0

5

10

15

20 0.0

0.2

0.4

0.6

0.8

1.0

11/26/12 0:00 11/27/12 0:00 11/28/12 0:00 11/29/12 0:00 11/30/12 0:00 12/1/12 0:00

Pre

cip

itat

ion

(m

m)

Stre

amfl

ow

(m

m)

Streamflow

Precipitation

Seasonal isotope hydrology of a coffee agroforestry watershed in Costa Rica

J. Boll1, K. Welsh*1,2, O. Roupsard2,3 1University of Idaho, Moscow, ID, USA; 2Centro Agronómico Tropical de Investigación y Enseñanza (CATIE), Turrialba, Costa Rica; 3UMR Eco&Sols, CIRAD, Montpellier, France

H21F-0794

Introduction

In the tropics, the “amount effect” has been

recognized as one of the primary influences on

stable isotope values in tropical precipitation for

monthly samples (Figs 1 and 2). However, when

examining event-based precipitation events at a

regional scale, lifting condensation levels and

surface relative humidity have a greater influence

(Sánchez-Murillo et al. 2014). The main goals of

this research were to examine what microclimate

factors influence local stable isotope (δ18O and δD)

values and whether stable isotopes can be used to

trace seasonality in this region at different temporal

scales. Methods

• Sampled precipitation (event-based samples),

groundwater (weekly samples), and stream water

(weekly samples at 4 locations) for 2+ years.

• Micrometeorological data were collected at an

eddy-flux tower located on site.

• Study Site: The Mejías watershed (Cartago,

Costa Rica) is an agroforestry microwatershed

(0.9 km2) located on the Aquiares Coffee Farm.

Results

Acknowledgments This work was funded by an NSF IGERT (Integrative

Graduate Education and Research Traineeship) and a U.S.

Borlaug Fellowship in Global Food Security. The

CoffeeFlux project provided site infrastructure and the

Aquiares Farm provided site access. The authors would

like to thank Alvaro Barquero, Alejandro Barquero, Alexis

Pérez, and Amilkar Moncada for field assistance.

-15

-10

-5

0

5

Sep-11 Mar-12 Oct-12 Apr-13 Nov-13

δ1

8O

(‰

)

Monthly δ18O values: Study data

MSD MSD

Fig. 5. Monthly averaged values for δ18O in precipitation at the study site. Precipitation isotope values were correlated with air temperature, wind speed, and wind direction.

Fig. 6. The Local Meteoric Water Line (LMWL) for precipitation samples (collected between Aug 2011 and Dec 2013) for the watershed.

y = 8.4987x + 18.02

-160

-140

-120

-100

-80

-60

-40

-20

0

20

40

-20 -15 -10 -5 0

δD

(‰

)

δ18O (‰)

LMWL Precipitation

Groundwater

Streamwater

-15

-10

-5

0

5

Jan-02 Apr-02 Jul-02 Oct-02 Feb-03 May-03 Aug-03 Dec-03 δ

18O

(‰

)

Monthly δ18O values: Historic GNIP data

MSD MSD

Dry Dry

Rainy Rainy

Dry Dry

Rainy Rainy

Fig. 4. Monthly averaged values of δ18O in precipitation from historic GNIP database for Turrialba. MSD = Mid-Summer Drought.

F. Gomez-Delgado et al.: Modelling hydrological behaviour of coffee agroforestry basin in Costa Rica 371Figure 1

Fig. 1. (a) Location of Reventazon river basin in Costa Rica, Central America. (b) Position of experimental basin inside of Reventazon

basin. (c) The “Coffee-Flux” experimental basin in Aquiares farm and its experimental setup to measure the water balance components.

In the present study we attempted to couple two lumped

models into a new integrative one, chosen to be scalable

and parsimonious: a basin reservoir model similar to the

CREC model (Cormary and Guilbot, 1969) and employing

the Diskin and Nazimov (1995) production function as pro-

posed by Moussa et al. (2007a,b), and the SVAT model pro-

posed by Granier et al. (1999). While the basin model was

considered appropriate for its simplicity and capacity to sup-

port new routines, the SVAT model was chosen for its par-

simony (three parameters in its basic formulation), its ro-

bustness (uses simple soil and stand data in order to pro-

duce model runs for many years, avoiding hydraulic param-

eters that are difficult to measure and scale up), its ability

to quantify drought intensity and duration in forest stands,

and for its successful past validation in various forest stands

and climatic conditions, including tropical basins (Ruiz et al.,

2010).

This paper aims to explain and model the hydrological

behaviour of a coffee AF micro-basin in Costa Rica, as-

sessing its infiltration capacity on andisols. The method-

ology consists of experimentation to assess the main water

fluxes and modelling to reproduce the behaviour of the basin.

First, we present the study site and the experimental design.

Second, we develop a new lumped eco-hydrological model

with balanced ecophysiological/hydrological modules (that

we called Hydro-SVAT model). This model was tailored to

the main hydrological processes that we recorded (stream-

flow, evapotranspiration, water content in the non-saturated

zone and water table level) and that are described in the sub-

sequent sections. Third, we propose a multi-variable calibra-

tion/validation strategy for the Hydro-SVAT model so we cal-

ibrate using the streamflow in 2009 and validate using the re-

maining three variables in 2009 and the streamflow in 2010.

Fourth, we make an uncertainty analysis to produce a confi-

dence interval around our modelled streamflow values, and

a sensitivity analysis to assess from which parameters this

uncertainty might come. Fifth, we attempt to simplify the

initially proposed model based on the results of the sensitiv-

ity analysis. Finally, we discuss the main findings concerning

the water balance in our experimental basin.

2 The study site

2.1 Location, soil and climate

The area of interest is located in Reventazon river basin, in

the Central-Caribbean region of Costa Rica (Fig. 1a and b).

It lies on the slope of the Turrialba volcano (central vol-

canic mountain range of the country) and drains to the

Caribbean Sea. The Aquiares coffee farm is one of the

largest in Costa Rica (6.6 km2), “Rainforest AllianceTM”

certified, 15 km from CATIE (Centro Agronomico Tropical

de Investigacion y Ensenanza). Within the Aquiares farm,

we selected the Mejıas creek micro-basin (Fig. 1c) for the

“Coffee-Flux” experiment. The basin is placed between the

www.hydrol-earth-syst-sci.net/15/369/2011/ Hydrol. Earth Syst. Sci., 15, 369–392, 2011

Fig. 1. Location of study site. Source: Gómez-Delgado et al. 2011

References Gómez-Delgado, F.., et al. 2011. Modelling the

hydrological behaviour of a coffee agroforestry basin in

Costa Rica. Hydrology and Earth System Sciences 15(1):

369-392.

Sánchez-Murillo, R., et al. 2014. Spatial and temporal

variation of stable isotopes in precipitation across Costa

Rica: An analysis of historic GNIP records. Open Journal of

Modern Hydrology 3: 226-240.

Fig. 9. Hourly values of δ18O (‰) measured in stream water (11/26/12 – 12/1/12), compared with stream flow and precipitation levels. Both δ18O and δD are positively correlated with precipitation amount (p=0.02 and 0.00005).

Discussion

• For monthly averaged samples (Figs 4 and 5),

dry season precipitation is enriched with respect

to the rainy season, but at an hourly or daily

basis other trends are evident, such as

correlation with meteorological measurements

and stream response to precipitation events (Fig

9).

• Air temperature, wind speed, and wind direction

are all significantly correlated with δ18O and δD

values in precipitation, which could be due to

seasonal differences in air mass sources.

• Examining stream water δ18O and δD values at a

finer scale (Fig. 9) shows how streams respond

to precipitation, as the stream becomes more

enriched with higher amounts of rain due to

source contribution of groundwater and overland

flow from spring and roads (Figs. 10 and 11).

• Stable isotopes are useful for examining

seasonality and watershed dynamics when

examined at different temporal scales. For

example, measuring isotopes in stream samples

during storm events assists in our understanding

of when storm flow reaches streams.

• Further work will include partitioning stream

water sources and quantifying spring contribution

to storm flow.

-9

-8

-7

-6

-5

-4

-3

3/1/12 6/1/12 9/1/12 12/2/12 3/4/13 6/4/13 9/4/13 12/5/13

δ1

8O

(‰

)

δ18O in groundwater WTL-1 WTL-2 WTL-4 WTL-5

Fig. 7. δ18O values in groundwater (weekly samples) at four groundwater wells: WTL-1 and WTL-2 (low elevations), WTL-4 (mid-elevation), and WTL-5 (high elevation).

*Contact: Kristen Welsh wels8152@vandals.uidaho.edu

-10

-9

-8

-7

-6

-5

-4

-3

-2

-1

0

2/1/12 6/1/12 10/1/12 1/31/13 6/2/13 10/2/13 2/1/14

δ1

8O

(‰

)

δ18O in stream water Flume

Lower

Middle

Upper

Fig. 8. δ18O values in stream water (weekly samples) at four locations in the watershed. Values generally become enriched in δ18O from upper elevations to lower elevations.

Fig 2. The land cover at the

site is dominated by Coffea

arabica coffee plants

interspersed with Erythrina

poeppigiana shade trees.

-7.3

-7

-6.7

-6.4

-6.1

-5.8

δ1

8O

(‰

)

δ18O, streamflow, and precipitation (hourly) O18

Fig 3. The study site is

part of the global

network of FLUXNET

micrometeorological

sites.

Storm flow Low flow

Fig. 11. Low flow conditions compared with storm flow conditions at the flume. Fig. 10. Intermittent springs (a) and road and path runoff (b) at the

site contribute significantly to storm flow.

a b