Spatiotemporal variability of stable isotopes in precipitation and stream water in a high elevation tropical catchment in the Central Andes of Colombia

The Colombian Andean Mountains include the headwaters of the main basins of the country. However, the isotope composition of water in these high mountain ecosystems has been poorly studied. In this study, we analysed the first set of stable isotope data collected along a wide elevation range (2600–4950 m a.s.l.) in the Central Andes of Colombia. The stable isotope composition of stream water and precipitation was determined for a period between 2017 and 2018 in the Upper Claro River basin. The driving factors influencing the spatial and temporal variability of δ2H, δ18O, and d‐excess were identified, and compared with daily air temperature and precipitation data from seven meteorological stations. The local regression line was described by δ2H = 8.2 δ18O + 12.3, R2 = 0.98. The δ2H and δ18O values showed more depletion in heavy isotopes, and the d‐excess values were more negative during the rainy season. An altitude effect of −0.11‰/100 m and −0.18‰/100 m was estimated for stream water and precipitation δ18O values, respectively, with the latter showing non‐linear behaviour. The dataset was compared with Colombian stations of the Global Network of Isotopes in Precipitation database, and a back‐trajectory analysis of air masses was conducted and compared with the d‐excess values. The δ18O weighted mean values changed with respect to the position in the Central Andes, indicating contrasting altitude effects depending on the moisture sources. The most positive d‐excess values were attributed to moisture recycling enhanced by local ecosystem conditions and the origin of precipitation from the Amazon basin, which change during the year and across the northern Andes. The results showed a high level of variation because of differences in elevation, seasonality, and atmospheric circulation patterns during the year. This study contributes to knowledge of spatial and temporal isotope composition data in the northern Andes, delineation of water supply basins, and to the definition of ecosystem boundaries in the high mountains of Colombia.


| INTRODUCTION
The interaction between environmental factors and the water cycle has been of increasing concern during recent decades. Climate change and anthropogenic factors are posing threats to water availability on local and global scales (IPBES, 2019;IPCC, 2022). The analysis of the stable isotopes in water (deuterium: 2 H; oxygen-18: 18 O) is essential to enable understanding of the factors affecting hydrological processes, and consequently the impacts of environmental change on local and regional moisture sources and water recharge areas. Understanding of the complex relationships between isotope signatures and meteorological, hydrological, geological, and land cover characteristics is particularly important in high elevation catchments including in the northern Andes, encompassing Ecuador, Colombia, and Venezuela. In this area water availability strongly depends on stream water from high mountain ecosystems including glaciers, páramos, cloud forest, and Andean forest (Buytaert et al., 2006;Escobar et al., 2022;Harden et al., 2013;Rodríguez-Morales et al., 2019).
Because of their relevance to the water cycle, páramo ecosystems have been widely studied (Hofstede, 2003;Sarmiento et al., 2013;Sklenar et al., 2010;van der Hammen & Cleef, 1986). The Andean páramos comprise a large variety of lakes, peat bogs, and wet grasslands where the soil can be up to several meters thick (Buytaert et al., 2006). In these environments the cold and wet conditions have favoured organic matter accumulation, which along with the input of volcanic ash has resulted in organic soils having porous structure (Buytaert et al., 2005), high water retention, and high hydraulic conductivity (Nanzyo et al., 1993;Rousseaux & Warkentin, 1976;Shoji & Fujiwara, 1984;). Páramos are known to have higher levels of precipitation than runoff and evapotranspiration (Buytaert, 2004); this is reflected in a sustained base flow and extreme water regulation capacity (Hofstede, 1995;Luteyn, 1992;Podwojewski et al., 2002;Poulenard et al., 2003). However, despite the importance of these ecosystems for Colombia, stable isotope investigations in páramos have been mainly conducted in Ecuador, Bolivia, and Perú (Bershaw et al., 2016;Fiorella et al., 2015;Valdivielso et al., 2020).
The spatial and temporal distribution of δ 2 H and δ 18 O isotopes in South America has been related to four main factors (Rozanski & Araguas-Araguas, 1995) including: (1) water input from three atmospheric moisture sources (the Atlantic and Pacific oceans and the Caribbean Sea); (2) the Andes Mountains blocking free air flow and leading to enhanced condensation along the slopes; (3) the proximity of the world's largest source of continental evapotranspiration (the Amazon Basin); and (4) the seasonal movement of the Intertropical Convergence Zone (ITCZ). The Amazon rainforest contributes large amounts of recycled moisture to the air masses transported by trade winds (Ampuero et al., 2020;Martinelli et al., 1996;Vallet-Coulomb et al., 2008). These conditions (mainly temperature and vapour, but also topographical conditions) determine the stable water isotope patterns in the hydrological cycle. Precipitation effects (including continental, seasonal, and altitude effects) cause predictable stable isotope compositions for local sites, landscapes, and regions. Additionally, a second order isotope parameter, termed deuterium excess (d-excess; δ 2 H-8 δ 18 O; Dansgaard, 1964), represents the y-intercept from the Global Meteoric Water Line (GMWL) (Bershaw et al., 2020;Gonfiantini, 1998), and characterizes the non-equilibrium fractioning during evaporation in the source area, temperature of condensation, and the degree of moisture recycling (Froehlich et al., 2008;Jouzel & Merlivat, 1984;Liotta et al., 2006;Rank & Papesch, 2005). On a global basis, relative humidity, sea surface temperature, and wind speed are the main controlling parameters (Benetti et al., 2014;Clark & Fritz, 1997;Rozanski et al., 1993).
Consistent with these factors, regional models of stable isotopes in precipitation over the tropical Americas have shown that variations in precipitation depend on both local meteorological factors and the interaction of several regional patterns, mainly the El Niño Southern Oscillation (ENSO) and the South American Summer Monsoon (SASM) (Vimeux et al., 2005;Vuille et al., 2003Vuille et al., , 2012Vuille & Werner, 2005). In the tropical Andes specifically, atmospheric circulation patterns varying in direction during the austral winter and austral summer also influence the isotope composition of atmospheric water vapour (Samuels-Crow et al., 2014;Villacis et al., 2008). For example, it has been suggested that stable isotope values for snow in the austral summer in the Peruvian Andes reflect regional-scale atmospheric forcing, but during the austral winter they are modulated by temperature, heating related to topography, or wind regimes on the ice cap surface (Hurley et al., 2015).
The spatial variability in isotope composition in the equatorial region is also dominated by the altitude effect. The δ 18 O composition in precipitation changes with elevation at a rate of À0.2‰/100 m (up to 3000 m a.s.l.), and declines above 3000 m a.s.l. to À0.5‰/100 m (Rozanski & Araguas-Araguas, 1995). A similar effect has also been observed for stream waters, where the isotope composition is generally modified by surface water mixing and/or evaporation (Bershaw et al., 2016). In high mountain ecosystems, including the páramos of Costa Rica and Ecuador, the isotope composition in precipitation has been found to be influenced by Amazon rainforest air masses during passage of the ITCZ in wetter seasons, and by orographic precipitation during the transition and drier seasons (Esquivel-Hernández et al., 2019).
In the Colombian Andes there is an absence of continuous and well-distributed stable isotope data, and the driving factors affecting stable isotope composition are still poorly understood. Only 11% (four) of the stations available in the Global Network of Isotopes in Precipitation (GNIP) have samples for at least five consecutive years (60 samples) been collected, and only one includes data up to 2021.
Stable isotope data have been commonly collected during time-limited campaigns, and consulting projects where data acquisition terminates with the cessation of funding. In most cases the data have remained unpublished (Vuille, 2018). Several stable isotopes studies have been conducted in different hydrological systems and using various investigative techniques. Based on 800 samples from a number of studies, Rodríguez (2004) developed a local meteoric water line (LMWL) equation for Colombia: δ 2 H = (8.03 ± 0.28) δ 18 O + 9.6, and reported that the δ-values during the dry season were higher than during the rainy season, and that they were more enriched near the oceans than in the inner part of the country, because of the lower elevation and proximity to the coast. The lowest δ-values did not coincide with the highest precipitation rates, which was explained by the various sources of rain fronts affecting the central part of the country (the Atlantic Ocean through either the Amazon basin or the Caribbean, and the Pacific Ocean).
Other stable isotope studies focused on groundwater exploitation in Colombia have mainly focused on the identification of recharge areas and residence times in the Eastern Andes (Castrillon et al., 2003;Mariño-Martinez et al., 2018); the Caribbean region (Campillo et al., 2021;Toro et al., 2009); the Cauca River (Betancur & Palacio, 2007;Medina et al., 2009), and the areas surrounding the Upper Claro River basin (UCRB) (Otálvaro et al., 2009). For instance, Saylor et al. (2009) found that the precipitation amount exerts the greatest effect on δ 18 O and δ 2 H at the GNIP stations in the vicinity of Bogotá. Nonetheless, the scope of these studies has been spatially and temporarily limited, and studies directed at understanding the driving factors that affect the stable isotope composition across the complex orography and climatic setting of the country have not yet been undertaken. In this report we provide the first stable isotope dataset for the area, derived by the Colombian Institute for Hydrology, Meteorology and Environmental Studies (IDEAM), as a contribution to the study of high mountain ecosystems highly threatened by environmental and anthropogenic change.
The dataset enabled analysis of spatiotemporal variations in the isotope composition of stream water and precipitation in the UCRB.
For this purpose the leading questions were: (i) how did the stable isotope values for precipitation and stream water vary in the UCRB over the period (2017-2018); (ii) how did the stable isotope composition vary as a function of elevation in the UCRB, and how was this related to other geographical regions of the country; and (iii) what were the main sources of precipitation during the 2017-2018 period in the UCRB, and how are moisture sources and land cover characteristics of the northern Andes related to the observed isotope patterns? 2 | MATERIALS AND METHODS

| Location and climate of the Upper Claro River basin
Colombia is located in the northwest of South America, and its land area extends to the Caribbean Sea in the north, the Pacific Ocean in the west, and the Amazon rainforest in the south. Three mountain ranges dominate the topography, including the Western, Eastern, and Central Cordilleras (Figures 1 and 2). The UCRB is a high elevation tropical catchment (65 km 2 ) in the Colombian Central Andes. It extends topographically from 2600 to 4995 m a.s.l., and provides water for industry, agriculture, and human consumption over a large area. There are seven meteorological stations in the basin over an elevation range of 2714-4699 m a.s.l. (Figure 2), and these are used to study the short hydrological response of the Conejeras glacier at the summit of the UCRB (Ceballos et al., 2006;Morán-Tejeda et al., 2018) ( Table 1). Mölg et al. (2017) have reported that this tropical glacier (0.2 km 2 ) is important in recording and understanding climate variability, as it plays a major role in modulating the arrival of water from the summit areas to the páramo (Morán-Tejeda et al., 2018), despite comprising only 1.3% of the watershed (WGMS, 2017). However, we were not able to investigate the effect of this glacier on the isotope composition, as snow accumulation on the glacier takes place over very short periods, and the snowmelt water resulting from thaw of the permanent glacier is always mixed with precipitation in the form of runoff (as found by analyzing the water at river water station S5, which was situated at the outlet of the Conejeras glacier).
The Nevado del Ruiz volcano (NRV) and Nevado de Santa Isabel volcano (NSIV) limit the upper basin at the top of the UCRB. They contribute to the hydrothermal and volcanic activity of the region (Boh orquez et al., 2005;Robertson et al., 2002), which influences the UCRB through a hot spring that actively contributes water to the basin. Geothermal and geological studies have described a complex tectonic system composed of faults, fractures, and unconformities that configure a network of shallow and deep-water circulation systems in the UCRB and across wider areas of the Central Colombian Andes (Arango et al., 1970).
The climate in the northern Andes is defined by topographic factors, but also by the meridional displacement of the ITCZ and ENSO phenomenon (Arias et al., 2015;Hastenrath, 2002;Sakamoto et al., 2011). Trade winds from the Atlantic and Pacific oceans and the Amazon forest, and a surface tropical westerly termed the Choc o Stream all affect the local meteorology (Poveda et al., 2006). The weather follows a bimodal regime involving two rainy and two dry seasons. Although precipitation occurs in the UCRB throughout the year, suggesting the absence of a strictly dry season, there is a clear contrast between months associated with continuous rain events, and months having less rain; here we refer to the latter as the dry period or season. The rainy periods occur from March to May (MAM) and September to November (SON), with the monthly average being higher for the latter period. The dry periods are from June to August (JJA) and December to February (DJF) (Rabatel et al., 2018). The annual average precipitation is 1300 mm, the environmental lapse rate is À0.52 C/100 m, and the average relative humidity is 83% for the entire basin (Peña, 2016). In the highest areas the mean temperature is À2 to 4 C (WGMS, 2017). The annual evapotranspiration is 300-

| Data collection and analysis
We integrated meteorological data (air temperature and precipitation 2 m above ground) with the water composition of stable isotopes.
Stream and precipitation water samples were periodically collected from January 2017-October 2018 from three rainfall collectors (P1: 3635 m a.s.l.; P2: 4128 m a.s.l.; and P3: 4413 m a.s.l.) and five stream sites (S1, S2, S3, S4, and S5) ( Figure 3; Table 2 and Table 3). The sampling was planned to occur monthly, but for logistical reasons this was not always achieved, and consequently our analyses were constrained by the actual sampling intervals. Each precipitation collector consisted of 0.5 L glass Erlenmeyer flasks fitted with a circular funnel (0.1 m diameter) containing a plastic sieve inside to prevent contamination with debris. The flasks were placed on flat ground, and surrounding vegetation was removed to avoid interference with incoming precipitation. Stream water samples (snapshot) were manually collected using a beaker from the middle of the stream to avoid the influence of bank vegetation. Following collection, samples were transferred to 50 mL brown glass vials closed with silicone septa caps, and stored in cooling boxes until laboratory analysis. In total 140 samples were col- . This illustrates the relationship between air trajectories and stable isotope composition in precipitation. In the west, the Pacific Ocean waters evaporate and travel to the east along the main wind direction, and generate one of the rainiest areas of the world on the Choc o Tropical Moist Forest (CTMF) and the western flank of the Western Colombian Andes (7000-9000 mm/year; IDEAM, 2011). Air masses reach the highest elevations of the UCRB helped by the Choco Stream following several moisture recycling events, and as δ 18 O and δ 2 H deplete with altitude and increased continentality. In the east, the Atlantic Ocean waters evaporate and travel to the west, crossing the Amazon Forest after several moisture recycling events. When vapour reaches the Andes it forms a high precipitation belt on the eastern flank of the Eastern Colombian Andes (5000-7000 mm/year; IDEAM, 2011). In the UCRB, four main ecosystems occur with increasing elevation (Andean forest, cloud forest, páramo, and glacier). The páramo is characterized by an abundance of lakes, wetlands, and peatlands. The páramo ecosystem has been formally divided into subpáramo, páramo, and superpáramo. For simplification the subpáramo is not shown ( Illustration by María Camila Botía, @mariabotia11).

laboratory standards were calibrated against Vienna Standard Mean
Ocean Water (VSMOW2) and Vienna Standard Light Antarctic Precipitation (SLAP2), and used for all measurements. All results were expressed in per mil (‰) using the conventional delta-notation relative to Vienna Standard Mean Ocean Water (V-SMOW). All plots are presented with the Global Meteoric Water Line (GMWL) defined as (Craig, 1961). The analytical precision (the standard deviation of a quality check sample measured in each run) was better than 0.3‰ for δ 18 O values and 0.8‰ for δ 2 H.
To estimate any altitude effects on river water we used the mean catchment elevation, calculated using ArcGIS. Two different approaches were considered to calculate variation in the rate of δ 18 O as a function of the elevation of the river water. In the first approach a monthly mean δ 18 O value, derived only from samples taken simultaneously at the five stations, was used to reduce the bias resulting from seasonality effects and irregular sampling periodicity, while in the second approach all samples were considered. In both approaches the gradients were calculated from the slope of the regression lines. method, which has been widely used to determine correlations between δ 2 H and δ 18 O in precipitation (Payne, 1992); this gives equal weighting to all data points regardless of their respective precipitation amounts (Crawford et al., 2014). The RL for precipitation was calculated based on the precipitation-weighted least square regression (PWLSR) approach, using the open software Local Meteoric Water Line freeware (Hughes & Crawford, 2012); this reduced the effect of small precipitation volumes and potential evaporation during sampling.

| Back-trajectory cluster analysis
To better understand the origin of the air masses arriving at our sampling sites we used the Stochastic Time Inverted Lagrangian Transport (STILT) model (Lin et al., 2003). The model was driven by 3-hourly meteorological fields from ECMWF short-term forecasts (following the contemporary IFS cycle development; more information   Ampuero et al. (2020), in which this time was considered sufficient for the back-trajectories to travel back to the point of last saturation (Hurley et al., 2012).
Based on the 100-particle ensemble, a mean back-trajectory was 3 | RESULTS

| Seasonality in stable isotope composition
Seasonal variation in the stable isotope composition of precipitation and stream water over the air temperature and precipitation variation of M3 is shown in Figure 4. values were conducted for all samples ( Figure S1), but these comparisons were not describing the amount effect as the sample collection period included several rainfall events, and therefore the isotope signature included a mixture of low and high rain events. The correlation coefficients (R) were generally <0.6. There was also wide variation in δ 18 O values as a function of the mean daily air temperature, and there was no clear correlation when individual precipitation samples were analyzed ( Figure S1).

| Altitude effect
An elevation effect on the stable isotope composition was found for precipitation and stream water (Figure 5a,b) Figure 5b). Using only the months corresponding to the five stations (selected months in Figure 5b), a gradient of À0.11‰/100 m (R = À0.79, p > 0.1) was calculated based on the arithmetic average.

| d-Excess in precipitation and origin of air masses
The relationship between d-excess and elevation for precipitation samples is shown in Figure 5c. This shows two linear correlations, the T A B L E 2 Characteristics of the precipitation stations in the UCRB (this study) and international stations (derived from GNIP).  (Rozanski et al., 1993).

| Local and regional comparisons of δ 18 O and RLs
A comparison between precipitation-weighted δ 18 O values and the elevations of stations in Colombia is presented in Figure 7 and

| Seasonality effect
The isotope composition of precipitation and stream water in the Andes Mountains is mostly driven by seasonal precipitation and atmospheric convective rain processes, and to a minor extent by temperature (Valdivielso et al., 2020). This is consistent with the findings of can also change across the ecosystems during the same day. The temperature effect could be better studied by measuring short-term stable isotope variations of rain or fog in páramo ecosystems.
The stream water stable isotope values were more depleted than those for precipitation. This pattern appears to be typical for rivers fed by higher elevation catchments , as has been reported for the high Andes mountains in Bolivia by Gonfiantini et al. (2001), and for the Peruvian Plateau by Bershaw et al. (2016) Figure 3). Thus, despite the records suggesting an influence of greater precipitation amounts during the wet season on variation in the isotope composition, the irregular scale of sampling did not enable specific events be directly correlated with isotope composition data. According to Risi et al. (2008), the amount effect is best observed at intra-seasonal or longer timescales, as it is related to the residence time of water vapour in the atmosphere. This highlights the necessity for developing longer sampling campaigns at study sites, with at least monthly periodicity.
Variation in the δ 2 H and δ 18 O values for stream water at stations S1, S2, and S3, which are located downstream of the boundary between the Andean forest and the páramo (3400 m a.s.l.) was attenuated compared with upstream samples (Figure 4). This attenuation effect has been described by Gomez et al. (2015) (Arango et al., 1970;Stewart et al., 1983).
While evaporation and hydrothermal activity might have dampened the isotope signals, other dampening factors have also been δ 18 O + 9.6); this was not unexpected because in the present study F I G U R E 9 Relationship between δ 18 O and δ 2 H for all precipitation samples, showing regression lines (based on the PWLSR method) and mean values for the stream water stations. The lowest δ-values accounted for those samples from P1 and P3 having the highest levels of accumulated precipitation; for P2 this corresponded to an average accumulated precipitation sample located in the middle of the plot. The P1 and P2 values were similar, but the P3 values were more depleted, probably because of an increased altitude effect. only precipitation from elevations above 3600 m a.s.l. during a different and relatively short time period were sampled. However, it is not clear if the evaporation or mixing occurred during rainfall, within the soil zone, or in the streams (Kendall & Coplen, 2001). Furthermore, while the occurrence of distillation processes can cause evaporation in the rainfall, the characteristics of precipitation and the origin of air masses could have an influence. In the UCRB the main wind direction is from the east during the rainy season and from the west during the dry season. The main wind patterns define the isotope composition observed in precipitation, because of the enhanced moisture recycling and mixing proportions that water vapour is subject to across differing orographic features in different months of the year.
The seasonal changes in the origin of air masses can be also interpreted according to the position of the ITCZ with respect to the study catchment area. During October and November, the ITCZ has a southward displacement, reaching its southernmost point in January (Poveda, 2004). During this time the ITCZ crosses the UCRB, and in January is south of our sampling locations, meaning that there is a stronger influence from the northeasterly trade winds into the northern part of the continent. This explains the northward shift of the clusters during the boreal winter described previously, which was related to the increased precipitation ( Figure 6, upper panel) from September onwards in both years, indicating the approach and arrival of the ITCZ. From April to May the ITCZ moves northward (Poveda, 2004), and in July reached its northernmost location as the precipitation increased again from February to May in 2018. At that time the catchment area was located south of the ITCZ, and the northeasterly winds had lost their dominance; consequently, the clusters showed a more eastward direction.
Because the stable isotope composition changed seasonally, with more depleted values during the rainy season, we hypothesize that the origin of air mass plays an important role. This cannot be explained solely by the change in precipitation. As explained above, air mass origins change seasonally, and this defines the wet and dry months. Therefore, air masses from the Atlantic Ocean travelling across the Amazon rainforest for a large distance, in contrast to those coming from the Pacific Ocean, will result in a more depleted isotope composition because of more moisture recycling along the Atlantic-Amazon trajectory.

| Altitude effect
More depleted stable isotope values were found for the UCRB with increasing elevation, as described by the altitude effect (Dansgaard., 1954;Mook, 2000 if there was a decrease in δ 18 O with increased elevation, but the slopes of the correlations between elevation and δ 18 O were not significantly different from zero. This effect was also described by Gonfiantini et al. (2001), who showed that the altitude effect deviates from linearity when elevation increases. Mountains (Jiao et al., 2019) and the Rocky Mountains (Moran et al., 2007) have interpreted these heterogeneities or inverse altitude effects to the influence of the local water cycle and mixed moisture sources. For example, an increase in the contribution of local moisture could mask an elevation effect (Jiao et al., 2019), or the contribution of vapour masses to both lee and windward slopes could cause a depletion of δ 18 O with distance as precipitation occurs (Moran et al., 2007). In the UCRB the very depleted δ 18 O values found for P1 caused the positive altitude effect with respect to P2; in this case the samples were collected during the first rainy season (MAM), when predominant air masses included a mixture of east and west sources ( Figure 6). We argue that such depleted values can be explained by an ongoing Rayleigh distillation of the moisture coming from the east, and precipitation occurring after the air masses cross over the top of the Andes. Our study did not investigate inter-seasonal effects, and therefore we were unable to quantitatively assess this hypothesis; however, we concluded that there was a significant influence of seasonality and the type of slope on the altitude effect. To cope with such uncertainty, more comprehensive (covering more elevation stations) and long-term (at least 5 years) stable isotope composition sampling should be conducted.

| d-excess
The most positive precipitation d-excess values occurred during dry months, and the most negative values occurred during the rainy months. The variation in d-excess suggests that high values are mainly a consequence of recycled moisture and relative humidity changes, as reported for the Mediterranean Sea region (Koeniger et al., 2016(Koeniger et al., , 2017 (Figure 1). However, these air masses do not contribute a considerable amount of humidity from the original evaporative source, which is reflected in the relative low precipitation play additional important roles, as they directly affect transpiration and atmospheric circulation by enhancing evapotranspiration (Escobar et al., 2022;Hoyos et al., 2018).
The influence on d-excess values of local processes including soil evaporation (Jouzel et al., 2013), mixing sources, and recycled air moisture (Froehlich et al., 2002) should be considered in future analyses. In practice, these can be investigated through longer term isotope monitoring programs that enhance understanding of the influence of open water and land cover on evaporation processes, as proposed by Henderson-Sellers and McGuffie (2006). An approach considering event-based sampling could investigate temporal and spatial convective impacts on the isotope composition of precipitation (Vimeux et al., 2011). Although Gat et al. (2001) argues that precipitation at the crest of the Andes slopes results from the Atlantic Ocean, and at lower elevations from the Pacific Ocean, the processes are likely to be more complex, particularly in the Colombian Andes. This study demonstrates a high level of complexity among factors influencing the isotope composition of rain and stream water in northern South America.

| CONCLUSIONS
This study provides a comprehensive analysis of the spatiotemporal variation of stable isotopes in the UCRB for the period 2017 to 2018.
The isotope values were more depleted during the rainy season. We showed that a seasonal effect was determined by wet-dry cycles, and overlaid by the amount effect. Temperature patterns did not determine variations in the stable isotope composition. A reduction in variation in the composition of stable isotopes in stream water was observed downstream from 3400 m a.s.l., below the boundary between the páramo and the Andean forest, suggesting that ecosystem type and land cover may be driving factors affecting the spatiotemporal distribution. The variety of source areas, and geological, topographic, and ecosystem features, provides a new focus for hydrological research in the northern Andes.
Our study fills information gaps in a region that has been poorly studied, and builds on several studies suggesting the influence of the Choc o Stream and the Tropical North Pacific basin as evaporative sources for precipitation in the Central Colombian Andes. The absence of stable isotope studies in the Choc o region has hindered analysis of the magnitude and seasonality of stable isotope variation. We argue that the hydrology of Andean ecosystems must be understood as part of a wider and more complex system. The páramos and glaciers provide clear evidence that precipitation originates not only from the evapotranspiration of water in high mountain areas, but also from ecosystems at lower elevations, including cloud forest, the Andean forest, and the Amazon rainforest. As a consequence, the ongoing delimitation of water supply basins and definition of boundaries between ecosystems in the high-mountains of Colombia, and long-term goals in territorial planning for the country, should consider boundaries beyond those strictly based on ground-based characteristics.