov- 14 7.45 523 245 -7.1 -45 -11.8 16 24-Dec- 7.8 507 285 16 19-Jan-17 7.7 520 260 28-Feb- 7.7 506 220 -12.3 17

115

18-May- 13.5 7.3 500 240 17

Santa 3-Jul-14 10.9 7.44 835 184 -8.6 -53 10.7 15.3 -10 Susann 15-Feb- - 7.6 851 - -8.4 -52 9.3 15.4 -9 a (SUS) 15 22-May- 10.3 7.1 798 193 -8.6 -54 10 14.9 -8.9 15 20-Sep- 10.4 7.4 805 192 -8.4 -53 9.6 15.5 -8.8 15 27-Aug- - 7.35 916 208 -8.6 -53 10.1 15.2 -9.1 16 28-Aug- - 7.22 918 195 -8.6 -53 10.3 15 -9 16 30-Aug- 10.3 7.14 918 190 -8.6 -53 9.3 15.2 -8.4 16 6-Sep-16 10.3 7.05 904 180 -8.6 -53 10.3 14.9 -10.3 22-Sep- 9.8 7.69 872 190 16 31-Oct- 10.3 7.94 852 175 -8.4 -53 10.6 15.3 -6.9 16 2-Nov-16 9.7 7.69 805 185 10.5 10-Nov- 180 -8.5 -53 10.3 15.2 -3.6 16 28-Nov- 9.6 8.07 878 195 -8.4 -53 10.1 14.2 -7.5 16 24-Dec- 7.4 783 200 16 19-Jan-17 7.6 814 213 28-Feb- 6.7 831 185 -6.4 17 28-Apr- 10.8 7.7 851 200 -7.0 17 18-May- 10.4 7.4 822 180 17

Peschie 3-Jul-14 12.8 7.2 615 320 -9 -57 8 10.2 -6.1 ra 15-Feb- 10 7 653 - -8.9 -55 9 12.2 -3.9 (PES) 15 22-May- 11.4 7 594 335 9 -57 8.3 12.1 -4.2 15

116

20-Sep- 12.4 7.2 605 336 -9 -57 9 12.1 -4.8 15 27-Aug- - 7.1 733 380 -8.9 -57 16 11.5 -5.9 16 28-Aug- - 7.11 727 385 -8.9 -57 11.75 12 -3.1 16 30-Aug- 15.9 7.33 690 350 -9 -57 9.9 12 -3.8 16 6-Sep-16 12.9 7.4 676 380 -9 -57 9.6 12.2 -3.5

22-Sep- 11.8 7.4 632 370 16 31-Oct- 11.2 7.43 624 360 -8.7 -56 8.8 11.9 -2.4 16 2-Nov-16 10.8 7.37 634 338 9 10-Nov- 350 -8.9 -57 8.3 11.1 -2.9 16 28-Nov- 9.8 7.4 665 360 -8.9 -57 8.9 11.5 -3.2 16 24-Dec- 7.2 580 365 16 19-Jan-17 7.7 645 370 28-Feb- 7.4 623 370 -3.8 17 28-Apr- 12.2 7.4 676 373 -2.9 17 18-May- 15 7 651 345 17

Table 3. Trace metal concentrations at SUS, PES, VIC, and NER. Limits of detection (LOD) are given in the firs row. All values are in ppb.

date Al Cr Mn Fe Co Ni Cu Rb Sr Pb U

2.4e- .05 .003 .002 .0002 .0047 .0010 .0015 .0002 LOD .00115 5 35 1 2 4 5 5 1 1

27-Aug- 37. 45.1 6607.3 SUS 2.05 0.65 0.12 4.65 1.81 1.36 0.59 1.56 16 61 9 4

117

30-Aug- 18. 10.0 4411.6 1.45 0.40 0.10 3.28 1.00 1.01 0.41 1.23 16 18 9 2

33. 17.5 7186.6 6-Sep-16 1.88 0.62 0.12 4.14 1.67 1.30 0.59 1.46 83 5 7

31. 10.1 5511.2 22-Sep-16 0.76 0.62 0.10 2.35 3.88 1.27 0.59 1.44 56 0 1

12. 4831.1 31-Oct-16 0.63 0.22 4.46 0.09 1.63 1.32 1.08 0.25 1.37 21 5

28-Nov- 4.8 5898.0 0.21 0.23 3.42 0.10 0.72 0.54 1.02 0.15 1.22 16 1 7

2.6 5012.9 24-Dec-16 0.27 0.27 3.84 0.09 0.45 0.74 1.02 0.17 1.15 8 9

2.4 5095.7 19-Jan-17 0.27 0.17 2.12 0.09 0.38 0.37 0.95 0.13 1.12 0 3

1.6 5322.3 28-Feb-17 0.26 0.16 3.58 0.09 0.37 0.17 0.95 0.12 1.19 2 5

0.4 7794.8 28-Apr-17 0.26 0.14 1.24 0.09 0.38 0.11 1.04 0.11 1.15 6 9

Al Cr Mn Fe Co Ni Cu Rb Sr Pb U

27-Aug- 25. 1.60 0.46 8.64 0.10 3.10 1.35 1.70 527.35 0.46 0.70 16 78

30-Aug- 18. 1.55 0.41 8.49 0.10 2.95 1.76 1.81 547.35 0.47 0.72 16 00

8.1 6-Sep-16 0.78 0.25 3.33 0.09 1.03 1.19 1.81 554.03 0.35 0.73 5

16. PES 22-Sep-16 0.96 0.30 6.30 0.10 1.82 1.69 1.81 569.47 0.38 0.71 59

27. 10.8 31-Oct-16 0.96 0.57 0.10 2.16 2.38 1.69 466.87 0.58 0.63 62 7

28-Nov- 2.4 0.45 0.19 1.75 0.09 0.52 0.57 1.63 123.65 0.20 0.63 16 9

2.0 24-Dec-16 0.45 0.22 9.40 0.09 0.44 1.10 1.62 109.35 0.18 0.63 4

118

7.5 19-Jan-17 0.47 0.21 1.33 0.09 0.41 0.85 1.56 234.89 0.23 0.63 9

7.0 28-Feb-17 0.46 0.18 1.25 0.09 0.33 0.77 1.55 216.01 0.21 0.63 2

0.3 28-Apr-17 0.42 0.11 0.39 0.09 0.28 0.74 1.44 400.28 0.15 0.59 0

Al Cr Mn Fe Co Ni Cu Rb Sr Pb U

27-Aug- 39. 20.0 2.57 0.85 0.17 4.99 2.37 1.43 164.59 0.64 0.68 16 26 9

30-Aug- 34. 14.2 2.71 0.59 0.15 3.91 1.50 0.78 106.71 0.53 0.45 16 66 3

22. 12.1 6-Sep-16 3.08 0.49 0.16 4.19 1.26 0.87 138.23 0.49 0.56 60 8

19. 22-Sep-16 1.82 0.53 7.54 0.15 2.00 2.25 0.86 125.09 0.66 0.53 07

20. VIC 31-Oct-16 1.71 1.74 9.14 0.15 2.12 1.83 0.83 118.53 0.41 0.54 77

28-Nov- 1.9 1.03 0.19 3.52 0.14 0.35 0.46 0.74 101.00 0.14 0.46 16 1

2.3 24-Dec-16 1.14 0.22 7.93 0.13 0.28 0.39 0.70 108.23 0.14 0.49 5

2.0 19-Jan-17 0.98 0.95 1.11 0.14 0.25 0.30 0.69 101.75 0.14 0.48 4

1.4 28-Feb-17 1.01 0.28 1.74 0.13 0.24 0.24 0.66 105.72 0.13 0.49 7

Al Cr Mn Fe Co Ni Cu Rb Sr Pb U

2.0 27-Sep-16 0.33 0.30 1.15 0.14 0.30 0.43 0.93 72.36 0.18 0.24 2

8.9 3-Oct-16 0.46 0.22 2.89 0.14 0.67 1.31 0.98 71.21 0.21 0.24 8 NER 8.4 18-Oct-16 0.43 0.21 2.57 0.14 0.56 1.19 0.87 69.21 0.21 0.23 9

6.9 26-Oct-16 0.40 0.17 1.79 0.12 0.45 1.00 0.88 69.71 0.18 0.22 9

119

7.0 28-Oct-16 0.38 0.18 2.06 0.12 0.50 1.09 0.88 67.92 0.20 0.22 2

14. 2-Nov-16 0.59 0.35 6.75 0.12 1.32 1.65 0.96 71.83 0.33 0.22 66

7.2 4-Nov-16 0.55 0.23 5.99 0.10 0.51 0.42 0.82 0.13 0.20 2

10-Nov- 2.1 0.27 0.12 1.80 0.10 0.20 0.62 0.83 0.11 0.20 16 4

15-Nov- 11. 0.33 0.28 4.03 0.45 1.35 1.09 0.95 67.89 0.30 0.25 16 07

23-Nov- 3.1 0.31 0.22 2.56 0.14 0.31 0.41 0.85 60.12 0.16 0.22 16 3

2.2 6-Dec-16 0.30 0.18 1.92 0.14 0.28 0.33 0.82 58.97 0.15 0.22 2

1.2 9-Jan-17 0.29 0.17 0.57 0.13 0.20 0.26 0.81 57.52 0.14 0.21 4

5.9 11-Jan-17 0.30 0.24 5.36 0.13 0.28 0.44 0.81 68.49 0.17 0.17 5

1.7 1-Feb-17 0.30 0.17 1.44 0.13 0.21 0.31 0.77 61.27 0.14 0.21 5

2.4 1-Mar-17 0.30 0.18 1.77 0.13 0.19 0.25 0.83 60.45 0.15 0.22 4

Table 4a. NER Pearson correlation matrix. Values >0.5, in bold, are potentially correlated; values >0.9 are strongly correlated.

Co U Precip Fe Mn Sr Cr Rb Cu Ni Al Pb . Co 1 0.408 0.036 0.216 0.299 0.090 -0.136 0.342 0.175 0.614 0.341 0.520 U 0.408 1 0.186 -0.250 0.202 0.219 0.307 0.649 0.392 0.436 0.243 0.379 Precip 0.036 0.186 1 0.172 -0.018 0.413 0.437 0.356 0.567 0.300 0.500 0.348 . Fe 0.216 -0.250 0.172 1 0.706 0.481 0.579 0.397 0.576 0.695 0.785 0.749 Mn 0.299 0.202 -0.018 0.706 1 0.620 0.544 0.667 0.491 0.705 0.613 0.766 Sr 0.090 0.219 0.413 0.481 0.620 1 0.661 0.751 0.739 0.536 0.694 0.633 Cr -0.136 0.307 0.437 0.579 0.544 0.661 1 0.696 0.911 0.676 0.834 0.735 Rb 0.342 0.649 0.356 0.397 0.667 0.751 0.696 1 0.777 0.752 0.725 0.776

120

Cu 0.175 0.392 0.567 0.576 0.491 0.739 0.911 0.777 1 0.811 0.941 0.850 Ni 0.614 0.436 0.300 0.695 0.705 0.536 0.676 0.752 0.811 1 0.908 0.983 Al 0.341 0.243 0.500 0.785 0.613 0.694 0.834 0.725 0.941 0.908 1 0.938 Pb 0.520 0.379 0.348 0.749 0.766 0.633 0.735 0.776 0.850 0.983 0.938 1

Table 4b. VIC Pearson correlation matrix. Values >0.5, in bold, are potentially correlated; values >0.9 are strongly correlated.

Precip. Mn U Cr Rb Cu Pb Sr Fe Al Ni Co Precip. 1 0.258 -0.085 -0.097 -0.136 0.090 0.007 -0.109 -0.301 -0.123 -0.105 -0.090 Mn 0.258 1 0.328 0.150 0.250 0.460 0.287 0.219 0.207 0.361 0.252 0.422 U -0.085 0.328 1 0.518 0.926 0.685 0.621 0.972 0.703 0.619 0.672 0.812 Cr -0.097 0.150 0.518 1 0.572 0.677 0.813 0.681 0.846 0.882 0.962 0.836 Rb -0.136 0.250 0.926 0.572 1 0.739 0.686 0.930 0.820 0.754 0.757 0.890 Cu 0.090 0.460 0.685 0.677 0.739 1 0.955 0.742 0.757 0.859 0.774 0.830 Pb 0.007 0.287 0.621 0.813 0.686 0.955 1 0.734 0.781 0.895 0.855 0.833 Sr -0.109 0.219 0.972 0.681 0.930 0.742 0.734 1 0.800 0.724 0.797 0.872 Fe -0.301 0.207 0.703 0.846 0.820 0.757 0.781 0.800 1 0.926 0.925 0.921 Al -0.123 0.361 0.619 0.882 0.754 0.859 0.895 0.724 0.926 1 0.958 0.946 Ni -0.105 0.252 0.672 0.962 0.757 0.774 0.855 0.797 0.925 0.958 1 0.950 Co -0.090 0.422 0.812 0.836 0.890 0.830 0.833 0.872 0.921 0.946 0.950 1

Table 4c. PES Pearson correlation matrix. Values >0.5, in bold, are potentially correlated; values >0.9 are strongly correlated.

Precip. U Sr Rb Fe Cu Al Mn Cr Pb Co Ni Precip. 1 -0.434 -0.199 -0.488 -0.775 -0.369 -0.324 -0.451 -0.612 -0.361 -0.580 -0.587 U -0.434 1 0.721 0.903 0.324 0.398 0.463 0.386 0.715 0.561 0.558 0.653 Sr -0.199 0.721 1 0.642 0.344 0.647 0.663 0.551 0.740 0.742 0.696 0.734 Rb -0.488 0.903 0.642 1 0.557 0.635 0.567 0.572 0.674 0.703 0.649 0.683 Fe -0.775 0.324 0.344 0.557 1 0.820 0.713 0.833 0.646 0.734 0.764 0.722 Cu -0.369 0.398 0.647 0.635 0.820 1 0.844 0.889 0.645 0.914 0.783 0.747 Al -0.324 0.463 0.663 0.567 0.713 0.844 1 0.955 0.824 0.953 0.896 0.893 Mn -0.451 0.386 0.551 0.572 0.833 0.889 0.955 1 0.783 0.957 0.913 0.867

121

Cr -0.612 0.715 0.740 0.674 0.646 0.645 0.824 0.783 1 0.822 0.932 0.983 Pb -0.361 0.561 0.742 0.703 0.734 0.914 0.953 0.957 0.822 1 0.916 0.891 Co -0.580 0.558 0.696 0.649 0.764 0.783 0.896 0.913 0.932 0.916 1 0.974 Ni -0.587 0.653 0.734 0.683 0.722 0.747 0.893 0.867 0.983 0.891 0.974 1

Table 4d. SUS Pearson correlation matrix. Values >0.5, in bold, are potentially correlated; values >0.9 are strongly correlated.

Al Pb Mn Rb Ni U Co Cr Fe Cu Sr Precip. Al 1 0.989 0.975 0.935 0.937 0.936 0.915 0.886 0.800 0.792 0.224 0.031 Pb 0.989 1 0.982 0.902 0.920 0.888 0.887 0.865 0.735 0.826 0.173 -0.021 Mn 0.975 0.982 1 0.920 0.892 0.871 0.901 0.841 0.771 0.806 0.221 -0.081 Rb 0.935 0.902 0.920 1 0.824 0.945 0.864 0.774 0.803 0.756 0.457 0.054 Ni 0.937 0.920 0.892 0.824 1 0.852 0.933 0.987 0.837 0.562 0.201 -0.054 U 0.936 0.888 0.871 0.945 0.852 1 0.845 0.790 0.801 0.747 0.267 0.221 Co 0.915 0.887 0.901 0.864 0.933 0.845 1 0.904 0.797 0.560 0.363 0.106 Cr 0.886 0.865 0.841 0.774 0.987 0.790 0.904 1 0.844 0.443 0.235 -0.142 Fe 0.800 0.735 0.771 0.803 0.837 0.801 0.797 0.844 1 0.406 0.301 -0.208 Cu 0.792 0.826 0.806 0.756 0.562 0.747 0.560 0.443 0.406 1 -0.025 0.131 Sr 0.224 0.173 0.221 0.457 0.201 0.267 0.363 0.235 0.301 -0.025 1 0.086 Precip. 0.031 -0.021 -0.081 0.054 -0.054 0.221 0.106 -0.142 -0.208 0.131 0.086 1

122

CHAPTER 3

Lakes as paleoseismic records in a seismically-active, low-relief area (Rieti Basin, central Italy) Abstract

Lakes Lungo and Ripasottile (LUN and RIP) are shallow lakes located in the tectonically active Rieti Basin within the central Apennines where large-magnitude earthquakes regularly occur. Sediment cores from these lakes were the subject of a paleoseismic investigation covering the past ~1000 years. Small basin lakes are atypical candidates for these studies, yet may contain seismically induced event layers (seismites) that were generated through strong ground shaking, sediment transport, hydrological reorganization and transient or persistent changes in groundwater chemistry and flow.

The physical and chemical core features formed by these processes, were identified by x- ray fluorescence (XRF) core-scanning, X-ray diffraction (XRD), smear slide analysis,

13 stable isotopic analysis (δ C of organic matter) and bulk geochemical analysis (organic

matter percent, calcium carbonate percent, organic carbon percent, total nitrogen, total

sulfur). Four event layers were identified, occurring in both lakes, and corresponding to

major documented earthquakes in 1298, 1349, 1639, and 1703 CE. The common physical

structure was a homogenous bed (homogenite) of resuspended sediment consisting of a

denser base with a concentration of high magnetic susceptibility (MS) and siliciclastic

grains. Organic matter was concentrated towards the top of these beds. Chemical

123

signatures, including peaks in sulfur, strontium, and barium may represent abrupt shifts to

a groundwater-dominated system, transient groundwater chemistry changes

communicated to the lakes, or changes in groundwater flow and/or spring discharge.

13 Excursions in δ Corg may represent disruptions or changes in carbon source. Not all

event layers had the same features, a result attributed to heterogeneities in epicentral

location, variable lake extent through time, and anthropogenic modification. These results

are interpreted within the context of changes in water column oxygenation, productivity,

and historical records of land use.

Introduction

1. Background

Lake sediment records are receiving increasing interest as archives of paleoseismic

activity. These records are capable of recording cyclicity of past events and aiding in the

development of earthquake recurrence intervals. In regions like the Central Apennines,

Italy, where destructive earthquakes with magnitude > 6 have occurred with high

frequency, determining the nature and extent of earthquake effects in the past may help mitigate future potential consequences. Italy has an advantage over many other

seismically active countries in its extensive and complete written records of past

earthquakes and their effects. The Rieti Basin has been a focus of studies on land use and

hydrological history (Calderoni et al. 1992, Leggio 1995, Mensing et al. 2015) as well as

124

paleoseismicity for the late Pleistocene-Holocene (Michetti et al. 1995). Recent major seismic events of magnitude >6 within 30 km of the Basin (i.e. August, October 2016,

Chiaraluce et al. 2017b.) renewed the necessity to supplement the records of paleoseismicity using additional proxy records, such as lake sediment cores.

Most of the research on lakes as paleoseismic records has focused on deep lakes in high-relief regions that are susceptible to slope failures (Monecke at al. 2004, Hubert-

Ferrari et al. 2012, Avᶊar et al. 2016). “Seismites” are classically considered sedimentary

features attributed to past earthquakes and are formed by mass wasting deposits such as

slumps, turbidites, or landslides (Monecke et al. 2004, Schnellman et al 2002, Wilhelm et

al. 2016). Seismic imaging and grain size analysis of sediment cores are common tools to

identify these deposits (Monecke et al. 2004; Daele et al., 2015). Fewer studies have

explored the potential for small, shallow lakes in areas without sufficient relief to

generate seismically-driven turbidites. The lake records in low-relief areas, however, may provide paleoseismic information in subtler ways and require development of a more expansive set of lake proxies. A paleoseismic investigation has previously been carried out in the Rieti Basin, though the study focused on hazard assessment and earthquakes prior to the historical period (Michetti et al. 1995). Their approach, seismic trenching and fault scarp interpretation, proved difficult at the edges of the Rieti Basin where the major faults are overlain by thick colluvial sequences and vegetative cover. One limitation was that the ages of the events identified by Michetti et al., (1995) were only broadly

125

constrained by using radiocarbon to date the colluvium, which provides only a maximum

age, and one that may have been influenced by old carbon contamination from the

underlying carbonate bedrock, a pervasive problem with radiocarbon date in the Rieti

Basin sediments (Mensing et al. 2015).

The physical effects of ground shaking can have a strong influence on low-relief lake basins. Co-seismic movement can change lake inflow and outflow (especially drainages), as well as disrupt sedimentation (Carrillo et al, 2008; Avşar et al. 2014; Daele et al. 2015). Studies have shown that strong earthquakes can produce ground motions that cause lake water oscillations, or seiche waves (Doig, 1998; Avesar et al. 2014). This effect is more pronounced in lakes with a high sediment accumulation rate and shallow, narrow basins (Avşar et al. 2014), two properties that are currently common to the lakes in the Rieti Basin. The oscillations re-suspend unconsolidated surface sediment composed of a mixture of clastic and organic material that is redeposited in a density gradient, creating a unit with higher density clastic or coarse-grained material at the base and low-density organic-rich sediment concentrated at the top. Water oscillations can also increase sediment focusing, or littoral sediment transport to the depocenter (Avşar et al.

2014).

When lakes receive inputs from multiple hydrochemically distinct sources, there is a potential for seismicity to alter the relative contributions of these inputs. The

126

chemical constituents of groundwater contributes influx through lakebed seepage and/or

peripheral spring discharge (Brunner et al. 2009). This component of a lake’s

hydrochemical budget is often neglected because of difficulties in measurement or

quantification (Rosen 2015; Rosenberry et al. 2015). The connection between aquifers

and movement along faults that crosscut them includes physical changes (flow, spring

discharge, pore pressure) and chemical changes (Table 1) that have been observed in the

central Apennines within similar faulted karstic aquifers (Quatrocchi 1999, Falcone et al.

2012). The lakes in the Rieti Basin have historically had significant inputs of

groundwater and surface water from a high sulfate karstic aquifer, low sulfate shallow

alluvial aquifer, and low sulfate river water (Archer et al. 2016) and thus provide a good

test case for recognizing seismically driven chemical changes caused by changing water

source. If the seismological effects on lake hydrochemistry are persistent and large

enough, it may in turn be exported to the sediment record. This paleo-hydrochemical

approach may provide a new means of detecting paleoseismicity, especially if other

processes can be ruled out as potential drivers.

The study of paleoseismic signatures in the Rieti Basin was instigated by two

factors. First, the recent central Italy seismic sequence (August 24 2016-January 18,

2017) had epicenters located 40-60 km from the Rieti basin and prompted the necessity and curiosity to investigate past events. The central Italy seismic sequence resulted in some transient hydrochemical effects in nearby springs (Chapter 2) as well as strong

127

ground shaking in the Rieti Basin (USGS 2016, INGV shake map or Paolo Belezza, personal communication). Second, there is an abrupt and sedimentological and geochemical change in both lake records that cannot be explained satisfactorily by other processes, and appears to be coeval with a major central Apennine seismic sequence in

1349 CE documented by Galli and Nasso (2009).

The chemical proxies used to identify the regional groundwater influence on these lakes are sulfur (S), calcium Ca), barium (Ba), and strontium (Sr). These elements have complex biogeochemistry in lakes, so possible transformations in the lake water column and sediment pore water must also be considered. In lake sediments, sulfur is derived from particulate S inputs from the catchment, dissolved sulfate from the water column, and/or S incorporated in settling organic matter (Cohen 2003, Couture et al. 2016). Once deposited, sulfur in the form of sulfate may be utilized by bacterial sulfate reduction

(BSR) when other potential electron acceptors become depleted (oxygen, nitrate, Mn, Fe) then hydrogen sulfide either diffuses to the water column or binds to sedimentary constituents (Nealson and Myers 1992). When sufficient ferrous iron and sulfide are present under anoxic conditions, iron monosulfides precipitate, followed by the more stable form, pyrite (Holmer and Storkholm 2001; Russell and Werne, 2009). Though reaction with reduced iron is kinetically favored, organic forms may predominate when lake sulfate concentrations are low (<5mg/L), while in lakes with higher sulfate concentrations, retention of sulfur in sediment as non-organic, primarily metal-sulfides,

128

predominates (Nriagu and Soon, 1985). Strontium is a metal that becomes enriched in

groundwater in this region following an earthquake (Chapter 2). Its concentration in

water and sediment in central Italy correlates well with calcium and barium because of

their incorporation into prevalent limestone massifs (De Vos et al. 2006). Strontium also

behaves much like Mg geochemically and can be used as a proxy for Mg changes

downcore in elemental data when Mg is not measured. (Rush 2010).

The goal of this study is to pick sedimentary and/or geochemical event layers with

distinct physical and/or chemical characteristics in dated sediment cores from two

adjacent lakes, compare these paleo-earthquake ‘picks” with a compiled catalog of past major earthquakes within a 50-km radius from the lakes under study that have shown evidence of regional effects, then examine both seismic and alternative non-seismic mechanisms that could have possibly created the identified event layers. Comparing lake sediment core records to historical records provides an opportunity to test the utility of lakes in seismically active areas as verification or documentation of past earthquakes.

129

Table 1. Earthquake signals in lake records—an overview The signals in lake sediments that studies have resolved as seismically-induced are outlined in Table 1, organized by their mechanisms (physical, chemical). SIGNAL/ SEISMITE MECHANISMS REFERENCE TYPE Physical A. Mass-wasting, Co-seismic weakening of Hubert-Ferrari Signals landslide or turbidite slope stability of steep et al. 2012, deposits: sides within a lake or the Brooks 2001; Slanted or curved bedding slope stability of Bertrand et al. structure, no grain-size landforms in the 2008; Fanetti et sorting or MS signature. surrounding watershed al. 2008; Daele May be chaotic in form. et al. 2015 B. Homogenites: Erosive Strong shaking- induced Chapron et al. or disturbed base, with a water oscillations, seiche 1999; Leroy density-sorted structure waves, or slope-failure 2002; Bertrand including a siliciclastic- cause upper sediment et al. 2008; enriched base and an resuspension. Re-settling Carrillo et al. organic-rich or lower occurs in a density 2008; Hubert- density upper. May gradient, though visually Ferrari et al. include macroscopic appears homogenous. 2012; Avᶊar et organic debris or littoral May cause enhanced al. 2014; macrophytes/ carbonate sediment focusing with Valero-Garcés more littoral debris at et al. 2014 depocenter. C. Increased sediment Decreased slope Holbrook et al. yield: More clastic and/or cohesion in the 2006; Hubert- riverine input, higher watershed yields Ferrari et al. sediment accumulation increased sediment & 2012; Avᶊar et rate, or increased soil soil delivery to rivers and al. 2014b; influx. “Muddy river streams. This decreased Wang et al. effect.” The opposite cohesion may also 2015; Daele et signal could be due to weaken river banks, al. 2015 river avulsion. causing changes in flow. River avulsion also may be caused by landslides in the channel.

130

D. Soft sediment Brittle or ductile Matsuda 2000; deformation: micro-faults, deformation of sediment Avᶊar et al., load casts, layers occurring due to 2016 liquefaction/fluidization, ground shaking. flame structures Chemical E. Transient increases in Fracture clearing in the Quattrocchi Signals concentrations of certain aquifer and/or “flushing” 1999; Amoruso chemical species (weeks- water and dissolved et al. 2011; months). constituents out of pore Falcone et al spaces, slower flow paths 2012; and/or karstic reservoirs. Voltattorni and Caused by co-seismic Quattrocchi increases in hydraulic 2012 conductivity. The groundwater carrying this signal is then contributed to the lake water, either by seepage, littoral springs or surface flow. F. Persistent increases in Enhanced permeability Carro et al. dissolved concentration of by seismic strain or 2005; Manga chemical species altered poroelasticity and Wang, associated with regional increases hydraulic 2015; Michetti groundwater (months- conductivity. This can et al. 2007; decades) cause higher water tables Jang et al. and/ or spring discharge, 2008; Falcone and increase et al. 2012 groundwater flow rate. In alluvial aquifers, co- seismic consolidation of soil particles and lateral dilation causes increased transmissivity and more horizontal groundwater flow, increasing groundwater contribution to the lakes

131

G. Anomalies in Eh, Hydrothermal fluid Quattrocchi sulfur gas emissions, escapes, or new available 1999; Leroy et decreased pH, increased conduit for hydrothermal al. 2002 222 dissolved Rn and CO2 fluids and volcanic gasses from deep seated sources (“seismic pumping”) that migrate upwards along faults. Increased CO2 and/or thermal water input to groundwater also could increase dissolution along flow paths, though on a longer time scales. H. New spring discharge Permanent aquifer Leroy at al. location, or cessation of deformation, formation 2002; Claesson existing spring. Lasting of new fractures, fracture et al. 2004; change in groundwater closing, or breaching of Manga and chemistry. aquitards. This may Wang 2015 cause mixing of different reservoirs previously isolated.

2. Tectonic and hydrogeological setting

The central Apennines of Italy were formed as an accretionary wedge during subduction in the Mesozoic. The region is now undergoing post-orogenic extension driven by the convergence of the Eurasian plate with the African plate and subsequent back-arc spreading and opening of the Tyrrhenian basin to the west. This spreading is accommodated by N-NW trending Quaternary normal faults (Galli and Nasso, 2009,

Chiaraluce et al. 2017). The study location, the Rieti Basin, is an extensional depression in the central Apennines (Figure 1a.) surrounded by Mesozoic carbonate platform,

132

Triassic-Miocene pelagic units and Miocene foredeep sediments (Roberts and Michetti,

2004). The basin has subsequently filled with 0.5-2 km of Quaternary fluvio-lacustrine

deposits (Cavinato and De Celles, 1999). The lithology of the surrounding faulted area

plays a role in the strength and location of seismic events, as shallow marine sedimentary

units are prone to deform seismically (Chiaraluce et al. 2017). Earthquakes are frequent

and shallow (<12 km below land surface) relative to other plate convergence zones

(Carannante et al. 2013), and typically occur as multiple-shock sequences. During August

2016- January 2017, a seismic sequence struck this region north of Amatrice, activating a

normal fault system from about 6-12 km in depth. Three mainshocks had magnitudes

greater than 6.0, and thousands of aftershocks (Mw >2.0) were recorded. This event

activated a fault segment in between the previous two large seismic sequences of 2009

near L’Aquila to the southeast and 1997 near the Colforito Plain to the northwest

(Chiaraluce et al. 2017). The mainshocks of the 2016-2017 seismic sequences occurred

with reported intensity of 6-7, while the 2009 mainshock outside of L’Aquila was felt

with an intensity of 5 in the Rieti Basin (USGS 2016).

The lakes of the Rieti Basin intersect the groundwater table and receive surface

inputs from groundwater springs in their current hydrological configuration (Archer et al.

2016). In the modern configuration, Lago Lungo (LUN) and Lago di Ripasottile (RIP),

located ~ 2 km from the basin-edge faults, have dissimilar chemistry owing to major

contributions by different aquifers. LUN is fed by groundwater from the shallow aquifer

133

within basin-fill alluvium that is more sensitive to seasonal changes in chemistry and

water level (Archer et al. 2016). The alluvial aquifer discharges into both LUN and an

active spring, Vicenna Riara, located just to the south of LUN. RIP receives input from

the spring on the NE edge of the Basin, Santa Susanna (SUS). This spring represents the

baseflow of a large recharge-area carbonate aquifer spanning over 300 km2. The aquifer

is contained in the Umbria-Sabine succession of pelagic deposits embedded with

Triassic-Miocene evaporite deposits. Upwelling of water from areas of intense gypsum-

anhydrite evaporite dissolution is responsible for the high sulfate (5-20 fold greater than

nearby alluvial springs) of SUS water (Spadoni et al. 2010, Boschetti et al. 2011). The

contribution of this spring to RIP in the modern day is unmistakable by the high sulfate ion concentration and similar stable isotopic values (Archer et al. 2016). This chemistry represents mixing with the deep, mineralized flow circuit where groundwater exhibits higher salinity, higher pCO2, and greater concentration of elements associated with

carbonite and evaporitic rocks, such as strontium and barium (Morgantini et al. 2009).

These extent and water levels of these lakes have fluctuated over the past two millennia

and were likely joined into one larger lake in the past. Reclamation efforts during the 17th

and 18th centuries drained excess surface water from the basin and permanently separated

the two basins. The lake level has been maintained at 369-370 m above sea level (a. s. l.)

since 1940 by pumping excess water to the Velino River, but at the lakes’ greatest extent

134

within the historical period the level was at 375 m a.s.l. (Figure 1b.; Camerieri and

Mattioli 2014).

The aquifers of the carbonate Apennines are hosted by the ridges of Mesozoic-

Cenozoic limestone and dolostone and have large recharge areas (>300 km2).

Groundwater circulates through preferential flow paths within fracture networks and karstic structures (Petitta et al. 2011). Major springs are located along the basin border faults or the permeability divides at the contact with recent fill deposits and are characterized by large, steady discharge (0.5-18 m3/s). Faults within these aquifers act as

low-permeability divides or conduits for upwelling of mineralized groundwater from a

deeper-flow circuit (Petitta et al. 2011, Falcone et al. 2012). SUS discharges along the

range-front fault that separates the plain from the higher-relief carbonate mountains to the northeast (Spadoni et al. 2010). Upstream in the Velino River Valley, several other major springs discharge at the contact between the carbonate ridges and the alluvial deposits of the valley. Thermal and/or mineralized springs comprise some of these, sourcing the deep flow circuit and characterized by high concentrations of dissolved ions and dissolved CO2

(Petitta et al. 2011).

3 Methods

3.1 Core recovery, correlation, and visual analysis

135

Sediment cores were recovered from Lago Lungo in 2009 and 2012 and from Lago di

Ripasottile in 2012 and 2013 using a hand operated square-rod Livingstone and a modified Livingstone (Bolivia) corer from a floating platform anchored in the deepest part of each lake (See Mensing et al 2015, for more detail regarding coring procedure).

Surface sediments were recovered with a plastic tube core specially fitted with a piston to produce minimal disturbance of the sediment-water interface. The unconsolidated surface sediments were stabilized with Zorbitrol (sodium polyacrylate absorbent powder) while the core was still in an upright position. Three overlapping cores were analyzed as the composite core from Lungo: LUN12-2C, the surface core, LUN09 and LUN12-2B. From

Ripasottile, 4 overlapping cores were used to develop the composite core: RIP1A-1P, the surface core, RIP13-1A, RIP13-1B, and RIP12-1C. The composite core for each lake were built by using the most complete core (the longest continuous core with the fewest gaps in recovery and coring artifacts) then adding correlated core section intervals where material was missing or recovered poorly. These composite cores are hereafter referred to as LUN and RIP.

Smear slides were made at regular intervals of approximately every 10cm of each core and also from select intervals of sedimentological change, then were inspected visually under a microscope with regular and cross-polarized light illumination. X-ray diffraction (XRD) analysis was conducted on discrete intervals from each core. Sub- samples were powdered and scanned from 5° to 65° using a Seimens D-500

136

diffractometer and a 2.2 kW sealed cobalt source. Phase identification was performed

using Match! Software after removal of background and filtering of data. Mineral phases

at select depths are semi-quantitative and are reported herein on a presence/absence basis.

3.1 Age model determination and correlation

The age model for the Lago Lungo core is based on historically documented

biostratigraphic markers and magnetic paleosecular variation (PSV). Refer to Mensing et

al. (2015) for a detailed description of the age model development for the Lago Lungo core. The magnetic trends (ChRM inclination, declination and relative paleointensity) were smoothed and fit to reference curves from PSV models of Europe over the past

3000 years (Figure 2). This correlation provides a high-resolution age model that was further constrained by specific pollen cultigens, or species with a well-documented first appearance of cultivation in the area that can serve as local biostratigraphic datums. The first application of the Ripasottile age model, presented here, is based on correlation between LUN and RIP magnetic parameters and application of the same PSV model,

SCH.DIF.3k of Pavón-Carrasco et al. (2009), to the RIP core.

Rock magnetic and paleomagnetic properties were measured at 1-cm spacing on u- channel samples collected from 3 distinct and partly overlapping cores (RIP12-1B,

RIP13-1A and RIP13-1B). The low-field magnetic susceptibility (k) was measured for each u-channel using a Bartington magnetic susceptibility probe MS2C in-line with the

137

rock magnetometer at the Istituto Nazionale di Geofisica e Vulcanologia (INGV), Rome,

Italy. The MS curves were used to correlate between the three cores and to develop one

continuous composite core using overlapping sections where ends were distorted or

shortened from of the core recovery process. The age model for RIP was completed first

by selecting the dominant tie points between LUN and RIP composite core magnetic

parameters, tying in major peaks and troughs between these tie points, then adjusting the fit to be succinct with the PSV curves and models developed for for Europe (Gallet et al.

2002; Pavón-Carrasco et al. 2009).

The temporal error for the PSV age model for LUN was determined by first calculating the temporal resolution of regional archeomagnetic model SCHA.DIF.3k of

Pavón-Carrasco et al. (2009) at the geographical coordinates of Lago Lungo, and then summing the Probability Density Function (PDF) of the various archeomagnetic parameters at each tie point in the age model. The 95% confidene interval for the summed PDF at each tie point ws then used to determine the temporal error for sections of the core. Since both LUN and RIP are essentially in the same geographic location, the

PDFs and temporal error developed for each LUN tie point (Mensing et al., 2015) were transferred to the RIP core by through correlation to LUN.

3.2 High-resolution XRF analysis

138

Elemental geochemistry analysis of the composite core sections was carried out using

the ITRAX X-ray fluorescence (XRF) scanner (Cox Analytical Instruments) at the

University of Minnesota, Duluth. The LUN12-2 core was scanned at a resolution of 0.5

cm throughout, with an increased resolution (0.2cm) for the middle 7 sections (2B-5L

through 11L) as well as the top section (LUN12-2C). The RIP13 and RIP12 composite

core was scanned at a resolution of 0.5 cm, with one section (RIP13-1B-5L) scanned at a

0.1cm resolution to pick out fine variation in the laminations. The scanner was operated

using a molybdenum source, 30s dwell-time, a voltage of 30 kV and an x-ray current of

30 mA to obtain peak areas for elements Si:Pb. XRF data (raw counts) were normalized by computing the log-ratio (base 10) of individual element counts and dividing by the total measured kilo-counts per second (kcps) to correct for the closed-sum effect or dilution by increases in unmeasured light elements such as organic matter and/or water

(Lowermark et al. 2011, Haenssler et al. 2013). A locally weighted scatterplot smoothing

Steve Juggins) software.

3.3 Stable isotope analysis

The composite cores of both lakes were sampled every ~10 cm for organic carbon 13C analysis. Dry, powdered sediment samples were moistened with DI water then placed in a glass fumigation desiccator containing concentrated HCl for 30 days to remove

139

carbonates, following the method of Yamamuro & Kayanne (1995). The concentration of organic C, total N and organic carbon-δ13C were then determined using a Eurovector

EA 3000 elemental analyzer interfaced to a Micromass IsoPrime stable isotope ratio mass spectrometer equipped with a helium diluter using the method of Werner et al. (1999).

Error was determined by running duplicates of samples and acetanilide standards. The standard deviation for each duplicated sample was calculated, and the average isotopic and elemental concentrations and corresponding standard deviations for the 12 acetanilide standards are: (δ13C) -33.66±0.02, (weight % C) 71.09±0.96, and (weight %

N) 10.36±0.15. All isotopic measurements are presented in the delta (δ) notation:

13 12 13 12 δa = [Ra/Rstd]*1000 , where Ra is the C/ C ratio in the sample and Rstd is the C/ C ratio in the standard. All values are expressed as parts per mil (‰).

3.4 Bulk Elemental Analysis

Organic matter percent (OM%) was determined through loss on ignition after the method of Dean, (1974). The cores were sampled at 10 cm intervals, dried at 100°C for

24h, weighed and combusted at 550°C. The weight after combustion at 550°C was divided by the dry sediment weight to calculate total percent of organic matter. Then, after the method of Heiri et al. (2001) the same samples were re-combusted at 1000°C and reweighed, and the difference in weights was used to calculate CaCO3%.

140

Weight percent sulfur (weight %S) was determined at UNR using an ELTRA CS-800 with induction furnace. Strong correlation between weight percent S and XRF counts of S allowed for calculation of a transfer function between discrete weight %S measurements and high-resolution XRF core-scanning S counts. The ratio of organic carbon to total sulfur (Corg:STS) was calculated for each discrete measurement of weight %C.

3.5 Event Layer Identification

For this investigation, the high-resolution photos taken immediately following core splitting were examined for signals listed in Table 1, except disturbed laminae. Disrupted or disturbed laminations are a common feature of the LUN and RIP cores due to the nature of the coring process, so there is no sure way to discern between seismic laminae disturbance and coring artifacts. This inspection combined with assessing the chemical trends was used for picking initial candidates for event layers. The first criterion for picking event layers was selection of atypical beds or discontinuities that deviated from background sedimentological and mineralogical characteristics. The next criterion was the presence of a signal from Table 1, particularly homogenites (Table 1-B) and chemical irregularities. The third criterion is that corresponding event layers should be present in both lakes to be attributed to a regional trigger, even if the features of these event layers differ slightly between lakes (Xoes et al. 2010; Daele et al. 2015). After the event layer candidates were picked in each core, the age model and age model errors were used to

141

determine how these events compared to historical seismic events. After comparing with

the age model and historical events, one additional iteration of event layer selection took

place to check the position of the event “picks”.

3.6 Historical earthquake chronology and GIS simulation

Earthquake historical records, along with on-fault trenching and paleoarchaeological

studies, compiled into databases of past events (Valensise et al. 2003, Galli et al. 2011,

Rovida et al. 2016), served as the data resources for identifying relevant paleoseismic events. The database CPTI15-DBMI15 (Parametric Catalogue of Italian Earthquakes-

Italian Macroseismic Database 2016) was the primary resource as it includes information from a wide range of sources. These events were then assessed in terms of to the selected event layers in the cores to evaluate potential synchronicity.

ArcGIS was used to create a basemap of the study area. The digital elevation model

(DEM) used was the TINITALY 1 spliced together. The spatial analyst toolset and cut fill tool was used area surrounding the two lakes that would be inundated if water was 1 m, 2 m, 3 m, or 5 meters higher than present. Figure 1b shows a shaded area of the lake’s greatest extent within the study period, 375 meters a.s.l. (Camerieri and Mattioli 2014).

142

4 Results

4.1 Historical Earthquakes

A list of historical earthquakes with magnitudes of 6.0 or greater for the past 1000 years was compiled (Table 2) and their epicenter plotted (Figure 1). One event with a smaller magnitude, Mw 5.78, was included because of its proximity (~5 km) to the basin. Each of

these events is evaluated in the context of the lake core records from Lungo and

Ripasottile.

Table 2. Earthquakes by year, magnitude and location Year Magnitude Epicenter Epicenter Intensity Distance Source Lat./Lon. (MMI) to Rieti (Km) 1298 6.3 Reatino 42.575 9-10 5 CPTI15- 12.902 DBMI15; CFTI 4 MED 1349 6.3 Flamignano 42.270 9-10 20 CPTI15- (Aquilano) 13.118 DBMI15; Galli and Naso 2009, Cello 1997 1639 6.2 Amatrice 42.639 9-10 30 CPTI15- (Monti 13.261 DBMI15; della Laga) 1703 6.8 Appennino 42.7 11 20 INGV Reatino (2 13.067, DBMI04, loc.) 42.617 CFTI 4 13.1 MED

143

1703 6.7 Aquilano 42.234 10 40 CPTI15- 13.292 DBMI15; INGV DBMI04, CFTI 4 MED 1785 5.8 Piedilucco 42.536 8-9 5 CPTI15- 12.788 DBMI15; INGV DBMI04 CFTI 4 MED 2009 6.3 E of 42.309 9-10 60 CPTI15- L’Aquila 13.510 DBMI15; USGS 2016

4.2 Core chronology

An age model for LUN was constructed using the combination of archaeomagnetic data and pollen cultigens. Unfortunately, due to “dead” carbon problems, radiocarbon dates were not useful. See Mensing et al. (2015) for a detailed discussion of the age model for

LUN. The error associated with the LUN age model is in two forms. First, the PSV age tie points have a minimum and maximum age that is based on the 95 percent confidence interval (2σ) of the combined magnetic parameters (See table 1 in Mensing et al. 2015).

This interval was not constant throughout the core, but increased with increasing core depth. The error associated with the biostratigraphic markers used in the age model was a range in dates for the particular cultigen based on historical records. The two cultigens used, Zea Mays and Cannabis, historically appeared in the Rieti Basin only after a certain

144

date that was tied to the first presence of these pollen types in the core. These first

appearance ages are reported as ranges, so these were assigned to the error range of the

cultigen tie points. Cultigen tie point age error ranges were used in the upper portion of

the core (post-1660 CE) and PSV error ranges were applied below.

The RIP age model is presented herein for the first time and is discussed in more detail.

The correlation between magnetic parameters in LUN and RIP cores (Figure 2) shows that the RIP core (1216cm) represents ~1020 years of sedimentation. This is less than

LUN, representing ~2700 years in 1438cm. The sedimentation accumulation rate (SAR), calculated based on this age correlation, has similar values in both LUN and RIP during the upper ~400 cm, ranging from 0.4 - 1.0 cm/year (Figure 4a. and b.). During the laminated interval, SAR increased in LUN to 1.2 cm/year and in RIP varies between 1.5 and 3.0 cm/year. The age of the base of the RIP core is at ~975 CE and LUN is -695

BCE.

4.3 Bulk core sedimentology

The predominant lithology in RIP is a laminated facies composed of laminated silt, marl, and clay, with lamination thickness ranging from 1 mm to 20 mm, and varying in color from gray, maroon, dark gray, light gray, and white. The white bands, photographed when the core was freshly opened (Figure 3) oxidized to orange after a year in core storage. XRF analysis shows frequent peaks of Mn, Fe, and Ca,

145

corresponding to the color banding, and close inspection of some of the lighter gray

layers show grain size variation from silt to clay, and rare fine sand layers. Smear slide

and XRD analysis of the laminated facies shows the major components are quartz,

calcite, and clay minerals. Intermittent calcite concretionary layers were observed in the

laminated facies, but they appear discontinuous and not correlative between core pairs in

the same lakes and are not deemed relevant to the paleoseismic interpretations. In RIP,

the laminated facies comprises the entire core from the base until 410cm, above which,

there is an abrupt change to a homogenous or discontinuously banded clay mottled with

black grains of iron sulfides. Lamination resumes intermittently above this interval, but

with fewer multicolored bands and some thicker more homogenous intervals (Figure 3).

The lithology in LUN is slightly more varied than RIP in that the base of the core

is composed of a silty streaked black marly facies (lower core sections not shown), the

middle of the core from 750 - 340 cm is composed of the laminated facies predominant in

RIP (Figure 2). At 340cm there is an abrupt change in LUN to a dark sulfidic clay,

correlative to the change in RIP at 410 cm. The other prominent feature that allows for

stratigraphic correlation occurs at 150 cm in LUN and 220 cm in RIP and consists of a

bed with a disturbed base and several centimeters of iron sulfide- rich sedimentation. The seismic event picks are annotated in Figure 3, and their salient features are listed in Table

4.4 Elemental geochemistry

146

Selected elements and chemical proxies are displayed in Figure 4. The continuous

data, including normalized XRF data, magnetic susceptibility, sediment accumulation

rate (SAR) and schematized core lithology are plotted with discrete downcore data including organic matter percent (OM%) and calcium carbonate percent (CaCO3%), both

13 measured by LOI, δ Corg, and the weight ratio of Corg/Ntotal. Data gaps in the continuous

curves (“No Data”) are due to the contruction of the composite core sequence after core

sections were submitted for XRF analysis and the correlated sections analyzed contain

gaps between sections. The elements displayed were selected based on their response

associated with the paleoseismic event picks. Conservative detrital input, represented by

Ti, is highest during the laminated background sedimentation and has an opposite trend

from Ca and CaCO3% in both lakes. Both Fe and Mn have similar profiles in both lakes

and track with Ti in the laminated facies and then separately above ~400 cm in RIP and

~350 in LUN. The Ca profile in both lakes matches well with CaCO3% determined by

sample combustion, indicating that XRF counts of Ca are not influenced by the presence

of other elements not measured during core scanning (Lowermark et al. 2011). The Sr in

both cores tracks with Ti until ~350cm in LUN and ~400cm in RIP, when it shifts to

behavior similar to Ca and at times sulfur. The profiles of sulfur in both lakes exhibit low

values in background sedimentation with several distinct peaks in the upper portions of

both cores. The sulfur XRF data correlated well with select samples of sulfur weight

percent (see section 3, Methods) so downcore XRF variations were interpreted as

147

changes in total sedimentary S. Weight percent sulfur was extrapolated from point samples based on this high correlation. Figure 4 uses these weight % S in the calculation of Corg/STS ratios. For the majority of both core sequences, Corg/STS values were low, with the major peaks concentrated between 1000-1350 CE and after 1800 CE. The OM% in

LUN is low (<10%) and relatively invariable during background sedimentation. Above event 2L, the OM% peaks, then reaches the highest in the core (17.5%) during the sedimentologically distinct interval at 200-230 cm. Another peak occurs concurrent with event 4L. In RIP, background sediment %OM is also relatively low and invariable

(<10%), and the first peak occurs after event 2R. The highest value of %OM in RIP is in the most recent sediment or core top (16.23%).

Ratios of carbon to nitrogen (C/N) at the base of LUN are variable, at times reaching values over 20. During the banded interval of sedimentation, values are lower and less variable, fluctuating between 7.3 and 8.4. Above, as the ratio again becomes more variable, the highest value (21.4) corresponds to the homogenous grey clay unit overlying event 2L. In RIP, values are generally lower, with a maximum value of 10. The largest variations are at the base and the top of the core, though the range is small (>6,

<10).

4.4.1 Stable Isotope Results

13 Values of δ Corg in RIP (figure 3) range between -26.3 and -25.6‰ at the base of the core, below 400 cm depth (year ~1400 CE). Above this depth, values are more

148

depleted and exhibit a greater range, from -29.2 to -25.4‰. In LUN, in the marly facies at the base of the core, before 900 CE, values were highly variable, ranging from -40.2 to

13 27.0‰. The laminated facies interval in the LUN core has Corg values similar to those in

RIP, ranging between -28.2 and -25.7‰. After 1400 CE in LUN, values again became

more negative and variable, ranging from -38.5 to -26.3‰. Event layers 1L, 2L and 3L

13 are concurrent with depletions in δ Corg. Event layers 1R and 2R are also concurrent with

13 more negative δ Corg, though these negative values are less in LUN than in RIP.

4.4 XRD and smear slide analysis results

XRD analysis of select depths, displayed in Figure 3, found calcite and quartz the

dominant minerals in both LUN and RIP. The uppermost ~400 cm of both cores contains

primarily calcite, while the lower portions contain primarily quartz. Pyrite, gypsum,

brushite, dolomite, and various clay minerals were also present as minor constituents at

various depths in the sediment (Figure 3).

Examination of smear slides from event intervals compared to laminated sedimentation (Figure 3S1) confirms the dominant form of carbonate present throughout the record is calcite, with sparse dolomite. At depths where XRD analysis found pyrite as a minor constituent, pyrite was observed in smear slides as opaque black grains, some framboidal. Depths with similar shaped opaque black grains were assumed to also contain pyrite. Grain size contrasts were evident in successive layers, as was the

149

prevalence of OM and diatoms in some event layers. Diatoms were absent and OM was low in laminated sediment and at the base of some event layers.

150

Table 3: Event layer description and selection - -

s LUN event Earthquake correlation (run Depth Core depth in cm) Characteristics events RIP (run Depth Core depth in cm) Characteristics Type Seismite 1) (From Table 4L 1703 LUN Homogenite; 4R RIP Homogenite; B; E, 140- • Abrupt and/or 175- • Abrupt and/or F or 155 erosive basal 190 erosive basal H contact contact • MS and • MS and siliciclastic siliciclastic element peak at element peak at base with pyrite base with pyrite • grading in OM%; • peak in S- highest at top 13 deposition, low depleted δ Corg at C:S base, graded to more • grading in enriched OM%; highest at top of the unit 13 • depleted δ Corg at base, graded to more enriched 3L 1639 LUN • Abrupt basal 3R RIP • Abrupt basal E 185- contact 222- contact; MS peak and/or 197 • Homogenous 232 • Homogenous F sediment black-mottled • S, OM% peaks; sediment low C:S • S, OM% peaks; • Minor pyrite low C:S

151

2L 1349 LUN Homogenite; 2R RIP Homogenite; B; E, 305- • MS and 370- • MS and siliciclastic F 330 siliciclastic 400 element peak at and/or element peak at base with pyrite H base with pyrite • Structure-less 20cm • S, Sr, and Ca gray-brown bed peaks; low C:S • grading in OM%; • grading in peak at top OM%; highest • S and Sr peaks 13 at top • depleted δ Corg • structure-less • Shell fragments 20cm gray bed throughout with low siliciclastic content 13 • depleted δ Corg and increased C/N • preserved gastropod shells within high OM% layer 1L 1298 LUN Homogenite; 1R RIP Homogenite; B or 385- • MS and 435- • MS and siliciclastic C; F 400 siliciclastic 440 element peak and/or element peak • 5cm gray bed H • Structure-less • S, Fe peaks 13 ~15cm gray • Depleted δ Corg bed • 8‰ depletion 13 in δ Corg, increased C/N

5. Discussion

152

Four event horizons are identified in the last ca. 700 years of Lago Lungo and

Ripasottile are strong candidates for low relief seismites. Of the initial event picks, all but one were confirmed during the iterative round. The position of 3R was adjusted upwards to better position it in light of concurrent reclamation efforts, discussed below. Within the error of our age model, all picks appear to correspond to the age of major earthquakes (>

6 magnitude) close to the Rieti Basin (epicenter within 40 km). These event horizons occur in both lakes, as would be expected if they were generated by a regional signal such as seismicity. Below, the discussion focuses on each earthquake for which physiochemical observations were made, consistent with changes delineated in Table 1, and features associated with each pick are summarized in table 2.

5.1 Event 1, the 1298 earthquake

5.1.1 Earthquake Description

This was a strong event (6.26 Mw) that occurred in 1298 CE with an epicenter

only 5 km to the north of the lakes. It would be expected to cause strong ground motions

and possibly large fault ruptures in the Reatino Mountains on the east of the plain (Figure

1b). Historical records of damage and perceived shaking were converted to intensity

values (Rovida et al. 2016), reporting that the event was felt with an epicentral intensity

of 9.5 throughout the Rieti basin, with higher local intensities of 10 in Poggio Bustone,

the small hillside town 1 km east of LUN and the Santa Susanna spring. Over the past

1000 years, this earthquake appears to have been the closest-distance largest-magnitude

153

event to the Rieti Basin. The alluvial basin-fill sediments are highly susceptible to seismic ground shaking during events with this intensity, especially near the basin-edge faults where the lakes are situated (Tertulliani 1999; Bozzano et al. 2009).

5.1.2 Event 1 Features

The distinguishing sedimentological feature picked as a seismite is a homogenite

(Table 1-B), occurring in both lakes and denoted as 1L and 1R (Table 3; Figure 3).

Homogenites form when a physical trigger causes strong mixing of sediment and pore water, resuspending less dense grains that re-settle at various velocities depending on particle density. This gives the homogenites characteristic partitioning, with the composition of the base and uppermost portions chemically distinct though without visible structure (Hubert-Ferrari et al. 2012). Seismically-induced triggering mechanisms

may include lake-slope mass movements or sediment fluidization from strong water

oscillations (seiche waves) or from large pulses of groundwater input (Valero-Garcés et

al. 2014).

During shaking-induced sediment resuspension at event 1, denser particles

including siliciclastic and ferromagnetic grains (high MS) rapidly settled and

concentrated at 1L and 1R, explaining the elevated Fe, Ti and MS at this event (Figures

13 4a. and 4b.). The depleted δ Corg in 1L relative to the background values (Figure 4a)

may also be a result of enhanced water column mixing and OM recycling (Talbot and

Lærdal 2000; Vreca and Muri 2006) during homogenite formation. During resuspension,

154

decomposed or partially decomposed OM would provide a source of relatively light C to

13 the water column, causing further depletion in δ Corg as the recycled carbon is utilized.

The C:N also increases slightly within 1L, above the range of aquatic algae (C:N<10,

Routh et al. 2004) that prevails in the background laminated facies (Figures 4a. and b.).

This signifies a contribution from macrophytes, possibly those occupying the littoral zone

transported to the depocenter by increased wave action (Hubert-Ferrari et al. 2012). The non-pollen palynomorph Glomus, an indicator of disturbance and/or erosion, also peaks in this interval in LUN (Mensing et al. 2015), supporting increased wave-action and erosion of the littoral zone. An alternative seismic trigger for 1L and 1R that would produce a similar signature is decreased slope cohesion or instability triggering a subaqueous slump or slide (Table 1-C). This mechanism is less likely because the lakes at present have low-relief basin morphology, though ~700 years ago the basin(s) may have been substantially deeper and/or steeper.

The prominent chemical feature of 1R is the S-peak above the homogenite (Figure

4b). The proposed seismite is ~10 cm thinner than 1L and appears texturally as a subtle anomaly within the RIP laminated facies (Figure 3). This interval, though, marks the earliest occurrence of S in the RIP core. The proposed source of increased S in RIP

2- following 1R is an influx of groundwater containing high SO4 (Table 1-F or 1-H), from the basal aquifer supplying S. Susanna spring located a few km to the east of RIP (Figure

1b.). The springs that discharge along normal faults at the northern and eastern-edges of

155

the Rieti Basin exist at variable elevations above permeability boundaries, suggesting the

current karst circuit is not in equilibrium with the base level (Soligo et al. 2002).

Movement along faults could permanently change the geometry of the groundwater flow

through the karst circuit and potentially the location of spring discharge (Table 1-H).

During this time, the lake extent was broader (Figure 1b.), so more proximal spring

2- discharge would have been highly communicable to the lake(s). High SO4 groundwater, therefore, feasibly could move to RIP with a seismic trigger (Table 1-F or 1-H).

5.1.3 Alternative interpretations

Beds 1L and 1R may instead be thick layers of rapid sediment accumulation caused by flooding. A flood deposit would also be homogenous and may exhibit the graded sedimentation observed (Gilli et al. 2013). The recurrence interval of floods, however, is more frequent than devastating earthquakes, so a flood signature would be expected more frequently and regularly throughout the record (Wilhelm et al. 2016).

Also, increased OM and detritus would be expected in a flood deposit (Arnaud et al.

2002) and are not observed here. The persistence of certain geochemical signatures following this event also suggests that the cause was not an ephemeral event such as a flood, but a lasting change potentially caused by seismicity. Strong storms are another physical mechanism for sediment disturbance and/or resuspension, but, like floods, would be expected to reoccur on the order of tens of years, not hundreds of years as strong earthquakes reoccur in this region. The physical disturbance causing the

156

homogenite formation at 1L and 1R instead could have been a slope failure external to the lake that caused increased sediment yield, but this scenario is still more likely with a seismic trigger (Table 1-C) given the non-regularity of beds such as 1L and 1R.

As core depth increases so does the error associated with both the age model and the correlation between the two cores, so the attribution of event intervals 1L and 1R to the 1298 CE earthquake hinges on a wider possibility of misidentification than the other events described below. The intensity of this event and the strength of shaking likely felt in the basin, though, encourages the likelihood of a signature in these lake sediments, and the depth intervals selected are the most suitable candidates.

5.2 Event 2, the 1349 Earthquake

5.2.1 Earthquake Description

This earthquake was part of a major seismic sequence that occurred in multiple parts of the Apennines. The 1349 seismic sequence had two main epicentral areas, with mainshocks greater than Mw 6 (Intensity>9) that occurred ~1 day apart (Rovida et al.

2016, Galli and Nasso 2009). The northern epicenter, the event of interest to this study, is located to the NW of L’Aquila on a fault segment that extends to the NW, connecting to normal faults on the eastern edge of the Rieti Basin (Figure 1a., Rovida et al. 2016,

Michetti et al. 1995).

5.2.2 Event 2 Features

157

Event 2 has the strongest expression of the four event layers and is a turning point in the geochemistry of both cores. Sedimentologically, 2L and 2R both are classified as homogenites with many of the same features seen in the event 1 homogenites (Table 1-B

;Figures 4a, 4b), including a grading of particle size within the bed seen in smear slides (

Figure 3S1). The peaks in redox-sensitive elements (Fe, Mn, S) at 2L and 2R and other

events (Table 3) also indicate a scenario where upper sediment is sufficiently disturbed to

induce ventilation of lake bottom water and/or pore water. This would allow reduced

species to form insoluble Fe or Mn oxides and oxyhydroxides. These authigenic minerals,

with higher density than OM, concentrate at the base of the seismite and may be the

source of the concurrent MS peaks (Figures 3b, 3c and 4a, 4b).

The strong peaks in S and Sr associated with 2L and 2R seismites (Table 2;

Figure 4) point to another effect of seismicity: a transient groundwater chemistry change

2- as with the S peak associated with event 1 (Table 1-G). Input of high-SO4 groundwater

is the most likely S source at 2L and 2R. This is supported by the presence of pyrite

(FeS2) at the base of 2L and 2R (Figure 3). Pyrite is a common feature of the base of event layers 2, 3 and 4 (Table 3) and is not pervasive in the rest of LUN and RIP cores.

FeS2 requires sufficient Fe, S, and OM to form (Davison et al. 1985; Routh et al. 2004),

and S content, nearly absent prior to event 2, appears to be the limiting factor. Sulfate

influx combined with OM resuspension could stimulate OM degradation, first aerobically

then utilizing sulfate (Holmer and Storkholm 2001). The sulfide produced by this

158

reaction is needed for the conversion of metastable iron-monosulfides to stable pyrite

(Davison et al. 1985; Wilkin et al. 1996; Morgan et al. 2012).

The numerous gastropod shells at the top of 2L are notable because they are rare in both lake cores. The presence and preservation, especially in 2L, may represent material transported from an oxygenated littoral zone of the lakes and surrounding marshes. Freshwater gastropods are most common in these zones (Hu et al. 2016) and exhibit a strong codependence with near-shore macrophytes (Pip and Stewart, 2011).

Sediment focusing, enhanced by seismic shaking, may have transported both the OM and shells into the lake depocenter (Leroy et al. 2002; Rosen 2015). In 2R, the homogenite instead contains numerous shell fragments, possibly signifying more distal transport.

13 Further evidence is provided by the coeval depleted δ Corg values in 2L and 2R (Table 3;

13 Figure 4). Like event 1, a depletion in δ Corg, especially coupled with increases in C/N

values suggests increased organic matter recycling.

The post-Event 2 core stratigraphy diverges strongly between LUN and RIP, and

indicates that in addition to transient groundwater chemistry changes, this event may

have induced sustained changes in water feeding the Rieti Plain and the degree of

connection between the lakes, either by mechanism 1-F or 1-H (Table 1). After Event 2

in LUN, the geochemically stable laminated facies is absent, suggesting a permanent shift

in the system. The closest mainshock of the 1349 seismic sequence was located along the

159

Salto River, a major river converging with the Velino River and the Rieti Plain. An earthquake with this high magnitude (Mw=6.27) are known cause landslides, slope instability, and increased sediment yield (Table 1-C; Wang et al. 2015). Or, an avulsion or channel abandonment on the Velino River could have triggered a similar response.

These effects could have significantly changed surface water flow into or through the

Rieti Basin. This disrupted water balance, with increased groundwater and decreased surficial flow, could cause the marsh belt surrounding the lakes to expand. It also potentially explains the problems with flooding cited during the14th century (Leggio and

Serva 1991), when excess water began entering from an unseen contribution rather than a manageable surface component. Further evidence for this source water transition is provided by reexamined Alnus pollen trends in LUN (Mensing et al. 2015; 2016).

Variations in Alnus pollen, a proxy for marsh extent and groundwater level (Claessens et al. 2010), increased during ~1400 to 1800 CE, potentially also benefitting from a system with increased groundwater input. This supports the paleoseismic driver for Event 2 rather than an anthropogenic cause related to water diversion of reclamation efforts

(Mensing et al. 2015, 2016).

5.2.3 Alternative interpretations

While the abruptness of all proxy changes at event 2 points to a catastrophic trigger, it is possible that a change in balance of surface to groundwater inflow/outflow could cause a similar signature and would hypothetically make groundwater chemical

160

constituents more prominent if dilution from siliciclastic riverine sediment input was

decreased. The subsequent lower water levels would reduce seasonal anoxia and

encourage more frequent lake water mixing, leading to peaks in redox-sensitive elements

as they remain oxygenated. Increased C:N, CaCO3% and OM% support this scheme, as a

shallower water column would be more favorable for calcite preservation and increased

light penetration for enhanced productivity. Instead of a physical transport mechanism,

the littoral debris (gastropods and shell fragments) may indicate a lake low stand where

the coring location was very close to the littoral or marsh zone. This may also explain the

source of S; as increased S burial from sulfate reduction in the marsh belt (Novak et al,

13 2013; Roeser et al. 2016). The δ Corg depletion concurrent with 2L and 2R could also be explained by residual OM after sulfate reduction and methanogenesis took place in a proximal low-energy wetland, both processes leaving a residual organic C pool more

13 depleted in δ Corg (Cohen 2003; Dean 2006). Alternative S sources were considered, including input from the catchment as increased soil flux, from air pollution, or from internal cycling. These are unlikely, however, because soil flux would likely be accompanied by increases in siliciclastic material (Russell and Werne 2009) and detrital

elements (Ti, Rb, K, Fe) are virtually absent at this interval (Figure 4a. and b). Air

pollution would contribute comparable S amounts (Nriagu and Soon, 1985, Urban et al.

1999), but pollution is not a viable source in the pre-industrial era. Internal potential

sources include early diagenetic formation of organic S compounds, but this is not the

161

most likely source because the S peak and OM% peaks are offset and ratios of Corg:STS

remain low (Figure 4; Davison et al. 1985; Nriagu and Soon 1985; Ding et al. 2016).

Human alteration is another potential explanation for the hydrological regime

changes at event 2. Over the past ~1000 years in the Rieti Basin, humans have been

attempting to control water flow (Mensing et al. 2015; 2016). These efforts were

primarily focused on canal construction to drain excess surface water to create more land

area for crops and livestock with less standing water for malaria vectors. While several

canal construction campaigns were documented during the 14th century, none were

reportedly successful in managing water flow or water levels (Lorenzetti 2009).

Subsequent and more lasting reclamation efforts are discussed in sections 5.3.2 and 5.4.2,

below.

5.3 Event 3, 1639 CE

5.3.1 Earthquake description

This earthquake, with a magnitude of 6.2, occurred near the town of Amatrice and has been cited as a paleoseismic analogue of the August 24th 2016 magnitude 6.0 event,

as the events ruptured the same fault segments and likely occurred in the same manner: as

a sequence of events characterized by a mainshock and several hundred lower magnitude

aftershocks (Chiaraluce et al. 2017). This earthquake sequence was different from the

1298 and 1349 events in epicentral locations, as it occurred to the East relative to the

Rieti Basin and at a distance of ~30km. The shaking was felt moderately in Rieti

162

(Intensity= 5), reportedly less than for the 1298 event (Intensity=9 to 10) or the 1349

event (Rovida et al. 2016).

5.3.2 Event 3 Features

The base of Event 3 is at the abrupt cessation of a distinct lithology in both LUN

and RIP. The event layers consist of peaks in S and Sr (3L, Table 3). Although 3L and 3R

beds are sedimentologically uniform, the mechanism for proposed formation, a change in

groundwater hydrology and chemistry (Table 1-E or 1-F) differs from the forces associated with forming homogenites (Table 1-B). The distinctive interval prior to Event

3 is characterized by low siliciclastic elements, high OM content and ~1mm-thick disrupted laminations of calcite and/or gypsum (Figure 3). Peaks in Sr and Ca are associated with these laminations in both lakes (Figure 4a., b.).

The paleoseismic signature of Event 3 must be considered in parallel with the first successful drainage project of the basin, the Cava Clementina, because the project, beginning in 1601 CE, just before the 1639 CE earthquake, had a significant impact on surface water dynamics in the study area (Lorenzetti 1989; Leggio and Serva 1991). We propose the core depth corresponding to this drainage project at ~200cm in LUN and~245cm in RIP, indicated by orange bars (“CC”) in Figures 4a. and b. The sedimentology is distinct at these depths where thin, discontinuous gypsum and calcite laminations or stringers persist for ~5 cm. Sudden lake level lowering caused by the

163

drainage project would could chemical saturation, causing calcite with minor gypsum

precipitation in the water column and/or uppermost sediments. Peaks in Sr with

concurrent peaks in Ca associated with these layers also signifies ion saturation in a water

level lowering scenario (Rush 2010; Kylander et al. 2011). This depth also corresponds

with a declining trend in Alnus pollen, cited as the successful reclamation of the basin

(Mensing et al., 2015; 2016).

The depth of Event 3, directly above this evaporite bed, is marked in both cores

by a distinct base, then peaks in S and pyrite. The proposed formation mechanism for 3L

and 3R is a transient increase in groundwater discharging to the lake co-seismically and

post-seismically (Table 1-E, 1-F). These earthquake effects could have temporarily

disrupted reclamation efforts by raising water levels, decreasing chemical saturation.

Event 3 has a different structure than other events (Table 3) because the Cava Clementina

was simultaneously working to change the system, and it also has a different

manifestation in 3L and 3R, likely because the project separated the lake basins. Above

both 3L and 3R, CaCO3% increases again (Figures 4a. and b.), suggesting reclamation

and water drainage resumed.

The similarity between the 1639 CE seismic sequence and the August 24th 2016

Amatrice earthquake is also integral for gauging potential event formation mechanisms.

These events ruptured the same fault system and had similar magnitude and depth

164

(Chiaraluce et al. 2017). The intensities in Rieti of both the 1639 and 2016 mainshocks

were reported as moderate (MMI= V; USGS, 2016, Rovida et al. 2016). Based on

observations made by colleagues, the ground shaking associated with the 2016

mainshocks in the Rieti Basin may have been insufficient to generate seiche waves in the

lakes. The effects of the Aug. 24th 2016 event on groundwater chemistry in the Rieti

Basin area, though, were notable (Chapter 2). Although flow was not explicitly measured,

increases in chemical constituents in spring waters were attributed to enhanced aquifer

hydraulic conductivity and pore pressure. This same increased flow and spring discharge

response has also been observed co-seismically and post-seismically in faulted carbonate aquifers in the region (Table 1-E; Carro et al. 2005; Falcone et al. 2012). It is plausible, then, that after the 1639 CE earthquake groundwater flow in the surrounding carbonate aquifer and/or alluvial plain-filling aquifer (Table 1-F) responded in the same way. The connection of the lakes to groundwater and springs at this interval is explicit, illustrated in a historical map dated 1632 (Leggio and Serva 1991) that shows a river from Cantalice spring (sourcing the same aquifer as S. Susanna). The Cantalice spring, today, is not connected to LUN and discharges along the major normal fault on the eastern edge of the plain (Calderini et al. 1998). The former connection between “fiume Cantalice” and LUN may have been a direct conduit for groundwater signals in the carbonate aquifer to reach the lakes. This may also explain why the S signal at Event 3 is stronger in LUN than in

RIP (Figure 4a. and 4b.).

165

5.3.3 Alternative hypotheses

There are two challenges in ascribing Event 3 to paleoseismicity; first, historical

maps of the Rieti Basin indicate that the lakes were probably separate water bodies in this

period (Leggio and Serva 1991) and may have responded differently to this event. Second

is that Event 3 may have been overprinted by anthropogenic changes to the hydrologic

regime because of the ~30 years between Cava Clementina construction and the

earthquake. Recognition of the depth where construction began aided in identifying Event

3, but the signature of the event, attributed to a pulse of increased groundwater influx,

could have also been caused by a meteorological, instead of seismic, driver. A year with

above average precipitation could have increased groundwater levels in the basin or

increased spring flow. When discerning between paleoseismicity and meteorological

trends, though, timescale and duration are important considerations. Earthquake

hydrology changes (Table 1-E, F and G) proposed to contribute to the chemical signature

of the LUN and RIP seismites, can be transient (Table 1-E) or lasting (Table 1-F and G).

Using the SAR determined by the age models, events 3L and 3R represent eight years and six years of sedimentation, respectively. The changes observed in response to the

2016 seismic sequence were likely transient, on the order of months, so if the 1639 CE event is truly analogous in terms of aquifer response, then the lasting sedimentary signature is likely not due to earthquake effects. Meteorological trends effecting groundwater flow, however, have a similar timescale, on the order of months or several

166

years depending on the phenomena. A climate trend, such as those controlled by the

North Atlantic Oscillation (NAO) in this region (Piovesan and Schirone 2000), would persist on the order of hundreds of years, so is not the best explanation for this apparent groundwater influx pulse. Timing and duration are important factors in attributing these signals to paleoseismicity, though added complication occurs due to the offset between groundwater seismic changes and in-lake processes leading to sedimentation and preservation.

Another potential difficulty in attributing Event 3 to a coeval trigger is that the age correlation before this depth interval (1550 – 1650 CE, Figure 2) is based on a tie point where an MS trough persists in LUN, but in RIP there is a peak within this trough.

There was no MS peak at the base of 3L, contrasting with other event layers, so the correlation between 3L and 3R is not as robust as for the other events.

5.4 Event 4, 1703

5.4.1 Earthquake details

The 1703 seismic sequence consisted of two large events both close (<30km) to the Rieti Basin. The first occurred west of Norcia, to the northeast of the basin (Mw 6.9) and the second, two weeks later, was located to the east of the basin (Mw 6.7). Rupture

167

occurred on the same fault system (Amatrice-Monte Vettore fault segments) as the 2016-

2017 sequence but also included segments closer to the Rieti Basin including those along the eastern edge (Galli et al. 2005). Like the 1639 and 2016 events, the 1703 events were part of a seismic sequence consisting of multiple earthquakes of considerable magnitude.

The 1703 mainshocks, though, represent the highest magnitude events to strike the region in the historical period. These events were felt strongly in the study area, with a reported intensity VII at Apoleggia, a small town 1.5 km to the north of the lakes (Rovida et al.

2016).

5.4.2 Event 4 Features

The Event 4 seismite is marked by an abrupt transition that includes both a homogenite (Table 1-B) and excursion in elemental chemistry profiles, attributed to groundwater changes (Table 1-E or 1-F). RIP and LUN were functioning as separate depocenters, creating some disparities between 4L and 4R, outlined in Table 3. The base of the 4R homogenite appears more erosive than 4L, and mottling by iron sulfides continues in the homogenous bed, possibly because of different water chemistry during re-settling. The more pronounced chemical signature in 4L could denote its proximity to the basin edge fault and associated groundwater springs. This seismite in LUN is most

13 similar to 2L and 2R in chemical and isotopic signature (Figure 4a.). The 4L δ Corg

values at the base are depleted then become ~7‰ enriched within the homogenite. The

168

depletion could be a sign of increased OM degradation and recycling, as observed at 2L.

13 The enriched upper may be a signal of the more enriched Corg signature of littoral

13 macrophytes, if water oscillations increased sediment transport. The δ Corg at 4R do not exhibit the same depletion, though this may be an artifact of sampling resolution.

Like the 1639 event, the 1703 seismic sequence may provide another comparison to the modern hydrochemical response (Chapter 2) because of the similar fault segments ruptured, though one of the 1703 mainshocks was closer to the basin. This event has similarities to both Event 3 and Event 2, and in terms of epicentral location it was intermediate between 1349 and 1639 but greater than both in earthquake magnitude. It is plausible, then, that the seismite at Event 4 signifies both strong shaking and groundwater changes related to the earthquake, and permanent aquifer deformation (Table 1H; Manga and Wang 2015), may have altered the connected to the lakes.

5.4.3 Alternative hypotheses

The elevated OM content of 4L and 4R, discussed above as a seismite feature, could instead be due to hemp cultivation and harvesting. This crop was not grown prior to this time, and part of the procedure for harvesting hemp is soaking the stalks in standing water, or “retting.” This practice was likely carried out in/around LUN and/or RIP

(Mensing et al. 2015), and could introduce POM that could have served as additional nutrients for productivity. This explanation, however, does not account for the chemical

169

signature of 4L and 4R, though, nor does it provide a transport mechanism to the

depocenter of the lakes.

Another alternative for this event signature is a shift in regional climate.

Temperature and precipitation records from Central Europe suggest that the period

preceding this event, the “Little Ice Age,” (~1400-1700 CE) was anomalously cool and

wet in many regions. Event 4 marks the transition to a warmer, more stable climate, and

the hydrologic regime could have reached a tipping point where gradual water level

decline in the basin suddenly cut off lake water inputs or connectivity (Randsaul-Wedrup et al. 2016). Also, the implementation or construction of Cava Clementina may have occurred over several years and its effects may not be observable in the sediment record until Event 4. The specific manipulations involved in this project are not known, so hydrological change caused by it can only be speculated.

5.5 The 1785 earthquake and more recent events

The historical earthquake in 1785 was strong (Mw 5.76) and very close (<5 km) to the

lake basins, yet no signature was observed in the sediment record in either lake. This may

indicate a decreased sensitivity of these lakes in recording regional signals following the

1703 event. Or, this could show a sensitivity threshold based on earthquake magnitude

and direction to epicenter. Previous work found a Mw 6.0 threshold for paleoearthquakes

to cause secondary ruptures other than the main fault, or Mw 6.5 for the normal fault

system by Pie di Colle (Michetti et al. 1995). Also, there is no evidence that this event

170

was a multi-shock sequence similar to the other four earthquakes discussed above. These factors may have limited an observable chemical response in nearby aquifers, and while physical shaking of the basin sediment was likely, these effects were not visually identifiable in this core section. The modern system of vegetation and agriculture was established by 1750, characterized by human activities overpowering natural signals of climate or environmental change (Mensing et al. 2016) This work suggests that after the mid- 18th century, anthropogenic activities may have been significant enough to overprinted signals of earthquakes.

Examples of these anthropogenic manipulations are the diversion of water into canals that drain surface water from the plain, artificial groundwater pumping, and the 20th century damming of the Salto and Turano Rivers that both fed into the Velino River

(Carrara et al. 2004; Iadanzo and Napolitano 2006) all masked natural flow variability.

The S. Susanna spring in its modern configuration flows from its source at the faulted contact with the carbonate aquifer rock along a man-made channel that is partially diverted out of the plain through the river and partially into RIP. This channel was constructed around 1940, prior to which the flow of S. Susanna water was uninhibited.

The trophic state of the lakes is also the highest in the lakes’ history (Franceschini et al.

2004), so biological influence on lake chemistry significantly dampens lake sensitivity to external forcing. Also, events 1-4 occurred during the Little Ice Age period, when plague

171

caused widespread population decline (Mensing et al. 2015; 2016), thereby limiting the

degree of anthropogenic overprint on natural phenomena such as earthquakes.

6 Conclusions

Shallow or small volume lakes are not typically investigated as records as paleoseismicity, especially in lower elevation areas where mass-wasting events do not

readily occur. In the Rieti Basin and central Apennine region, large magnitude

earthquakes over the last millennium have been well documented, providing a means of

investigating whether paleoseismic events register in sediments cores from the two lakes

studied. This combined sedimentological and geochemical proxy investigation identified

event layers that coincided with major earthquakes in the region. These earthquakes,

1298, 1349, 1639, and 1703 CE, had mainshocks greater than Mw=6.2 and epicenters

within 30km of the Rieti Basin. The event layers were therefore proposed as seismites

that feature a signature of physical shaking and/or sediment disturbance manifested as

homogenous deposits with density-graded sedimentation. All events have chemical signatures associated with them, the dominant theme being increased inputs from regional groundwater sources with associated chemical signatures. The 1298 and 1349

CE events have seismites in both lakes that indicate sediment resuspension caused by

2- ground shaking and water oscillation as well as increased contributions of high SO4

groundwater. Changes following the 1349 CE earthquake were persistent, reshaping

sediment delivery and lake hydrography. The 1639 event is more difficult to discern

172

because it occurs near the time of initiation of a major reclamation project, Cava

Clementina, a canal network implemented to drain excess water from the basin that also

undoubtedly affected lake sedimentation and hydrochemistry. The 1703 CE seismite was

remarkably similar in geochemical structure to the 1349 CE event, although modern

reclamation projects and landscape alterations were likely beginning to play an

overriding role in lake sedimentation. Events 1 and 3 are more noticeable in LUN than in

RIP, another results suggesting lake-basin separation during this period and possibly the effect of the LUN depocenter’s proximity to the basin-edge faults. Varied effects of major earthquakes in this area are controlled by the epicentral location in relation to major aquifers, rivers, and basins of alluvial fill that may be more susceptible to seismic shaking.

Alternative causes for event layers include major flooding events, channel avulsion of the Velino River and its relation to the lakes in relation to the 1349 event, lake level changes beyond certain thresholds, and anthropogenic landscape modification.

These different scenarios cannot be ruled out, though collectively the concurrence of each event with major earthquakes, the combined geochemical and sedimentological component of most events, and the synchronicity in both lake records is best explained by seismic forcing mechanisms. Subsequent large magnitude earthquakes, including the

1785 CE event, did not leave a discernible signal, likely because anthropogenic manipulation including eutrophication muted or overprinted the paleoseismic signatures.

173

The attribution of physical and chemical signals in event layers to major earthquakes also

has the potential to confirm the core chronology or constrain the age model tie-points in these records. A future investigation of sediment cores from similar sized lakes in this region would aid in attributing the geochemical signature of the event layers to major earthquakes.

174

References Amoruso, A., Crescentini, L., Petitta, M., Rusi, S., & Tallini, M. (2011) Impact of the 6 April 2009 L’Aquila earthquake on groundwater flow in the Gran Sasso carbonate aquifer, Central Italy. Hydrologic Processes, 25, 1754–1764 Archer, C.C, Noble, P.J., Kreamer, D.K., Piscopo, V., Petitta, M., Rosen,M., Poulson, S., Piovesan, G., & Mensing, S., (2016) Hydrochemical determination of sourcewater contributions to lakes Lungo and Ripasottile (central Italy). Journal of Limnology Arnaud, F., Lignier, V., Revel, M., Desmet, M., Beck, C., Pourchet, M., Charlet, F., Trentesaux, A. and Tribovillard, N., (2002). Flood and earthquake disturbance of 210Pb geochronology (Lake Anterne, NW Alps). Terra Nova, 14(4), pp.225-232. Avşar, U., Hubert-Ferrari, A., Batist, M., Lepoint, G., Schmidt, S., & Fagel, N. (2014) Seismically-triggered organic-rich layers in recent sediments from Göllüköy Lake (North Anatolian Fault, Turkey). Quaternary Science Reviews, 103, 67–80 Avşar, U., Jónsson, S., Avşar, Ö., & Schmidt, S. (2016). Earthquake‐induced soft‐sediment deformations and seismically amplified erosion rates recorded in varved sediments of Köyceğiz Lake (SW Turkey). Journal of Geophysical Research: Solid Earth, 121(6), 4767-4779. Bertrand, S., Charlet, F., Chapron, E., Fagel, N., & De Batist, M. (2008). Reconstruction of the Holocene seismotectonic activity of the Southern Andes from seismites recorded in Lago Icalma, Chile, 39 S. Palaeogeography, Palaeoclimatology, Palaeoecology, 259(2), 301- 322. Boschetti, T., Cortecci, G., Toscani L., & Iacumin, P. (2011). Sulfur and oxygen isotope compositions of Upper Triassic sulfates from northern Apennines (Italy): paleogeographic and hydrogeochemical implications. Geologica Acta 9(2),129-147. Bozzano, F., Caserta, A., Govoni, A., Marra, F., & Martino, S. (2008). Static and dynamic characterization of alluvial deposits in the Tiber River Valley: new data for assessing potential ground motion in the city of Rome. Journal of Geophysical Research: Solid Earth, 113(B1). Brothers, L. A., Engel, M. H., & Elmore, R. D. (1996). The late diagenetic conversion of pyrite to magnetite by organically complexed ferric iron. Chemical Geology, 130(1-2), 1-14. Brunner, P., Simmons, C. T., & Cook, P. G. (2009). Spatial and temporal aspects of the transition from connection to disconnection between rivers, lakes and groundwater. Journal of Hydrology, 376(1), 159-169.

175

Calderoni, G., Carrara, C., Ferreli, L., Follieri, M., Gliozzi, E., Magri, D., Narcisi, B., Parotto, M., Sadori, L., & Serva, L. (1994). Palaeoenvironmental, palaeoclimatic and chronological interpretations of a late Quaternary sediment core from Piana di Rieti (central Apennines, Italy). Giornale di Geologia 3(56/2), 43-72. Camerieri, P., Mattioli T. (2014) Archaeology and environmental changes along the Velino River. Mem. Descr. Carta Geol.d’It. XCVI, 169-188. Caraco, N. F., Cole, J J., & Likens. G. E. (1993) Sulfate control of phosphorus availability in lakes. Hydrobiologia 253: 275. doi:10.1007/BF00050748 Carannante, S., Monachesi, G., Cattaneo, M., Amato, A., & Chiarabba, C. (2013). Deep structure and tectonics of the northern‐central Apennines as seen by regional‐scale tomography and 3‐D located earthquakes. Journal of Geophysical Research: Solid Earth, 118(10), 5391-5403. Carrara, C. Ferreli, L., Guerrieri, L. & Serva, L. (1994). The Rieti Intermountain Basin and St. Francesco d’Assisi. Field Trip Guide From the 32nd International Geological Congress, Florence, Italy. Vol. 3, D01 to P13 Carrillo, E., Beck, C., Audemard, F. A., Moreno, E., & Ollarves, R. (2008). Disentangling late Quaternary climatic and seismo-tectonic controls on Lake Mucubají sedimentation (Mérida andes, Venezuela). Palaeogeography, Palaeoclimatology, Palaeoecology, 259(2), 284-300.. Carro, M., Amicis, M. & Luzi, L. (2005) Hydrogeological Changes Related to the Umbria Marche Earthquake of 26 September 1997 (Central Italy). Nat Hazards 34(3), 315–339. doi:10.1007/s11069-004-2074- Chapron, E, Beck, C, Pourchet, M & Deconinck, JF. (1999) 1822 earthquake‐triggered homogenite in Lake Le Bourget (NW Alps). Terra Nova. doi:10.1046/j.1365- 3121.1999.00230. Chiaraluce L, Stefano R, Tinti E, Scognamiglio L, Michele M, Casarotti E, Cattaneo M, Gori P, Chiarabba C, Monachesi G, Lombardi A, Valoroso L, Latorre D, & Marzorati S. (2017). The 2016 Central Italy seismic sequence: a first look at the mainshocks, aftershocks, and source models. Seismological Research Letters, 88(3), 757-771. Claessens, H., Oosterbaan, A., Savill, P., & Rondeux, J. (2010). A review of the characteristics of black alder (Alnus glutinosa (L.) Gaertn.) and their implications for silvicultural practices. Forestry, 83(2), 163-175.

176

Claesson, L., Skelton, A., Graham, C., Dietl, C., Mörth, M., Torssander, P., & Kockum, I. (2004). Hydrogeochemical changes before and after a major earthquake. Geology, 32(8), 641–644. doi:10.1130/G20542. Cohen A. S. (2003) Paleolimnology: The History and Evolution of Lake Systems. Oxford University Press, 528 pp. Couture, R. M., Fischer, R., Van Cappellen, P., & Gobeil, C. (2016). Non-steady state diagenesis of organic and inorganic sulfur in lake sediments. Geochimica et Cosmochimica Acta, 194, 15-33. Daele, V. M., Moernaut, J., Doom, L., Boes, E., Fontijn, K., Heirman, K., Vandoorne, W., Hebbeln, D., Pino, M., Urrutia, R. & Brümmer, R., (2015) A comparison of the sedimentary records of the 1960 and 2010 great Chilean earthquakes in 17 lakes: Implications for quantitative lacustrine palaeoseismology. Sedimentology, 62(5), 1466- 1496. Davison, W., Lishman, J. P., & Hilton, J. (1985). Formation of pyrite in freshwater sediments: Implications for C/S ratios. Geochimica et Cosmochimica Acta, 49(7), 1615-1620. Dean Jr., W.E., 1974. Determination of carbonate and organic matter in calcareous sediments and sedimentary rocks by loss on ignition: comparison with other methods. Journal of Sedimentary Petrology 44, 242-248 Dean, W. E. (2006). Characterization of Organic Matter in Lake Sediments from Minnesota and Yellowstone National Park. USGS Publications No. 2006-1053. De Vos W., Tarvainen T. (Chief-editors), Salminen R., Reeder S., De Vivo B., Demetriades A., Pirc S., Batista M.J., Marsina K., Ottesen R.-T., O'Connor P.J.1, Bidovec M., Lima A.4, Siewers U., Smith B., Taylor H, Shaw R., Salpeteur I.1, Gregorauskiene V., Halamic J., Slaninka I., Lax K., Gravesen P., Birke M., Breward N., Ander E.L., Jordan G., Duris M., Klein P., Locutura J., Bel-lan A., Pasieczna A., Lis J., Mazreku A., Gilucis A., Heitzmann P., Klaver G., Petersell V. (2006) Geochemical atlas of Europe. Part 2– interpretation of geochemical maps, additional tables, figures, maps. and Related Publications. Geological Survey of Finland. Espoo, Finland. Geol, 98(2), 69-82. Doig, R. (1998) 3000-year paleoseismological record from the region of the 1988 Saguenay, Quebec, earthquake. Bulletin of the Seismological Society of America 88:5, 1198-1203. Falcone A., Carucci, V., Falgiani A., Manetta M., Parisse B., Petitta M., Rusi S., Spizzico M., Tallini M. (2012). Changes on groundwater flow and hydrochemistry of the Gran Sasso carbonate aquifer after 2009 L’Aquila earthquake. Italian Journal of Geosciences, 131(3), 459-474.

177

Fanetti, D., Anselmetti, F.S., Chapron, E., Sturm, M., & Vezzoli, L. (2008) Megaturbidite deposits in the Holocene basin fill of Lake Como (southern Alps, Italy). Palaeogeography, Palaeoclimatology, Palaeoecology 259, 323-340. Galassi, D. M., Lombardo, P., Fiasca, B., Di Cioccio, A., Di Lorenzo, T., Petitta, M., & Di Carlo, P. (2014). Earthquakes trigger the loss of groundwater biodiversity. Scientific reports, 4, 6273. Gallet, Y., Genevey, A., & Le Goff, M. (2002). Three millennia of directional variation of the Earth’s magnetic field in western Europe as revealed by archeological artefacts. Physics of the Earth and Planetary Interiors, 131(1), 81-89. Galli, P., Galadini, F., & Calzoni, F. (2005). Surface faulting in Norcia (central Italy): a “paleoseismological perspective”. Tectonophysics, 403(1), 117-130. Galli, P. A. C., & Naso, J. A. (2009). Unmasking the 1349 earthquake source (southern Italy): paleoseismological and archaeoseismological indications from the Aquae Iuliae fault. Journal of Structural Geology, 31(2), 128-149. Galli, P. A., Giaccio, B., Messina, P., Peronace, E., & Zuppi, G. M. (2011). Palaeoseismology of the L'Aquila faults (central Italy, 2009, M w 6.3 earthquake): Implications for active fault linkage. Geophysical Journal International, 187(3), 1119-1134. Gilli, A., Anselmetti, F. S., Glur, L., & Wirth, S. B. (2013). Lake sediments as archives of recurrence rates and intensities of past flood events. In Dating torrential processes on fans and cones (pp. 225-242). Springer Netherlands. Heiri, O., Lotter, A., Lemcke, G., 2001. Loss on ignition as a method for estimating organic and carbonate content in sediments: reproducibility and comparability of results. Journal of Paleolimnology 25, 101-110. Holbrook, J., Autin, W. J., Rittenour, T. M., Marshak, S., & Goble, R. J. (2006). Stratigraphic evidence for millennial-scale temporal clustering of earthquakes on a continental-interior fault: Holocene Mississippi River floodplain deposits, New Madrid seismic zone, USA. Tectonophysics, 420(3), 431-454. Holmer, M., & Storkholm, P. (2001). Sulphate reduction and sulphur cycling in lake sediments: a review. Freshwater Biology, 46(4), 431–45 Hu, Z., Sun, X., Cai, Y., Guo, L., Chen, Q., Liu, T., Shi, F., & Yang, L. (2016). The habitat type and trophic state determine benthic macroinvertebrate communities in lowland shallow lakes of China. Journal Of Limnology, 75(2).

178

Hubert-Ferrari, A., Avşar, U., El Ouahabi, M., Lepoint, G., Martinez, P., & Fagel, N. (2012). Paleoseismic record obtained by coring a sag-pond along the North Anatolian Fault (Turkey). Annals of Geophysics, 55(5), 929-953. Iadanza, C., & Napolitano, F. (2006). Sediment transport time series in the Tiber River. Physics and Chemistry of the Earth, Parts A/B/C, 31(18), 1212-1227. Jang, C. S., Liu, C. W., Chia, Y., Cheng, L. H., & Chen, Y. C. (2008). Changes in hydrogeological properties of the River Choushui alluvial fan aquifer due to the 1999 Chi-Chi earthquake, Taiwan. Hydrogeology Journal, 16(2), 389-397. Kylander, M. E., Ampel, L., Wohlfarth, B., & Veres, D. (2011). High‐resolution X‐ray fluorescence core scanning analysis of Les Echets (France) sedimentary sequence: new insights from chemical proxies. Journal of Quaternary Science, 26(1), 109-117. Leggio, T. (1995) Trasformazioni del paesaggio dei monti Sabini dall'eta romana al medioevo. In: Leggio, T., Marini, M. (Eds.), Il paesaggio della conca reatina. Problemi ed esperienze di una ricerca multidisciplinare, pp. 51-70. Rieti. Leggio, T. & Serva, L. (1991). La bonifica della Piana di Rieti dall'et a romana al Medioevo: influenze sui mutamenti del paesaggio. Not. dell'Enea 25-70. Leroy, S., Kazancı, N., İleri, Ö., Kibar, M., Emre, O., McGee, E. & Griffiths, H. (2002) Abrupt environmental changes within a late Holocene lacustrine sequence south of the Marmara Sea (Lake Manyas, N-W Turkey): possible links with seismic events. Marine Geology 190(1-2), 531–552. Liu, C.-P., Wang, C.-H. & Hwang, L.-S. (2010) Temporal variation of seepage water chemistry before and after the Hengchun Ms 7.2 earthquake in south Taiwan. Geoderma 155(1-2), 107–114. Lorenzetti, R., (2009) La terra e le acque. Trasformazioni e persistenze del paesaggio della Valle Reatina, Rieti. Löwemark, L., Chen, H. F., Yang, T. N., Kylander, M., Yu, E. F., Hsu, Y. W., ... & Jarvis, S. (2011). Normalizing XRF-scanner data: a cautionary note on the interpretation of high- resolution records from organic-rich lakes. Journal of Asian Earth Sciences, 40(6), 1250- 1256. Manga M., Wang C.-Y (2015) Earthquake Hydrology. In: Gerald Schubert (editor-in-chief) Treatise on Geophysics, 2nd edition, Vol 4. Oxford: Elsevier. 305-328. Matsuda, J.-I. (2000). Seismic deformation structures of the post-2300 a BP muddy sediments in Kawachi lowland plain, Osaka, Japan, Sedimentary Geology 135, 99-116.

179

Mensing, S. A., Tunno, I., Sagnotti, L., Florindo, F., Noble, P., Archer, C., Zimmermann, S., Pavón-Carrasco, F., Cifani, G., Passigli, S., & Piovesan, G. (2015) 2700 years of Mediterranean environmental change in central Italy: A synthesis of sedimentary and cultural records to interpret past impacts of climate on society. Quaternary Science Reviews, 116, 72-94. Mensing, S., Tunno, I., Cifani, G., Passigli, S., Noble, P., Archer, C. & Piovesan, G. (2016). Human and climatically induced environmental change in the Mediterranean during the Medieval Climate Anomaly and Little Ice Age: A case from central Italy. Anthropocene, 15, 49-59. Michetti A.M., Brunamonte F., Serva L., & Whitney R.A. (1995). Seismic Hazard Assessment from Paleoseismological evidence in the Rieti Region, Central Italy. Bulletin of the Association of Engineering Geologists: Perspective in Paleoseismology, Ch. 6. Michetti A.M., Esposito E., Guerrieri L., Porfido S., Serva L., Tatevossian R., Vittori E., Audemard F., Azuma T., Clague J., Comerci V., Gürpinar A., Mc Calpin J., Mohammadioun B., Morner N.A., Ota Y., Roghozin E. IN: Serva, L., Blumetti, A., Esposito, E., (2015) Earthquake Environmental Effects, intensity and seismic hazard assessment: the lesson of some recent large earthquakes. Mem. Descr. Carta Geol. D’It. XCVII (2015), pp. 9 – 10 Minissale, A. (2004). Origin, transport and discharge of CO2 in central Italy. Earth-Science Reviews 66: 89- 141. Monecke, K., Anselmetti, F., Becker, A., Sturm, M., & Giardini, D. (2004) The record of historic earthquakes in lake sediments of Central Switzerland. Tectonophysics 394(1-2), 21–40. Morgan, B., Burton, E. D., & Rate, A. W. (2012). Iron monosulfide enrichment and the presence of organosulfur in eutrophic estuarine sediments. Chemical Geology, 296, 119-130. Nealson, K. H., & Myers, C. R. (1992). Microbial reduction of manganese and iron: new approaches to carbon cycling. Applied and Environmental Microbiology, 58(2), 439–443. Novak, M., Jacobson, G., Norton, S., Stepanova, M., Grimm, E., Jackova, I. & Buzek, F. (2013) Sulfur isotope evidence for changing input of continental and marine aerosols in a 60,000-year sediment core from Lake Tulane, central Florida, USA. Chemical Geology 349, 110–116. Nriagu, J.O., Soon, Y. K. (1985) Distribution and isotopic composition of sulfur in lake sediments of northern Ontario, Geochimica et Cosmochimica Acta, 49:3, 823-834.

180

Pavon-Carrasco, F.J., Osete, M.L., Torta, J.M., Gaya-Piqu_e, L.R. (2009). A regional archeomagnetic model for Europe for the last 3000 years, SCHA.DIF.3K: applications to archeomagnetic dating. Geochemistry Geophysics Geosystems 10, Q03013. Petitta, M., Primavera, P., Tuccimei, P., & Aravena, R. (2011). Interaction between deep and shallow groundwater systems in areas affected by Quaternary tectonics (Central Italy): a geochemical and isotope approach. Environmental earth sciences, 63(1), 11-30. Piovesan, G., & Schirone, B. (2000). Winter North Atlantic oscillation effects on the tree rings of the Italian beech (Fagus sylvatica L.). International Journal of Biometeorology, 44(3), 121-127. Pip, E., & Stewart, J. (2011). The dynamics of two plant-snail associations. Canadian Journal of Zoology 54(7):1192-1205. Quattrocchi, F. (1999) In search of evidence of deep fluid discharge and pore pressure evolution in the crust to explain the seismicity style of the Umbria-Marche 1997-1998 seismic sequence (Central Italy). Annali di Geofisica 42(4). Randsalu-Wendrup, L., Conley, D. J., Carstensen, J., & Fritz, S. C. (2016). Paleolimnological records of regime shifts in lakes in response to climate change and anthropogenic activities. Journal of Paleolimnology, 56(1), 1. Roberts, G. P., & Michetti, A. M. (2004). Spatial and temporal variations in growth rates along active normal fault systems: an example from The Lazio–Abruzzo Apennines, central Italy. Journal of Structural Geology, 26(2), 339-376. Roeser, P., Franz, S. O., & Litt, T. (2016). Aragonite and calcite preservation in sediments from Lake Iznik related to bottom lake oxygenation and water column depth. Sedimentology, 63(7), 2253-2277. Rosen, M.R. (2015) The influence of hydrology on lacustrine sediment contaminant records. In: Blais, J.M., Rosen M.R., Smol J.P. (eds) Environmental Contaminants: Using natural archives to track sources and long-term trends of pollution. Springer, Dordrecht Rosenberry, D. O., Lewandowski, J., Meinikmann, K., & Nützmann, G. (2015). Groundwater‐ the disregarded component in lake water and nutrient budgets. Part 1: effects of groundwater on hydrology. Hydrological Processes, 29(13), 2895-2921. Routh, J., Meyers, P., Gustafsson, Ö., Baskaran, M., Hallberg, R. & Schöldström, A. (2004) Sedimentary geochemical record of human–induced environmental changes in the Lake Brunnsviken watershed, Sweden. Limnology and Oceanography 49(5), 1560–1569.

181

Rovida A., Locati M., Camassi R., Lolli B., Gasperini P. (eds), (2016) CPTI15, the 2015 version of the Parametric Catalogue of Italian Earthquakes. Istituto Nazionale di Geofisica e Vulcanologia. doi: http://doi.org/10.6092/INGV.IT-CPTI15 Rush, RA. (2010) High Resolution geochemical XRF data from Elk Lake, Minnesota: A Holocene paleoclimate record from varved lacustrine sediments. Retrieved from http://conservancy.umn.edu/handle/11299/10271 Russell, J. M., & Werne, J. P. (2009). Climate change and productivity variations recorded by sedimentary sulfur in Lake Edward, Uganda/DR Congo. Chemical geology, 264(1), 337- 346. Schnellmann, M., Anselmetti, F.S., Giardini, D., McKenzie, J.A., & Ward, S.N. (2002) Prehistoric earthquake history revealed by lacustrine slump deposits. Geology 30(12), 1131– 1134. Soligo, M., Tuccimei, P., Barberi, R., Delitala, M. C., Miccadei, E., & Taddeucci, A. (2002). U/Th dating of freshwater travertine from Middle Velino Valley (Central Italy): paleoclimatic and geological implications. Palaeogeography, Palaeoclimatology, Palaeoecology, 184(1), 147-161. Spadoni, M., Brilli, M., Giustini, F., & Petitta, M. (2010). Using GIS for modelling the impact of current climate trend on the recharge area of the S. Susanna spring (central Apennines, Italy). Hydrological processes, 24(1), 50-64. Talbot, M. R., & Lærdal, T. (2000). The Late Pleistocene-Holocene palaeolimnology of Lake Victoria, East Africa, based upon elemental and isotopic analyses of sedimentary organic matter. Journal of Paleolimnology, 23(2), 141-164. Tarquini S., Isola I., Favalli M., Mazzarini F., Bisson M., Pareschi M. T., Boschi E. (2007). TINITALY/01: a new Triangular Irregular Network of Italy, Annals of Geophysics 50, 407 - 425. Tarquini S., Vinci S., Favalli M., Doumaz F., Fornaciai A., Nannipieri L., (2012). Release of a 10-m-resolution DEM for the Italian territory: Comparison with global-coverage DEMs and anaglyph-mode exploration via the web, Computers & Geosciences 38, 168-170. doi: doi:10.1016/j.cageo.2011.04.018 Tertulliani, A. (1999). Site effects as inferred from damage severity observation. Geophysical Research Letters 26(13) 1989–1992. Urban N., Ernst K., & Bernasconi S. (1999) Addition of sulfur to organic matter during early diagenesis of lake sediments. Geochimica et Cosmochimica Acta 63: 837–853.

182

USGS Earthquake Hazard Program. (2016) Retrieved from: https://earthquake.usgs.gov/earthquakes/ Valero-Garcés, B., Morellón, M., Moreno, A., Corella, J.P., Martín-Puertas, C., Barreiro, F., Pérez, A., Giralt, S. & Mata-Campo, M.P. (2014) Lacustrine carbonates of Iberian Karst Lakes: Sources, processes and depositional environments. Sedimentary Geology 299, 1– 29. Voltattorni, N & Quattrocchi, F. (2012) Soil gas degassing during the 2009 L’Aquila earthquake: study of the seismotectonic and fluid geochemistry relation. Italian Journal of Geosciences (Boll. Soc. Geol. It.), 131(3): 440-447 Vreca, P., & Muri, G. (2006). Changes in accumulation of organic matter and stable carbon and nitrogen isotopes in sediments of two Slovenian mountain lakes (Lake Ledvica and Lake Planina), induced by eutrophication changes. Limnology and Oceanography, 51(2) 781- 790. Wang, J., Jin, Z., Hilton, R., Zhang, F., Densmore, A., Li, G. & West, A. (2015) Controls on fluvial evacuation of sediment from earthquake-triggered landslides. Geology 43(2). Wilhelm, B., Nomade, J., Crouzet, C., Litty, C., Sabatier, P., Belle, S., ... & Anselmetti, F. S. (2016). Quantified sensitivity of small lake sediments to record historic earthquakes: implications for paleoseismology. Journal of Geophysical Research: Earth Surface, 121(1), 2-16. Wilkin, R. T., Barnes, H. L., & Brantley, S. L. (1996). The size distribution of framboidal pyrite in modern sediments: an indicator of redox conditions. Geochimica et cosmochimica acta, 60(20), 3897-3912.

183

Figure 1a. Map of the study area, including location within the Central Apennines, major faults, lakes, cities, historical earthquake epicenters including epicenters of the 2009 and 2016 seismic sequences.

184

Figure 1b. Close-up map of the Rieti Plain and associated watercourses. The modern lake extent is shown in blue, the white-hatched area surrounding the lakes is the GIS simulation of water level with three addition meters of water occupying the lake basin(s). Major rivers and springs referred to in the text are also indicated. DEM base maps courtesy of Tarquini etal. (2007) and Tarquini et al. (2012).

185

186

Figure 2. Magnetic parameters (inclination, declination, and relative paleointensity) from LUN and RIP cores (LUN from both 2009 and 2012 recoveries) showing correlation of different cores from the same lake, correlation between LUN and RIP, and correlation of the measures parameters to three PSV reference curves and models (Archeomagnetic data from France, Gallet et al., 2002; scha.dif.3k of Pavon-Carrasco et al., 2009).

187

RIETI-RIP13-1B-3B-1-W 1B-2B 1B-3B 1B-4L 1B-5L 1A-1P 4.5

C, Q, Py 4L 1298 C, 1349 4R Mg-C 2R 1703 Q, C

C, Q Gy, 1703 cl

C, Q, Br 1639 Q, C, cl C, Q, 3L 1L Q, C Py C, D, cl 1298 3R Q, Py 2L

C, Py 1639 C, Q, 1349 D, Br C, Q, C, Q, 1R cl D, Gy 1298 C, Q, Mg-C 1.0

1.1

1.2 C, Q

1.3

1.4

LUN RIP 1.5

B) Detailed view of Events 2L and C) 2R with unsmoothed proxy trends discussed in the text 2L 2R

1x105 2.5x105 4% 10% 2x105 4.5x105 6% 14% 0 160 300 600 0 120 100 600

Figure 3. A) High resolution images of upper core sections from LUN and RIP. Run depth (RD) in meters is displayed on the left of each section. XRD results are indicated by blue diamonds and blue text, with abbreviations for major and minor phases present

(C=calcite; Mg- C=magnesium calcite; cl=clay; D=dolomite; Gy=gypsum; Py=pyrite; Q=quartz; Br=brushite). Event layers are denoted by teal brackets on the right. Years in

188

red indicate the major earthquakes correlated to these event layers. Red whiskers from these ages indicate the error ranges associated with these years according to the age model. B) and C) are close-up detailed views of events 2L and 2R showing un-smoothed proxies: magnetic susceptibility (units (10-5 SI), XRF data (titanium, iron, sulfur, in counts per second) and percent organic matter that are related to the event layer formation and discussed in the text.

189

190

191

Figure 4. Multi-parameter plots of LUN (4a.) and RIP (4b.). Units for magnetic susceptibility (MS) are (10-5 SI), and sediment accumulation rates are cm/year. Elemental data from scanning XRF are reported at log10 (element counts/total kilo-counts per second). Loss-on ignition results are percent (%) organic matter and calcium carbonate, stable carbon isotopes of organic matter are in per mil (‰). Dashed horizontal lines indicate pollen zones identified by Mensing et al. (2015). Gray fields indicate event layers, and orange lines (CC) indicates the pick for the initiation of the Cava Clementina reclamation project and (M) indicates the modern reclamation project implemented during the Mussolini Era (1940’s).

192

CHAPTER 4 Evidence of palaeohydrological change in the Rieti Basin, Italy, from lake sediment core stable isotope analysis

Abstract

The organic and inorganic carbon fractions of sediment cores recovered from lakes

13 Lungo and Ripasottile (LUN and RIP) were analyzed using stable isotopes (δ Corg,

13 18 δ Ccarb, δ Ocarb) and bulk chemical parameters (C, N, and carbonate weight percent) to

understand past carbon cycling and hydrological change. Zones previously defined by

palynological and sedimentological characteristics were used to help interpret periods of

major shifts in the carbon data. The results show that the basal zone of LUN, representing

13 the oldest sediment recovered, had negatively correlated and oscillating δ Corg and C:N

values with low organic C concentration. These are indicative of a high degree of OM

degradation coupled with episodic influences from a terrestrial or macrophyte OM

source. The laminated zone, where both LUN and RIP exhibit similar sediment patterns

and values of all parameters, consists of OM mainly within-lake production. This period

13 was probably a merged single lake that occupied both depocenters. The δ Ccarb and

18 δ Ocarb values reflect either longer water residence time or primarily external DIC input.

During the LIA (Little Ice Age) zone, all parameters become highly variable and the

13 inverse correlation between δ Corg and C:N is disrupted. LUN and RIP values also

diverge, indicating hydrological separation. The modern zone, particularly the past 100

193

13 18 years, exhibit the highest Corg%, signifying eutrophication. The δ Ccarb and δ Ocarb values approach modern lake water in the most recent samples, suggesting shorter water residence times and the lowest lake levels of the past 2000 years.

Introduction

The change in surface and groundwater hydrology over time is a critical part in understanding the paleolimnology of lakes Lungo and Ripasottile (LUN and RIP) in the

Rieti Basin of Italy. This area has been the focus of studies on anthropogenic change and landscape evolution (Calderini et al. 1998; Mensing et al. 2015; 2016) and hydrologic change through the Pleistocene and Holocene (Ferelli et al. 1992; Calderoni et al. 1994;

Gliozzi and Mazzini 1998), but a clearer picture of the hydrological evolution over the past ~2000 years has not been explicitly studied. The type and amount of carbon deposited and preserved in lake sediment through time is widely used as a proxy of environmental change (Leng and Marshall 2004; Routh et al. 2004; Thevenon et al. 2012;

Lacey et al. 2016). Carbon system proxies have also been used to link hydrological factors to civilization shifts (Hillman et al. 2015; Douglas et al. 2015). Hydrologic balance and carbon cycle changes are an essential backdrop for past human-environment interaction studies.

A combined proxy approach is ideal when inferring system changes, especially in lake sediments that integrate multiple signals. To gain the most complete view into the

194

past, this study examined both the organic and carbonate fractions of sediment. Stable isotopes of carbon in organic matter (OM) are commonly used as proxies for lake paleoproductivity and/or organic carbon source. Different types of plants, namely algae,

C3 and C4 classes of terrestrial plants, and aquatic macrophytes have characteristic

13 isotope signatures (δ Corg) that can be utilized for OM source identification in sediment

(Meyers 1994; Cohen 2003). The weight ratio of organic carbon to total nitrogen (C:N) is

13 another proxy commonly used in conjunction with δ Corg to determine the origin of organic carbon. Aquatic algae and terrestrial vascular plants have large differences in the proportion of cellulose and proteins in their structure, yielding characteristic ratios of C:N

(Meyers and Lallier-Vergès 1998; Vreca and Muri 2006). Internal processes, such as OM decomposition and recycling, can also effect the amount and/or stable isotope composition of OM in the sediment (Meyers and Ishiwatari 1993; Cohen 2003; Khan et al. 2015).

The organic fraction of the sedimentary carbon pool is related to the inorganic, or carbonate, fraction by several processes. Out of the three main types of carbonate in lake sediments (authigenic, biogenic, detrital) authigenically precipitated carbonate stable isotopes offer the most integrated signal of whole-lake C and O dynamics (Cohen 2003).

The amount of authigenic carbonate precipitation depends on water column and sediment saturation state, redox state, and level of primary productivity. Down-core trends in the

13 18 amount of CaCO3 and stable isotopes (δ Ccarb, δ Ocarb) can yield information on past

195

carbon system dynamics. Isotopic variation, in particular, is indicative of lake water

balance, stratification, changes in water temperature and/or inflowing water source

(McKenzie 1985; Talbot 1990; Leng and Marshall 2004; Pueyo et al. 2011).

The stable isotopes of sulfur in sediment may also lend information on organic matter

recycling, as large isotope fractionations occur during dissimilatory sulfate reduction

(Nriagu and Soon 1985; Eimers et al. 2006). The extent of isotope fractionation has been

linked to lake salinity (Ryu et al. 2006;)and marsh extent, so changes in the bulk isotopic

34 composition of sedimentary sulfur (δ STS) may also lend information on hydrological

regime changes.

The goal of this work is to re-examine the major zones from LUN and RIP cores, previously defined as major shifts in the surrounding vegetation and/or regional climate, and to identify how the lake hydrological regime also transformed. This work expands on previous findings by Mensing et al. (2015; 2016) by providing stable isotope data from both LUN and RIP cores, and applying the correlated age model for RIP (Chapter 3). A hydrological scheme is proposed for the evolution of these lake systems over the past

~2700 years based on the experimental results.

Methods

CORING AND ZONE DELINEATION

196

A detailed description of core recovery, initial description, and correlation is included in Chapter 3. The stratigraphic zones in LUN compared here have the same boundaries as those Mensing et al. (2015), section 4.3 but the two zones at the base of

LUN (Transition Zone, Roman Era) have been combined and here are referred to as

“base”. The Medieval Period is referred to here as “laminated.” RIP zones were defined use the same ages as LUN, except for the base zone that is absent in the RIP core.

STABLE ISOTOPES

The composite cores of both lakes were sampled every ~10 cm for organic carbon

13C analysis. Dry, powdered sediment samples were moistened with DI water then placed

in a glass fumigation desiccator containing concentrated HCl for 30 days to remove

carbonates, following the method of Yamamuro & Kayanne (1995). The analyses were

then performed using a Eurovector EA 3000 elemental analyzer interfaced to a

Micromass IsoPrime stable isotope ratio mass spectrometer equipped with a helium

diluter using the method of Werner et al. (1999). Error was determined by running

duplicates of samples and acetanilide standards for δ13C. The standard deviation for each

duplicated sample was calculated, and the average isotopic and elemental concentrations

and corresponding standard deviations for the 12 acetanilide standards are: (δ13C) -

33.66±0.02, (weight % C) 71.09±0.96, and (weight % N) 10.36±0.15. The percent

carbonate (%CaCO3) was determined using the loss on ignition methods of Dean (1974)

197

and Heiri et al. (2001). Samples were weighed, dried at 100°C for 24 hours, then combusted at 550°C for four hours followed by combustion at 1000°C for two hours.

δ13C and δ18O values of carbonates from discrete depths were analyzed at the stable isotope laboratory at the University of California, Berkeley (UCB) using a

MultiCarb system in line with a GV IsoPrime mass spectrometer in Dual Inlet. Replicates of the international standard NBS19, and two lab standards (CaCO3 I & II) were analyzed along with the samples for each run. Long term external precisions for C and O are ±

0.05‰ and ± 0.07‰, respectively. Results are reported relative to the VPDB standard for both δ13C and δ18O.

Bulk powdered samples from select depths were also analyzed at the UCB stable isotope laboratory for δ34S of total sedimentary sulfur. Values were determined using the

SO2 EA-combustion-IRMS method using a GV IsoPrime isotope ratio mass spectrometer in line with the Eurovector Elemental Analyser (EuroEA3028-HT). The long-term external analytical precision is better than ±0.2‰ and the data are reported versus the

CDT standard. All isotopic measurements are presented in the delta (δ) notation:

13 12 13 12 δa = [Ra/Rstd]*1000 , where Ra is the C/ C ratio in the sample and Rstd is the C/ C ratio in the standard. All values are expressed as parts per mil (‰).

Results.

198

The downcore plots of LUN and RIP (Figure 1) show how the parameters assessed vary with depth and stratigraphic zone and are briefly summarized below. Discrete stable

34 isotope data sampled at core depths of interest are listed in Table 1, including δ STS,

13 18 δ Ccarb and δ Ocarb of LUN and RIP.

34 δ STS

34 The LUN δ STS values are positive for all depths sampled, ranging between 1.05 and

34 10.72‰ (Table 1). In RIP, δ STS reach as low as -22.86 and as high as +15.64‰. The

34 sampled depths of δ STS do not always coincide with depths of the event layers, but

34 during 2R and 4R, the values of δ STS (-10.61 and -7.13‰) are more negative relative to

34 the core average (-1.9‰). In LUN, the δ STS during 1L, 2L and 4L (+1.05, +1.09,

+2.14‰, respectively) are also more negative relative to the core average (+4.35‰).

13 ORGANIC CARBON (weight %Corg, C/N, and δ Corg):

The weight % Corg in LUN is highest in the core top (4.65%), with another prominent peak (4.48%) at 227 cm run depth (Figure 1). Values are low and relatively non-variable during the laminated interval and below. In RIP, weight % Corg is also relatively low and non-variable throughout the base of the core, with a rapid increase in the upper 50cm and apeak value (3.8%) in the core top. Ratios of organic carbon to nitrogen (C/N) at the base of LUN are variable, at times reaching values over 20. The

199

laminated interval C/N is low and considerably less variable, ranging from 6.92 to 8.4.

The highest value (21.4) in LUN occurs at 278 cm, when % Corg is also high (3%). In

RIP, values are generally lower, with a maximum value of 10. The largest variations are at the base and the top of the core, ranging from 6.3 near the base to 10.0 at the top of the

13 core. Values of δ Corg in RIP range between -26.3 and -25.6‰ at the base of the core and during the laminated interval. During the modern interval, values exhibit a greater range,

13 from -29.2 to -25.4‰. In LUN, the base of the core has highly variable δ Corg, ranging from -40.2 to 27.0‰. During the laminated interval, values were similar to those in RIP, ranging between -28.2 and -25.7‰. During the modern interval in LUN, values are relatively more negative and variable, ranging from -38.5 to -26.3‰.

13 CO-VARIATION OF δ Corg VERSUS C:N

13 The δ Corg versus C:N plots (Figures 2a. and b.) show how these two parameters co- vary in each core. The plots also include relative changes in %Corg as circle size. Fields indicate typical ranges for freshwater algae, aquatic macrophytes, and terrestrial vascular plants. In LUN, samples in the laminated zone all fall within the range of freshwater algae. Some from the base of the core also fall within this field, while most have higher

13 13 C:N and more negative δ Corg. The general trend in LUN is a negative correlation δ Corg and C:N (R2 =0.73). In RIP, all samples fall in the range of algal productivity, and the organic fraction is contrastingly stable in C:N values, which do not exceed 10.

200

18 13 STABLE INORGANIC ISOTOPES (δ Ocarb, δ Ccarb)

18 The values of δ Ocarb in RIP increased from -6.83‰ at the top of the core to values that raged between -2.98‰ to -5.79‰ throughout the core (Table 1, Figures 1, 3). The

18 base of LUN contains the most negative δ Ocarb of the core (-8.77‰) as well as a decreasing trend through this interval to the base. Values in the laminated interval in both

LUN and RIP exhibited less variability (±0.4) than in the modern interval (±1.35‰). In

18 LUN, δ Ocarb values were also lower at the core top (-7.83‰), though with little variation

(±0.26‰) through the modern. The laminated interval exhibits more enriched values, similar to RIP, and ranges from -4.15 to -2.98‰. After the cessation of laminations, both

13 18 13 lakes show a divergence in values of both δ Ccarb and δ Ocarb, with RIP δ Ccarb values becoming more enriched and LUN more negative in both isotopes (Figures 1 and 3).

13 LUN δ Ccarb values in the modern interval are generally more negative than in RIP (-

6.11‰ to -4.41‰, and -1.88‰ to -0.21‰, respectively). RIP values have a small range for the entire core (-0.17‰ to -1.88‰), while LUN values exhibit more variation with depth (-6.11‰ to +0.52‰).

Discussion.

ORGANIC MATTER SOURCES

201

13 The results of δ Corg and C:N in LUN and RIP indicate that carbon isotopes of OM serve as a proxy for OM source more than paleo-productivity. The negative correlation

13 between δ Corg and C:N in LUN demonstrates that for most of the record, changes in

13 δ Corg values are accompanied by changes in C:N, and these values are controlled by changing OM source to sediment (Meyers 1994). The lack of a paleoproductivity signal is

13 evidenced by the poor correlation between δ Corg and %Corg in both cores, where

R2=0.28 (Figures 1 and 2a, b). The Rayleigh-type isotope fractionation model used to

13 explain δ Corg enrichment during high productivity periods is not valid for LUN and RIP because the lakes are shallow and the epilimnetic DIC pool is renewed or recycled rapidly with frequent mixing compared to sediment accumulation rates (Cohen 2003).

All but the perimeters of RIP and LUN are presently phytoplankton dominated, and algae is an important source of OM. Additionally, macrophytes are abundant in the littoral zones and together with terrestrial plants, are important contributors of OM . Four of the most common non-algal plant species contributing to lake OM in Lazio, Italy, are

Alnus glutinosa, Phragmites australis, Quercus cerris, and Fagus sylvatica, all with

13 δ Corg ranging from -31.52 to -28.79‰ (Rossi et al. 2010). These values are within the

13 range of LUN and RIP values. The most negative δ Corg values of lake algae are found in chrysophytes and diatoms (-34.4 to -26.6 ‰), while cyanobacteria are the most variable

(-32.4‰ to -5.9‰) (Vuorio et al., 2006). Prevalent algae species in the LIA and modern sediment were Tetraedron minimum and Pediastrum (Mensing et al. 2015). These have

202

13 distinctly negative δ Corg values of -47.9± 0.1 and -40.1± 0.1‰ respectively. These ranges alone cannot identify the bulk OM source shifts among the different core zones, but together with C:N values the periods with major OM source changes in LUN can be identified (Figure 2a.).

There are two possibilities for the observed trend in LUN samples in the LIA and

13 base zones (Figure 2a). First, samples with more negative δ Corg and higher C:N could be

13 a mix of macrophytes and algae species with particularly low δ Corg and low C:N, such as

T. minimum. Second, the process of OM decomposition could be influencing these two

13 parameters. OM degradation results in sedimentary δ Corg negative excursions of 2-6‰

(Wilson et al. 2005; Oehlert and Swart 2014) as more labile OM is decomposed and refractory components remain (Meyers and Ishiwatari 1993). This occurs with increases in C:N of up to 6, as N is more readily lost during this process (Khan et al. 2015). Early diagenetic effects, displayed as grey arrows in Figures 2a. and 2b., may have concurrently altered the sedimentary OM of samples from the LIA and base (making a linear trend outside of the aquatic algae field in figure 2), possibly causing the negative correlation observed in LUN.

The concurrence of high C:N with a more pronounced diagenetic signature in LUN could be the sign of more OM contribution from a marsh belt surrounding the lake, where

OM amounts and decomposition rates are high. Shallow lakes or those with larger littoral zones or peripheral marsh have a greater contribution by macrophytes (Cohen 2003), so

203

the base and LIA zones could be periods marked by marsh expansion and/or lower lake level. The proximity of RIP to the Velino River may have allowed more through-flow throughout the LIA, preventing standing water and substantial marsh belt development.

CARBONATE DYNAMICS

13 18 The interpretation of δ Ccarb and δ Ocarb downcore hinges on the assumption that the carbonate was precipitated authigenically or diagenetically within the lake (Cohen

2003), because detrital carbonate retains its original isotopic signature and biogenic carbonate (carbonate shells or tests) does not always form in equilibrium with lake water.

LUN and RIP watersheds include higher elevation areas composed of carbonate rock, so there is a possibility that in the past, this carbonate was transported to the lakes. For detrital carbonate to be a significant portion, however, a high-energy river or stream must feed into the lake with sufficient energy for transport (Leng and Marshall 2004). It is more likely that in these cores, the surrounding bedrock contributes to the isotopic signature as dissolved inorganic carbon (DIC) from groundwater or river water flowing into the lakes. While the riverine input may have been more significant in the past than in modern day, it reaches the lakes after meandering through the low-relief basin and is therefore not considered a high energy environment. Another consideration is the differences in isotopic values of the common lacustrine carbonate phases: calcite,

204

aragonite, and dolomite (Talbot and Kelts 1986). XRD and smear slide results from these cores (Chapter 3), however, found low-Mg calcite to be the dominant form. Dolomite was a minor constituent at discrete depths, and aragonite was not present in significant

18 quantities at any depth. Dolomite δ Ocarb values may be +3‰ more enriched than calcite

(Land 1980; Rosen et al. 1995) when precipitated under the same conditions (at 25°C), so

18 the potential for this phase to mistakenly enrich bulk δ Ocarb values at certain depths in these cores must be considered.

In this work, the range of isotopic values of surrounding carbonate bedrock are used as an end member to determine relative contribution of external DIC input. The

13 other endmembers displayed in Figure 3, the modern lake water values of δ CDIC and

18 δ Owater-VSMOW, represent the signature of authigenic carbonate if precipitated in

18 equilibrium with modern lake water (Talbot 1990). The δ Ocarb of authigenic carbonate is not only dependent on the isotopic composition of the water (modern values for LUN of

18 and δ Owater-VSMOW are -7.0 to -5.3‰ and RIP are -8.2 to -7.7‰; Chapter 1), but also depends on lake water temperature during precipitation (Craig 1965; Kim and O’Neil

1997). Using the equation of Leng and Marshall (2004) adapted from Kim and O’Neil

18 (1997), the surface sample of LUN δ Ocarb precipitated from lake water at 21.5°C, which is within 1° of LUN water summer temperature measured in 2014 (Chapter 1). This confirms that carbonate is precipitating in equilibrium with the lake water. In RIP, the

18 same calculation for surface sediment δ Ocarb yielded a lake water temperature of 8.7°C,

205

or 9 degrees less than measured summer water temperatures. This could be due to water having a depleted oxygen isotope signature at the time of carbonate formation because increased precipitation or groundwater input, carbonate precipitation during a season other than summer, or disequilibrium effects caused by biogenic carbonate forming at the same time in the water column (Leng and Marshall 2004).

18 The primary and secondary controls of δ Ocarb values are likely the ratio of precipitation to evaporation or changes in the source of inflowing water. In the laminated

18 zone, the δ Ocarb of both lakes is enriched (Figures 1 and 3). This is may be explained by the presence of a single lake with larger water volume and surface area occupying the basin, allowing for longer residence time and more evaporative enrichment.

Alternatively, the primary source of inflowing water to the lakes could have switched during this period to surface water, instead of the groundwater input that dominated previously then again during the LIA zone. The surface water represents was likely direct river inflow containing DIC that is closer to the carbonate bedrock endmember (Figure

18 3). Ranges of calcite δ Ocarb precipitated authigenically in closed basin lakes are between

-7‰ to +5‰, while in open basin lakes are between -14‰ to 0‰ (Talbot 1990; Roberts

18 et al. 2008). The amplitude of δ Ocarb downcore variations in LUN and RIP are not large enough to identify periods of evaporative enrichment based on this proxy alone, though relative to the laminated interval, the base and modern zones of LUN, may have been more connected to groundwater with more contant through-cycling present. The

206

laminated interval, if hydrologically dominated by riverine connection as the sedimentology suggests, may only have had high inflow in certain years or seasons and lost water as evaporation as well as to groundwater outflux. In addition, the possibility of

18 control of δ Ocarb in LUN and RIP by water temperature (T) trends cannot be ruled out, though for these relatively short time scales (RIP= 1300, LUN=2700 years) the amplitude of T changes are likely too small (Lacey et al. 2016; Mensing et al. 2016).

13 The factors influencing δ Ccarb are lake productivity and the source and residence time of DIC (McKenzie 1985; Roberts et al. 2008). There is little fractionation between

13 13 13 lake water δ CDIC and the authigenic δ Ccarb (<1‰; Leng and Marshall 2004), so δ Ccarb is a good indicator of processes affecting lake water dissolved carbon. Because some of

13 18 the same factors control both δ Ccarb and δ Ocarb values, covariance is often observed

13 18 between δ Ccarb and δ Ocarb when lakes are hydrologically closed (Talbot, 1990; Leng et al. 1999). The lack of strong covariance in LUN and RIP (Figure 3), though, could be due to sample size and does not necessarily indicate hydrologically open conditions. Also, on these relatively short time scales, changes in the DIC pool often overprint trends in

13 evaporation (Li and Ku, 1997; Cohen 2003; Horton et al., 2016). The δ Ccarb values are

13 all heavier than lake water δ CDIC, (Figure 3), suggesting either varying inputs of heavy

DIC from dissolution of carbonate in the catchment, exchange with atmospheric CO2

(Horton et al. 2016), and/or influence from biological processes that preferentially remove light carbon (12C) from the lake water DIC pool (Rosen et al. 1995).

207

SULFUR ISOTOPES

34 The δ STS values of LUN and RIP vary significantly within each core and between the two lakes. While these data were initially collected to aid in interpretations of S dynamics related to paleoseismicity (Chapter 3), the limited number of samples and potential micro-scale effects (Louca and Crowe 2017) further complicated the story rather than clarifying it. The results can, however, support the divergence of sedimentary processes in LUN and RIP during the LIA, where differences between the two lakes in

34 34 δ STS values are up to 20‰ (Table 1). The extent of fractionation of δ STS during sulfate reduction between sulfate and sulfide is mostly dependent on the amount of sulfate available (Mayer and Schwark 1999; Ryu et al. 2006). In RIP during the LIA, the sulfate reduction pathway for organic matter mineralization was more prominent with more sulfate and/or more labile organic C available in the sediment. Greater amounts of reduced sulfide and metal-sulfides produced during this reaction recorded the greater

34 fractionations, and δ STS exhibit strong depletion (Nriagu and Soon 1985; Ding et al.

2016). The difference in fractionation between LUN and RIP may be because in LUN, more organic matter was available and sulfate reduction rates were more rapid, limiting fractionation (Böttcher and Lepland, 2000). The general tendency for more negative and

34 variable values of RIP δ STS than in LUN (Table 1) may be due to one or a combination of these scenarios. The extremely negative value of RIP in the most recent sample (-

22.86‰, 1998 CE, Table 4.1) was deposited when sulfate-rich Santa Susanna spring

208

water is known to flow directly into RIP (Chapter 1), suggests that high lake sulfate concentrations allow for more sulfate reduction and possibly a greater degree sulfur isotope fractionation. Alternatively, eutrophic conditions and high OM supply to the sediment encourages the sulfate reduction pathway as alternative electron acceptors are rapidly consumed.

Conclusions and Proposed Paleohydrological Scheme

A scenario for hydrological evolution of the system is offered based on these data.

The base of LUN represents the greatest marsh extent. During this time, macrophytes and

OM decomposition contributed to the OM pool episodically, evidenced by periods with

13 more negative δ Corg, high C:N and high CaCO3% . This zone of the core represents the longest time period (~1500 years), so could have captured oscillating wetter and drier climate periods when more or less of the marsh or littoral signal was preserved. RIP may also show the same pattern, though this zone is beyond depths successfully recovered during coring. The general decreasing trend in CaCO3% from the bottom of LUN to

~1200 cm RD represents rising lake level, where the oxic-anoxic boundary was higher and gradually more CaCO3 dissolution occurred in an anoxic hypolimnion.

The laminated interval represents a lake(s) with riverine inluence. The individual bands have distinct amounts of detrital and redox elements such as Ti, Fe and Mn (see

209

Figure 10 in Mensing et al. 2015), likely caused by annual or seasonal variability in riverine inflow. The carbon isotope proxies are the least variable in this interval, and

13 δ Corg and C:N values indicate organic sedimentation dominated by algae without marsh or littoral vegetation influence. The depocenter(s) may have also been farther from the lake edge than during the LIA and/or modern interval, so littoral OM and CaCO3 were not major components of the sediment during this interval. The carbonate stable isotopes are more indicative of DIC from the surrounding bedrock probably due to riverine inflow dominating the water budget and lower productivity, indicated by low OM% throughout this interval, caused less authigenic carbonate production. To further demonstrate how this interval was hydrologically distinct for both lakes and likely represented a period with surface inflow and outflow, supplementary figures 4S1 and 4S2 show a core photo from this zone in one core section from LUN and RIP with the downcore trends of major element constituents of the sediment (Ca, Fe, Mn, Ti) from the scanning XRF data (refer to Chapter 3 for XRF methods).

The LIA contains the most variable and complex signals in all proxies, suggesting this was the interval with the most rapid change in the lakes’ footprint and carbon cycling. The lakes’ surface water input was cut off intermittently, if not permanently, and the depocenters were separated. The lowering of lake level allowed for more communication with the marsh belt and OM transport, whether from catastrophic physical triggers (earthquakes, Chapter 3) or other high-energy events. The modern

210

interval begins abruptly but then is relatively less variable in LUN than in RIP. Water

level in RIP during the 1800’s was highly variable, probably due to continued input from

the Velino River. The 1900’s are marked by unprecedented increases in Corg%, as

eutrophication occurs in both lakes and water levels reach their lowest historical extent.

REFERENCES

Aichner, B., Herzschuh, U., & Wilkes, H. (2010). Influence of aquatic macrophytes on the stable carbon isotopic signatures of sedimentary organic matter in lakes on the Tibetan Plateau. Organic Geochemistry, 41(7), 706-718.

Bade, D. L., Carpenter, S. R., Cole, J. J., Hanson, P. C., and Hesslein, R. H. (2004) Controls of d13C DIC in lakes: Geochemistry, lake metabolism and morphometry. Limnology and Oceanography, 49(4), 1160-1172.

Barreiro-Lostres, F., Brown, E., Moreno, A., Morellón, M., Abbott, M., Hillman, A., Santiago Giralt, & Blas Valero-Garces, B. (2015). Sediment delivery and lake dynamics in a Mediterranean mountain watershed: Human-climate interactions during the last millennium (El Tobar Lake record, Iberian Range, Spain). Science of the Total Environment, 533, 506-519.

Böttcher, M. E., & Lepland, A. (2000). Biogeochemistry of sulfur in a sediment core from the west-central Baltic Sea: evidence from stable isotopes and pyrite textures. Journal of Marine Systems, 25(3), 299-312.

Calderini, G., Calderoni, G., Cavinato, G.P., Gliozzi, E., & Paccara, P. (1998) The upper Quaternary sedimentary sequence at Rieti Basin (central Italy): a record of sedimentation response to climatic changes. Palaeogeography, Palaeoclimatology, Palaeoecology 140, 97-111.

Calderoni, G., Carrara, C., Ferreli, L., Follieri, M., Gliozzi, E., Magri, D., Narcisi, B., Parotto, M., Sadori, L., & Serva, L., (1994) Palaeoenvironmental, palaeoclimatic and

211

chronological interpretations of a late Quaternary sediment core from Piana di Rieti (central Apennines, Italy). Giornale di Geologia 3 (56/2), 43–72.

Cohen A. S. (2003) Paleolimnology: The History and Evolution of Lake Systems. Oxford University Press, 528 pp. Craig, H., 1965. The measurement of oxygen isotope palaeotemperatures. In: Tongiorgi, E. (Ed.), Stable Isotopes in Oceanographic Studies and Palaeotemperatures. Pisa, Consiglio Nazionale delle Ricerche Laboratorio di Geologia Nucleare, pp. 161–182. Dean Jr., W.E., 1974. Determination of carbonate and organic matter in calcareous sediments and sedimentary rocks by loss on ignition: comparison with other methods. Journal of Sedimentary Petrology 44, 242-248. Ding, X., Li, D., Zheng, L., Bao, H., Chen, H. F., & Kao, S. J. (2016). Sulfur Geochemistry of a Lacustrine Record from Taiwan Reveals Enhanced Marine Aerosol Input during the Early Holocene. Scientific reports, 6. Douglas, P. M., Pagani, M., Canuto, M. A., Brenner, M., Hodell, D. A., Eglinton, T. I., & Curtis, J. H. (2015). Drought, agricultural adaptation, and sociopolitical collapse in the Maya Lowlands. Proceedings of the National Academy of Sciences, 112(18), 5607-5612.

Eimers, M. C., Paterson, A. M., Dillon, P. J., Schiff, S. L., Cumming, B. F., & Hall, R. I. (2006). Lake sediment core records of sulphur accumulation and sulphur isotopic composition in central Ontario, Canada lakes. Journal of Paleolimnology, 35(1), 99-109.

Ferreli, L., Brunamonte, F., Filippi, G., Margheriti, L., Michetti, A.M. (1992) Studi Geologici Camerti, vol. Speciale 1, 127-135.

Gliozzi, E & Mazzini, I. (1998) Palaeoenvironmental analysis of Early Pleistocene brackish marshes in the Rieti and Tiberino intrapenninic basins (Latium and Umbria, Italy) using ostracods. Palaeogeography, Palaeoclimatology, Palaeoecology 140, 325–333.

Heiri, O., Lotter, A., Lemcke, G., 2001. Loss on ignition as a method for estimating organic and carbonate content in sediments: reproducibility and comparability of results. Journal of Paleolimnology 25, 101-110.

Horton, T. W., Defliese, W. F., Tripati, A. K., & Oze, C. (2016). Evaporation induced 18 O and 13 C enrichment in lake systems: A global perspective on hydrologic balance effects. Quaternary Science Reviews, 131, 365-379.

212

Khan, N. S., Vane, C. H., & Horton, B. P. (2015). Stable carbon isotope and C/N geochemistry of coastal wetland sediments as a sea-level indicator. Handbook of sea-level research. Wiley, Oxford, 295-311.

Kim, S. T., & O'Neil, J. R. (1997). Equilibrium and nonequilibrium oxygen isotope effects in synthetic carbonates. Geochimica et Cosmochimica Acta, 61(16), 3461-3475.

Lacey, J., Leng, M., Höbig, N., Reed, J., Valero-Garcés, B. & Reicherter, K. (2016) Western Mediterranean climate and environment since Marine Isotope Stage 3: a 50,000-year record from Lake Banyoles, Spain. J Paleolimnol 55(2), 113–128. doi:10.1007/s10933- 015-9868-9

Leng, MJ, Roberts, N, Reed, JM & Sloane, HJ. (1999) Late Quaternary palaeohydrology of the Konya Basin, Turkey, based on isotope studies of modern hydrology and lacustrine carbonates. Journal of Paleolimnology.

Leng, M. J. & Marshall, J. D. (2004) Paleoclimate interpretation of stable isotope data from lake sediment archives. Quaternary Science Reviews 23, 811–83

Li HC, & Ku TL. (1997). Delta C-13-delta O-18 covariance as a paleohydrological indicator for closed-basin lakes. Palaeogeography Palaeoclimatology Palaeoecology 113: 69–80

Louca, S., & Crowe, S. (2017) Microscale reservoir effects on microbial sulfur isotope fractionation. Geochimica et Cosmochimica Acta 203, 117–139. Mayer, B. & Schwark, L. (1999). A 15,000-year stable isotope record from sediments of Lake Steisslingen, Southwest Germany, Chemical Geology, 161(1), 315-337. McKenzie, J.A. (1985) Carbon isotopes and productivity in the lacustrine and marine environment. In: Stumm, W. (Ed.), Chemical processes in lakes. Wiley, New York, pp. 99–118. Mensing, S. A., Tunno, I., Sagnotti, L., Florindo, F., Noble, P., Archer, C., Zimmerman, S., Pavón-Carrasco, F.J., Cifani, G., Passigli, S., Piovesan, G. (2015). 2700 years of Mediterranean environmental change in central Italy: a synthesis of sedimentary and cultural records to interpret past impacts of climate on society. Quaternary Science Reviews, 116, 72-94.

213

Mensing, S., Tunno, I., Cifani, G., Passigli, S., Noble, P., Archer, C., & Piovesan, G. (2016). Human and climatically induced environmental change in the Mediterranean during the Medieval Climate Anomaly and Little Ice Age: A case from central Italy. Anthropocene, 15, 49-59. Meyers, P.A., & Ishiwatari, R. (1993). Lacustrine organic geochemistry—an overview of indicators of organic matter sources and diagenesis in lake sediments, Organic Geochemistry, 20(7), 867-900. Meyers, P. A. (1994) Preservation of elemental and isotopic source identification of sedimentary organic matter. Chem. Geol. 114: 289–302.

Meyers, P. A. & Lallier-Vergès, E. (1999). Lacustrine sedimentary organic matter records of Late Quaternary paleoclimates. Journal of Paleolimnology 21: 345–372

Moreno, A., Valero-Garcés, B. L., González-Sampériz, P., & Rico, M. (2008). Flood response to rainfall variability during the last 2000 years inferred from the Taravilla Lake record (Central Iberian Range, Spain). Journal of paleolimnology, 40(3), 943-961.

Morettini, E., Santantonio, M., Bartolini, A., Cecca, F., Baumgartner, P. O., & Hunziker, J. C. (2002). Carbon isotope stratigraphy and carbonate production during the Early–Middle Jurassic: examples from the Umbria–Marche–Sabina Apennines (central Italy). Palaeogeography, Palaeoclimatology, Palaeoecology, 184(3), 251-273.

Nriagu, J. O., & Soon, Y. K. (1985). Distribution and isotopic composition of sulfur in lake sediments of northern Ontario. Geochimica et Cosmochimica Acta, 49(3), 823-834.

Oehlert, A. & Swart, P. (2014) Interpreting carbonate and organic carbon isotope covariance in the sedimentary record. Nature Communications 5, 4672.

Rosen, M. R., Turner, J. V., Coshell, L., & Gailitis, V. (1995). The effects of water temperature, stratification, and biological activity on the stable isotopic composition and timing of carbonate precipitation in a hypersaline lake. Geochimica et cosmochimica acta, 59(5), 979-990.

Pueyo, J. J., Sáez, A., Giralt, S., Valero-Garcés, B. L., Moreno, A., Bao, R., & Taberner, C. (2011). Carbonate and organic matter sedimentation and isotopic signatures in Lake Chungará, Chilean Altiplano, during the last 12.3 kyr. Palaeogeography, Palaeoclimatology, Palaeoecology, 307(1), 339-355.

214

Rossi, L., Costantini, M. L., Carlino, P., Di Lascio, A., & Rossi, D. (2010). Autochthonous and allochthonous plant contributions to coastal benthic detritus deposits: a dual-stable isotope study in a volcanic lake. Aquatic sciences, 72(2), 227-236.

Talbot M. R. (1990) A review of the palaeohydrological interpretation of carbon and oxygen isotopic ratios in primary lacustrine carbonates. Chemical Geology: Isotope Geoscience Section 80:261–279

Talbot, M. R., Kelts, K., (1986) Primary and diagenetic carbonates in the anoxic sediments of Lake Bosumtwi, Ghana. Geology 14 (11), 912-916

Thevenon, F., Adatte, T., Spangenberg J. E., & Anselmetti, F. S. (2012) Elemental (C/N ratios) and isotopic (15Norg, 13Corg) compositions of sedimentary organic matter from a high- altitude mountain lake (Meidsee, 2661 m a.s.l., Switzerland): Implications for Lateglacial and Holocene Alpine landscape evolution. The Holocene 22(10), 1135–1142

Van Dongen, BE, Schouten, S, & Damsté, J. (2002). Carbon isotope variability in monosaccharides and lipids of aquatic algae and terrestrial plants. Marine Ecology Progress 232:83-92.

Wilson, G. P., Lamb, A. L., Leng, M. J., Gonzalez, S., & Huddart, D. (2005). δ 13 C and C/N as potential coastal palaeoenvironmental indicators in the Mersey Estuary, UK. Quaternary Science Reviews, 24(18), 2015-2029.

Vreca, P., & Muri, G. (2006). Changes in accumulation of organic matter and stable carbon and nitrogen isotopes in sediments of two Slovenian mountain lakes (Lake Ledvica and Lake Planina), induced by eutrophication changes. Limnology and Oceanography, 51(2) 781- 790. Yamamuro, M., & Kayanne, H. (1995). Rapid direct determination of organic carbon and nitrogen in carbonate‐bearing sediments with a Yanaco MT‐5 CHN analyzer. Limnology and Oceanography, 40(5), 1001-1005.

215

Table 1. Stable isotope results from discrete sediment depths.

34 13 18 Depth (cm) Year (CE) δ STS (‰) δ Ccarb (‰) δ Ocarb (‰) LUN 23 1970 1.82 -4.41 -7.83 100 1843 -6.11 -7.27 143 1732 2.14 -1.66 -7.44 227 1574 -4.1 -7.58 240 1557 7.73 278 1494 7.21 327 1415 1.09 331 1408 10.72 -1.08 -4.15 370 1326 1.73 420 1256 1.05 542 1093 -0.31 -3.68 552 1085 -1.38 -3.78 562 1076 -1.06 -3.23 793 887 -0.21 -7.02 808 875 0.52 -7.15 815 867 3.4 1005 485 -1.76 -7.68 1216 -204 0.26 -7.57 1403 -633 0.28 -8.77 RIP 10 1998 -22.86 -1.88 -6.83 70 1913 1.94 100 1855 -.21 -4.47 140 1785 7.13 160 1754 -4.11 180 1723 -10.61 210 1676 1.51 -0.17 -4.12 306 1532 0.55 -0.49 -5.79 336 1492 15.64 387 1412 -7.13 407 1380 0.22 477 1286 -1.36 -2.98 478 1285 -1.7 -3.78 535 1259 -0.65 -3.44 536 1258 -1.57 -3.19

216

and

card C 13 δ org, C 13 δ

depth (cm). Zones are indicated by shadedindicated by depth Zones are (cm). - year (CE) and run and (CE) year show of axes both cores ). The y - ‰ . Downcore plots of organic carbon and carbonate parameters in LUN and RIP. LUN Units carboncarbonate inof. Downcore plots of parameters organic and are permil ( permil are

carb O 18

Figure 1 δ green=laminated,of top LUN. fields, aqua=LIA, togrey=modern, frombottom: purple=base 217

Figures 2a. and 2b. 13 Plots of C:N versus δ Corg in A) RIP and B) LUN. Fields indicate freshwater phytoplankton (black), freshwater macrophytes (dashed), and terrestrial C3 plants (grey)

218

typical ranges (Meyers1and Lallier-Vergès 1999; Aichner et al. 2010; Khan et al. 2015). Grey arrows indicate isotopic and C:N ratio shifts potentially induced by early diagenesis (Wilson et al. 2005; Khan et al. 2015). Circle diameter is relative to the weight percent organic carbon. The range in LUN is 0.8-4.7%, and in RIP is 0.7-3.8%. Scale for circle diameter is 1cm = 5% and 0.1cm= 0.5%.

18 13 Figure 3. Stable isotopes of oxygen (δ Ocarb) versus carbon (δ Ccarb) of carbonate in the different zones of LUN and RIP. Endmembers are also plotted, including the carbonate 13 18 18 bedrock and modern lake water (δ CDIC versus δ Owater). The values for δ Owater (average LUN and RIP values from Chapter 1) were converted from VSMOW, the standard for waters, to VPDB, the standard for carbonates, using the equation for fractionation between water and the mineral phase given by Leng and Marshall (2004) using summer water column temperatures measured in 2015. Shaded boxes indicate zones; blue is the LUN base zone, green is the laminated zone, and grey is the modern and LIA zones.

219

SUMMARY AND RECOMMENDATIONS FOR FUTURE WORK This work contributes a clearer understanding of how water enters Lungo and

Ripasottile (LUN and RIP), how this has changed through time and the major controls on water and sediment chemistry. Identifying potential groundwater earthquake responses and lake sediment seismites were unanticipated major findings. The conceptual model

developed in Chapter 1 provided a framework for subsequent chapters and their

conclusions. These data were particularly and unforeseeably useful as pre-conditions before the 2016-2017 seismic sequence in the central Apennines. During and immediately after the two largest-magnitude mainshocks (Aug. 24th and Oct. 30th), increases in physiochemical, isotopic and trace elements of spring water were measured, including the Santa Susanna and Peschiera springs. These discharge locations are at least

30 km from the mainshock epicenters, so the proposed earthquake effect on aquifer properties shows heavily faulted and fractured karstic aquifers may be particularly susceptible to earthquake related strain and/or shaking. There were heterogeneities in the observed responses, though, and these may yield important thresholds for epicentral distance, magnitude, or mainshock placement within the seismic series for effects on karstic aquifers. These responses also suggest that far-reaching earthquake effects may have impacted lakes LUN and RIP physically and/or chemically during major past earthquakes.

220

With this motivation, the background work for Chapter 3 began by compiling a list of

earthquake effects in lakes their sediments worldwide and a list of major past earthquakes

in the central Apennines surrounding the Rieti Basin. Then, reinvestigation of sediment

cores from both lakes led to the discovery of potential earthquake signatures that

correspond to years with earthquakes of Mw>6.0 within 30 km of the basin. These four

seismites in LUN and RIP could add novel instances of shallow or small lakes recording

paleoseismicity. The alternative explanations for the anomalous beds could be other

catastrophic events, such as floods or abrupt human landscape alteration. Or, these may be points where gradual limnological change reached a tipping point, leading to a new depositional regime and chemical signatures. The likeliness of an earthquake trigger hinges on coevality in the two lakes, yet the age model for RIP is not as well constrained as in LUN. With more refinement of the core chronology, the alternative hypotheses may be ruled out in favor of the presence of seismites in these lakes.

This results and inferences included in Chapter 4 offer a scenario for hydrological

regime evolution in the Rieti Basin lakes over the past ~2000 years. The entire core

records form both lakes were studied to determine how the lakes’ level, extent, connection with groundwater and surface water has evolved. This stable isotopic and bulk geochemical investigation of organic and inorganic C fractions in the LUN and RIP cores yielded a refined history of the hydrological change that expands on past work that focused on the paleoecological changes over the same time period (Mensing et al. 2015;

221

2016). The results suggest that 2000 years ago one lake with larger surface area and an

extensive surrounding wetlands occupied the lakes’ footprint. This environment

transitioned to a higher flow-through depositional center with low productivity during the

medieval time period, where both lakes exhibit prominent laminations and low amounts

of organic matter. During the Little Ice Age, the lakes were periodically separated and

had more groundwater influence that also may have been highly impacted by earthquake

events. Finally, entering the modern period human influence increased to a point where

efforts to control water flow and levels in the basin overprints many of the lake evolution

proxies. The last ~50 years of lake history marks the highest degree of hydrological

regulation and lake eutrophication.

Future considerations

Future work could expand on the data collected and may gain a better understanding of the hydrogeology and associated processes. Specifically, a longer time series for all lake and spring water sample collection could better capture variability in the measured chemical parameters. The natural variability in chemical concentrations and isotopic compositions may not have been adequately characterized with only one sampling per season over the two year sampling period. Climate change is influencing precipitation patters throughout the Mediterranean, and decreases in amount are predicted for central

Italy (Dragoni and Sukhija 2008; Brunetti et al. 2012). Groundwater resources are

222

susceptible to decreases in supply (Taylor et al. 2013), especially the springs sampled in

this work that source aquifers with large and increasing water demand (Fiorillo and

Guadagno 2010; Mastorillo and Petitta 2014). Collecting and analyzing water samples over a longer time scale may capture how aquifer storage may be changing over time and if this affects groundwater chemistry. Comparing this range to the pre-and post- earthquake range can confirm a seismic cause for chemical changes and also may help predict future response. Another addition to the work in Chapter 2 is adding flow data for the studied springs. An example of this is in progress, with sensors that were installed post-seismic sequence in area wells that measure water level changes and can be used to infer flow change. Depending on the results of this pilot data collection, an array of these sensors in groundwater wells throughout the seismically active central Apennines could be crucial in establishing the timing and magnitude of response.

The stable isotope analysis of the sediment cores is another area that could be

13 expanded and improved upon in the future. The stable isotopes of carbonate (δ Ccarb and

18 δ Ocarb) should be analyzed at more depths and at regular intervals throughout the cores.

13 34 The isotope analyses of organic C and S, δ Corg and δ STS, were measures of the total

organic and total sulfur fractions, respectively, but more advanced methods exist where

specific components of these fractions are isolated and analyzed. These compound-

specific isotope analyses yield detailed information on the sources of these components to

sediment (Cohen 2003; Aichner et al. 2010) as well as processes within the sediment

223

(Fang et al. 2014). For organic C, the analysis of specific lipid biomarkers, particularly n-

alkenes and fatty acids, can distinguish among algal, bacteria, fungal, terrestrial and

macrophytic organic inputs (Cohen 2003; Muri et al. 2004; Glaser and Zech 2005; Fang

et al. 2014). For the sedimentary S pool, the largest fractionations occur with changes in

redox state, often accompanied by microbial activity. If the downcore series of δ34S is

expanded in the future, these data would be most useful if a multiple-S isotope is used so

that the different biogeochemical pathways for sediment S cycling may be studied

(Canfield et al. 2010; Pellerin et al. 2015). Trends or major changes in S biogeochemistry

would be useful in expanding the work described in Chapter 3, where paleoseismic

inferences were based on S in the sediment.

In addition to the isotopic analysis of S, determining the S speciation and oxidation

state of the sediment would add a fascinating piece to this paleolimnological story. A

relatively new technique, synchrotron based soft X-ray absorption spectroscopy

(XANES) can directly determine S speciation and quantify the different S pool

proportional changes with depth (Prietzl et al. 2010; Morgan et al. 2012; Couture et al.

2016). This non-invasive technique might provide a record of S cycling that could help answer outstanding uncertainties involving sulfur source and biogeochemistry. The expansion of this work using these newer techniques combined with adding another lake sediment core record for another lake in the seismically active central Apennines would extend the lacustrine paleoseismic portion of this dissertation in an interesting direction

224

and could add information to the Italian paleoseismic database where historical written records do not exist.

225

References Aichner, B., Herzschuh, U., & Wilkes, H. (2010). Influence of aquatic macrophytes on the stable carbon isotopic signatures of sedimentary organic matter in lakes on the Tibetan Plateau. Organic Geochemistry, 41(7), 706-718. Brunetti, M., Caloiero, T., Coscarelli, R., Gullà, G., Nanni, T., & Simolo, C. (2012). Precipitation variability and change in the Calabria region (Italy) from a high resolution daily dataset. International Journal of Climatology, 32(1), 57-73. Canfield, D. E., Farquhar, J., & Zerkle, A. L. (2010). High isotope fractionations during sulfate reduction in a low-sulfate euxinic ocean analog. Geology, 38(5), 415-418. Couture, R.-M., Fischer, R., Cappellen, P., & Gobeil, C. (2016). Non-steady state diagenesis of organic and inorganic sulfur in lake sediments. Geochimica et Cosmochimica Acta, 194, 15–33. Dragoni, W., & Sukhija, B. S. (2008). Climate change and groundwater: a short review. Geological Society, London, Special Publications, 288(1), 1-12. Fang, J., Wu, F., Xiong, Y., Li, F., Du, X., An, D., & Wang, L. (2014). Source characterization of sedimentary organic matter using molecular and stable carbon isotopic composition of n-alkanes and fatty acids in sediment core from Lake Dianchi, China. Science of the Total Environment, 473, 410-421. Fiorillo, F., & Guadagno, F. M. (2010). Karst spring discharges analysis in relation to drought periods, using the SPI. Water resources management, 24(9), 1867-1884. Glaser, B., & Zech, W. (2005). Reconstruction of climate and landscape changes in a high mountain lake catchment in the Gorkha Himal, Nepal during the Late Glacial and Holocene as deduced from radiocarbon and compound-specific stable isotope analysis of terrestrial, aquatic and microbial biomarkers. Organic Geochemistry, 36(7), 1086-1098. Mastrorillo, L., & Petitta, M. (2014). Hydrogeological conceptual model of the upper Chienti River Basin aquifers (Umbria-Marche Apennines). Italian Journal of Geosciences, 133(3), 396-408. Morgan, B., Burton, E., & Rate, A. (2012). Iron monosulfide enrichment and the presence of organosulfur in eutrophic estuarine sediments.Chemical Geology, 296, 119–130

226

Muri, G., Wakeham, S. G., Pease, T. K., & Faganeli, J. (2004). Evaluation of lipid biomarkers as indicators of changes in organic matter delivery to sediments from Lake Planina, a remote mountain lake in NW Slovenia. Organic Geochemistry, 35(10), 1083-1093. Pellerin, A., Bui, T. H., Rough, M., Mucci, A., Canfield, D. E., & Wing, B. A. (2015). Mass-dependent sulfur isotope fractionation during reoxidative sulfur cycling: A case study from Mangrove Lake, Bermuda. Geochimica et Cosmochimica Acta, 149, 152-164. Prietzel, J., Thieme, J., Salomé, M. (2010). Assessment of sulfur and iron speciation in a soil aggregate by combined S and Fe μ-XANES: microspatial patterns and relationships. Journal of Synchrotron Radiation 17, 166–172. Taylor, R.G., Scanlon, B., Döll, P., Rodell, M., Van Beek, R., Wada, Y., Longuevergne, L., Leblanc, M., Famiglietti, J.S., Edmunds, M. and Konikow, L. (2013). Ground water and climate change. Nature Climate Change, 3(4), p.322.