A feather hydrogen isoscape for Mexico

Journal of Geochemical Exploration 102 (2009) 167–174 Contents lists available at ScienceDirect Journal of Geochemical Exploration j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / j g e o ex p A feather hydrogen isoscape for Mexico Keith A. Hobson a,⁎, Steven L. Van Wilgenburg b, Keith Larson c, Leonard I. Wassenaar a a b c Environment Canada, 11 Innovation Blvd., Saskatoon, Saskatchewan, Canada S7N 3H5 Environment Canada, 115 Perimeter Road, Saskatoon, Saskatchewan, Canada S7N 0X4 Department of Biology, University of Lund, Lund, Sweden a r t i c l e i n f o Article history: Received 13 June 2008 Accepted 3 February 2009 Available online 10 July 2009 Keywords: Deuterium Feather Groundwater House Sparrow Isoscape Mexico a b s t r a c t Developing useful biological isoscapes for areas of the world is a priority. This is the case for Mexico that hosts a large percentage of North America's Neotropical migrant birds. Here we investigated the use of House Sparrow (Passer domesticus) feathers to create a spatially explicit feather deuterium isoscape for that country using samples (n = 461) that were collected across Mexico. Considerable and useful spatial hydrogen isotopic structure was observed, suggesting that isotopes may be a potential forensic tool for evaluating origins of Mexican derived fauna and flora. The most positive feather δD values occurred in the northeast and most negative in the south-central part of the country, roughly matching δD patterns observed in groundwater. A weak negative isotopic relationship was found with altitude in both the Pacific and Atlantic drainage systems. The most parsimonious model describing isotopic spatial variation in feathers between 300 and 3000 m a.s.l. included groundwater δD (δDgw; precipitation proxy), sex, amount of precipitation, and the coefficient of variation in amount of precipitation. Overall, δDgw was a poor predictor of sparrow δDf values for all of Mexico. However, this relationship was considerably strengthened when we considered sex separately, removed the Baja peninsula from our sample, and considered the Atlantic and Pacific drainage basins separately. The strongest relationship between δDgw and δDf was found for female sparrows in the Atlantic drainage basin (r2 = 0.464). We recommend that researchers interested in inferring origins of migratory birds and other animals in Mexico create species specific isotopic basemaps that may be guided by the isotopic patterns we have observed for House Sparrows and groundwater. © 2009 Published by Elsevier B.V. 1. Introduction Fundamental to the practical application of “isoscapes” for tracking migrant organisms over large geospatial scales is that fixed tissue stable isotope values can be directly linked to geographical regions of known origin (Hobson and Wassenaar, 2008). For Neotropical migrant birds, these tissues are often feathers formed in the northern summer breeding or southern wintering sites. The isotopic composition (e.g. 13 C, 15N, 2H) of tissue is linked to discrete and continuous underlying spatial geological or hydrological isotopic patterns through local diet and foodwebs. To date, the long-term growing-season average patterns in the hydrogen isotopic composition of rainfall (δDp) at continental scales have proven to provide the most useful predictable spatial foundation for biological samples (Bowen et al., 2005; Hobson 2008). For example, in North America the strong latitudinal gradient in δDp across much of the USA and Canada is directly reflected in feathers (δDf) grown by birds prior to migration. This hydrosphere– biosphere isotopic linkage provides a powerful means of inferring DOI of original article: 10.1016/j.gexplo.2009.02.002. ⁎ Corresponding author. Tel.: +1 306 975 4102; fax: +1 306 975 5143. E-mail address: Keith.Hobson@ec.gc.ca (K.A. Hobson). 0375-6742/$ – see front matter © 2009 Published by Elsevier B.V. doi:10.1016/j.gexplo.2009.02.007 origins of individuals captured elsewhere (Kelly et al. 2002; Rubenstein et al. 2002; Hobson et al. 2006, 2007). Currently, the strength of the relationship between δDp and δDf has defined the utility of the isotope approach for tracking migrant birds. It is clear that such relationships will be influenced initially by our ability to accurately predict δDp for a given year and region, by the degree to which δDp reflects the δD value of local waters most relevant during the time of feather or tissue growth, and by ecological and physiological processes that may alter the relationship between these two parameters for the focal species. Despite several examples showing excellent and robust correlations between δDp and δDf in temperate regions of North America, much more research is required to elucidate the nature of the variance associated with such regressions (Hobson 2008; Wunder and Norris 2008). In addition, while some regions (e.g. central Europe) have reasonably good isotopic coverage of rainfall, other areas (Africa, Asia, high-latitude regions) have relatively poor data coverage. In North America, Mexico is represented by only two IAEA Global Network for Isotopes in Precipitation (GNIP) stations, and the complex terrain of that country makes an interpolated δDp basemap as a starting point problematic. This is unfortunate because Mexico hosts one of the greatest proportions of all Neotropical migrant songbirds that annually migrate there from temperate areas in the USA and Canada to winter (Petit et al. 1995). Knowledge of the patterns of δD/δ18O in rain and 168 K.A. Hobson et al. / Journal of Geochemical Exploration 102 (2009) 167–174 surface waters of Mexico and how those ultimately relate to tissue values in fauna would greatly assist inferring origins of animals wintering there or for resident species moving within Mexico. In this paper, we present the first avian feather-based deuterium isoscape for Mexico created from spatially extensive collections of sparrow feathers. We further explored the utility of using isotope proxies as an alternative to the long-term GNIP database to create a usable isoscape for inferring origins of birds growing tissues in Mexico. We chose the groundwater δD patterns (δDgw) described in a companion paper (Wassenaar et al., this volume) since groundwater was demonstrated to be a good proxy for annual precipitation in Mexico (Clark and Fritz 1997). At many of the same locations where groundwater samples were collected in Wassenaar et al. (this volume), we also captured House Sparrows (Passer domesticus) and sampled their feathers. Since House Sparrows are non-migratory, experience limited dispersal, and are extensively distributed throughout Mexico below 3000 m elevation, we reasoned that their feathers would reflect local baseline foodweb water δD values during growth. If precipitation and/or shallow groundwater drives the foundation of local foodweb δD values, then we would expect a strong relationship between House Sparrow δDf and δDgw. Besides the pattern observed in the feather basemaps, we hoped to ascertain a useable relationship between shallow ground water and feathers since a good relationship with ground water would provide a spatially explicit proxy for a GNIP-like database for Mexico, which could be used to aid in inferring the origins of birds and other wildlife. 2. Methods 2.1. Field sampling House Sparrows were selected as the target species due to their broad distribution across Mexico and their affinity to both populated and agricultural areas. Birds were captured using mist nets during February to March, 2007 primarily along roadways and conveniently accessible areas (Fig. 1). We collected several feather types from each individual but used the inner (P1) primary for isotope analysis since this is one of the first to be molted and so had the highest probability of being related to the location of capture. Field sampling locations were coordinated with groundwater sampling locations described in Wassenaar et al. (this volume). 2.2. Stable isotope methods All feathers were cleaned of surface oils in a 2:1 chloroform: methanol solvent rinse and prepared for stable-hydrogen isotope analysis at the Stable Isotope Hydrology and Ecology Laboratory of Environment Canada in Saskatoon, Canada. Stable-hydrogen isotope analyses of feathers were conducted using the comparative equilibration method described by Wassenaar and Hobson (2003) through the use of calibrated keratin hydrogen-isotope reference materials. Stablehydrogen isotope measurements were performed on H2 derived from high-temperature (1400 °C) flash pyrolysis of 350 ± 10 μg feather subsamples using continuous-flow isotope-ratio mass spectrometry. All results are for non-exchangeable δD expressed in the typical delta notation, in units of per mil (‰), and normalized on the Vienna Standard Mean Ocean Water — Standard Light Antarctic Precipitation (VSMOW-SLAP) standard scale. Measurement of three keratin laboratory reference materials (CFS, CHS, BWB) (corrected for linear instrumental drift) were both accurate and precise with typical mean δD ± SD values of −147.4 ± 0.79 (n = 5), − 187 ± 0.56‰ (n = 5) and −108 ±0.33‰ (n = 5) per autorun, respectively. A control keratin reference yielded a 6-month SD of ± 3.3‰ (n = 76). All results are for non-exchangeable δD expressed in the typical delta notation, in units of per mil (‰), and normalized on the Vienna Standard Mean Ocean Fig. 1. Location of House Sparrow feather sampling sites in Mexico, January–March 2007, in relation to drainage basin. K.A. Hobson et al. / Journal of Geochemical Exploration 102 (2009) 167–174 Water — Standard Light Antarctic Precipitation (VSMOW-SLAP) standard scale. Corresponding ground water samples used here are reported in Wassenaar et al. (this volume). 2.3. Statistical analysis We tested for spatial autocorrelation in δDf using a permutationbased test of Moran's I index of autocorrelation. Spatial structure in δDf was then modeled using semivariance analysis and kriging interpolation. Numerous models were considered, and we attempted to optimize the interpolation by minimizing the cross-validated Root Mean Square (RMS) error and optimizing the cross-validated regression between observed and predicted δDf. 169 We considered only second-year (SY) and after-second year (ASY) sparrows in our analyses. The country-wide sample consisted of 461 individuals from 54 locations (Fig. 1). Elevation is known to influence precipitation and groundwater δD (Clark and Fritz 1997) and has also been shown to correlate with δDf (Hobson et al., 2003). We examined the relationship between δDf and elevation using linear regression. Linear regression was also used to examine the relationship between δDf and the predicted δDgw from the General Linear Model (GLM) derived in Wassenaar et al. (this volume). We considered groundwater to be a proxy for weighted-average annual precipitation δD and so influence δDf through foodweb processes (Bowen et al., 2005). These analyses suggested that δDgw was a better predictor of δDf than elevation. Since we had multiple measurements for each site, we Fig. 2. Kriged surface of δDf values from all House Sparrow feathers (n = 461) and sampling sites. Sample δDf values were averaged for each site. 170 K.A. Hobson et al. / Journal of Geochemical Exploration 102 (2009) 167–174 2006). Amount of precipitation was used since this factor is known to influence precipitation δD values (Clark and Fritz 1997). Scatterplots suggested that δDf was non-linearly related to CVprecip, therefore we included models with a quadratic relationship between δDf and CVprecip. The most parsimonious models were selected based on quasi-likelihood under the independence model criterion (QIC) to select the working correlation matrix (Pan, 2001) and QICc and model weights to select amongst model subsets (Burnham and Anderson 1998). We only considered models within four QIC units of the top model as potentially useful models (Burnham and Anderson, 1998). We also considered other plausible variables (e.g. evapotranspiration, mean monthly temperature) but they were highly correlated (r N 0.5) with other variables (e.g. mean annual precipitation) and thus were not included to avoid multicollinearity. Moreover, the modeled δDgw values used were based on a parameter set that included additional meteorological data. Prior to analysis, total mean annual precipitation was log(10) transformed to achieve normality. In addition to modeling δDf, we also examined correlations between δDf and precipitationweighted mean annual and monthly predictions of δD in precipitation from a precipitation layer for Mexico and the GIS-based models of δDp (www.waterisotopes.org; Bowen et al., 2005). 3. Results Fig. 3. Relationship between elevation (m) and δD in House Sparrow feathers (n = 280) from Mexico in the Atlantic and Pacific drainage basins. Only birds above 300 m included. modeled δDf using 14 a priori and 19 a posteriori candidate Generalized Estimating Equation (GEE) models, each including δDgw as a proxy for precipitation. We chose GEEs to model δDf since multiple measurements at a site could reasonably be expected to be correlated, and GEEs can account for this correlation by treating site as a subject and each bird as a repeated measurement, and provides robust standard errors (SE) that are appropriate when data are correlated. A posteriori models were fit after discovering δDf followed different patterns in the drainage basins on either side of the continental divide. The models included sex of the bird, total mean annual precipitation, total precipitation in July– September (rainy season), the coefficient of variation of precipitation (hereafter CVprecip), drainage basin (east vs. west of continental divide), and interactions up to the third order. We did not simultaneously consider models with elevation and groundwater δD, because groundwater δD was largely a function of elevation (Wassenaar et al., this volume). Precipitation during July–September was used since the majority of precipitation occurs in this period and this also overlaps with the expected molt period for House Sparrows (Lowther and Cink There was considerable variation in δDf for House Sparrows across Mexico. δDf varied − 14 to −95‰ (mean −61.4‰, SD = 13.0‰, variance = 168.2‰). We found considerable spatial hydrogen isotopic structure in the feathers of sparrows (Moran's I = 0.08, Z = 15.35, p b 0.01). Spatial structure in δDf was subsequently modeled by Universal kriging with 1st order detrending of the data. An anisotropic (direction = 330.4°) spherical semivariogram was fit to the data (cross-validated RMS = 7.8). The model-estimated parameters were a range of 568.9 km, a lag distance of 77.5 km, a sill of 70.0, and nugget variance of 34.8. Our feather isoscape basemap for Mexican House Sparrows is presented in Fig. 2A. In general, there was a northeast to southwest trend from more positive to more negative δDf values. In addition, the Sierra Madre had a large area of negative values, corresponding to high elevations in that region (Fig. 2A). The Baja peninsula was more positive in δD relative to the rest of the western coast, but negative relative to the Atlantic/Gulf coast (Fig. 2A). Despite the strong spatial isotopic structure, there was significant uncertainty associated with these estimates, as standard error of the predictions varied from 6.9–17.4 (Fig. 2B); however, the greatest uncertainty was Table 1 Parameter estimates and robust standard errors (SE) for the QICc selected Generalized Estimating Equation for feather δD for House Sparrows (n = 280) in Mexico. Parameter Estimate Robust SE δDgw Sexa Drainageb δDgw ⁎ Sex Precipitation coefficient of variation Precipitation coefficient of variation squared 48.915 0.591 8.301 − 0.430 − 1.774 0.009 27.528 0.152 2.402 0.159 0.586 0.003 a b 1 = Male, 0 = Female. 1 = Atlantic, 0 = Pacific. Fig. 4. Relationship between δD in House Sparrow feathers (δDf) and the coefficient of variation (CV) in precipitation. Displayed values are residuals from a Generalized Estimating Equation (GEE) controlling for sex, drainage basin, δD in ground water (δDgw), and the interaction between δDgw and sex. Parameter estimates are given for GEE analysis of the residuals; however, see Table 1 for parameter estimates for all parameters. 171 K.A. Hobson et al. / Journal of Geochemical Exploration 102 (2009) 167–174 associated with those areas lacking samples (Fig. 2B). The interpolation explained 41% of the variance in the data (observed δDf = 1.16 predicted δDf + 10.01‰). We found no relationship between δDf and elevation below 300 m and variance for those low elevation data was extremely high. So analyses of the effect of altitude on δDf were restricted to the elevational gradient between 300 and 3000 m (n = 280 individuals). There was an expected but weak negative relationship between δDf and elevation for both the Atlantic and Pacific drainages. Elevation explained 17% and 10% of the variance in δDf in the Atlantic and Pacific drainages, respectively (Fig. 3). We also restricted our GEE model analysis to this subset of 280 individuals. Based on models with the lowest QIC, modeling repeated measures was most parsimoniously accounted for by using independent correlation matrix, and all subsequent GEE models used this form of model. Of the GEE models we explored to explain variance in δDf, one model received 100% of the support based on model weight, and was separated from the next best model by N400 QICc units. The top model included δDgw, sex, drainage basin, interaction between δDgw and sex, and a quadratic relationship with CVprecip. The top model explained 41.6% of the variance in the Fig. 5. Relationship between δDf and A) ground water (δDgw), B) growing season δDp and C) mean annual δDp for House Sparrow in Mexico (n = 461). Table 2 Regressions between feather δD (for House Sparrows in Mexico) and predicted δDf from the top QICc selected Generalized Estimating Equation (see Table 1 for model details). Category Observed vs. predicted N r2 All All without Baja All malesa All femalesa Atlantic Atlantic males Atlantic females Pacifica Pacific malesa Pacific femalesa δDf = − 11.820 + 0.750x δDf = − 2.444 + 0.907x δDf = − 8.377 + 0.805x δDf = 4.049 + 1.031x δDf = 5.357 + 1.045x δDf = 11.086 + 1.131x δDf = 4.512 + 1.036x δDf = − 154.155 + 0.714x δDf = 53.537 + 1.630x δDf = − 2.145 + 0.942x 461 387 221 166 222 132 90 165 89 76 0.232 0.341 0.191 0.468 0.405 0.259 0.464 0.106 0.127 0.222 a Excluding Baja samples. feather data (Table 1). Parameter estimates suggest that δDf was positively correlated with δDgw, males were more isotopically negative compared to females, the Atlantic drainage basin was more positive in δD than the Pacific drainage, and males were not as well correlated with δDgw as females (Table 1). The parameter estimates also suggested that δDf is most negative in areas with intermediate variance in precipitation amount, but was more positive in areas with lowest and highest variance (Table 1). To display the relationship between δDf and CVprecip, we ran a GEE using all of the same parameters as the top model except for the quadratic relationship with CVprecip, and show the residuals of this regression in Fig. 4. The relationship between δDf and δDgw was positive for the complete dataset (n = 461), with δDgw (hence weighted precipitation) explaining 23.4% of the variance in δDf, when drainage and the interaction between drainage and δDgw were included (Fig. 5A). However, variance explained within drainages separately indicated that the QICc selected GEE performed better when considering regional populations and sex (Table 2). Removal of Baja samples from the analysis resulted in the model explaining 34.1% of the variance. The Atlantic drainage showed a better overall regression than the Pacific excluding Baja (40.5% vs. 10.6% of variance explained). The greatest variance in δDf explained by δDgw was for females from the Atlantic drainage (46.4% of variance explained). Generally, regression of the observed values of δDf against predictions from our top GEE suggested that for most combinations of drainage and sex, our predictions showed little bias toward either high or low predictions (i.e. slopes 1, however see Table 2 for exceptions). However, there remained substantial unexplained isotopic variance that may be related to other factors (see below). Residuals from predictions for the samples (n = 491) showed positive spatial autocorrelation (Moran's I = 0.03, Z = 12.26, p b 0.01). Poor model fit in some regions such as Baja was reflected in the interpolated model residuals (Fig. 6), which suggested poor fit in the Baja and Yucatan regions. The relationship between δDf and growing-season δDp (www. waterisotopes.org) was similar to that with δDgw, and explained 24.6% of the variance when drainage and the interaction between drainage and δDp was included (Fig. 5B). In contrast, a similar model using mean annual δDp explained only 19.6% of the variance in the data (Fig. 5C). The poor relationship between δDf and either δDgw, growingseason δDp, or mean-annual δDp was due to high within-site isotopic variance, which ranged from a standard deviation of 3.7 to 16‰ (Fig. 7). Although the strength of the relationships between δDf and δDp were generally weak overall, δDf for all Mexican birds was best correlated with predicted weighted-monthly δD in precipitation for the months of June through August (Table 3). Values of δDf were also correlated with mean annual δDp but more weakly than for individual months during which House Sparrows molt (Table 3). We also examined these monthly and annual relationships for females from the Atlantic since that subgroup showed the best correlation with 172 K.A. Hobson et al. / Journal of Geochemical Exploration 102 (2009) 167–174 Fig. 6. Interpolated residuals from the top Generalized Estimating Equation (GEE) predicting House Sparrow δDf (n = 491). Residual were interpolated using an anisotropic (direction = 347.5 degrees) spherical semivariogram (cross-validated RMS = 8.8), with a range of 796.6 km, a lag distance of 95.3 km, a sill of 57.0, and nugget variance of 33.4. δDgw. That analysis showed a broader range of months with significant but weak relationships between δDf and monthly δDp and also showed a stronger relationship between δDf and mean annual δDp (Table 3). 4. Discussion Here we have presented the first species-specific feather hydrogen isoscape basemap for Mexico. Overall this δDf surface shows considerable and easily detectable geospatial structure that generally agrees with our previous findings on ground water (Wassenaar et al., this volume). The steepest gradient in δDf values was from the northeast through the west-central portion of the country. Although the Yucatan peninsula was not sampled for birds, based on the groundwater patterns, we would anticipate that the Yucatan will show comparatively positive δDf values compared to the rest of the country. In general, we found there was significant potential to use feather δD measurements in Mexico to infer origins of birds and other migrant organisms (Perez and Hobson 2007). Certainly, inferring east and west provenance seems to be the most promising outcome. In general, we found a poor causal relationship between δDgw and δDf for House Sparrows in Mexico. This suggests that while both groundwater and feathers show strong and similar broad spatial isotopic structure in Mexico (Wassenaar et al., this volume), that pattern is likely more complicated than a direct linear correlation, and is not transferred to local foodwebs or to House Sparrows during their period of molt. Our findings for Mexico contrast with other studies that have shown very strong correlations between δD in drinking water and human hair across the USA (Ehleringer et al., 2008) and the strong relationships found between mean annual growing season δD based largely on the GNIP data base and δDf values for forest insectivorous birds in the USA and Canada (Hobson 2008). There are several possible reasons for this poorer causal outcome, ranging from the appropriateness of using House Sparrows to the fact that few studies of hydrogen flow from water through food webs have been conducted at large spatial scales (Wassenaar and Hobson, 2003). Agriculture is widespread throughout Mexico and local diet used by sparrows may have been differentially driven by irrigation using groundwater or from rivers draining large watersheds. Such practices Table 3 Correlation (Pearson's r) between precipitation weighted monthly and annual δD in precipitation (from models of Bowen et al. 2005) and δD of feathers from House Sparrows (n = 461) in Mexico. All birds Fig. 7. Histogram of within-site standard deviation in δDf between individual House Sparrows. January February March April May June July August September October November December Mean annual Atlantic females r p r p 0 0.01 0.01 0.05 0.01 0.28 0.25 0.17 0.00 0.00 0.10 0.00 0.22 0.44 0.24 0.54 0.14 0.24 0.00 0.00 0.00 0.43 0.15 0.03 0.44 0.00 0.11 0.17 0.33 0.20 0.16 0.42 0.32 0.31 0.11 0.25 0.11 0 0.42 0.96 0.16 0.00 0.00 0.19 0.00 0.03 0.04 0.00 0.02 0.69 0.97 0.00 Bolded values indicate pb 0.05. K.A. Hobson et al. / Journal of Geochemical Exploration 102 (2009) 167–174 can certainly contribute to departures from local groundwater and foodweb δD. Unfortunately, we currently do not have detailed landuse or crop statistics for our collection sites and so cannot test for such effects. The use of other isotopes (e.g. 13C, C3 vs. C4) may reveal insights into this possibility. It is possible that the use of a different species may reveal better linkages. Rainfall in Mexico is highly seasonal with most precipitation occurring between June and October. Indeed, δDf values showed the highest correlations with monthly estimated precipitation δD for those months. However, these precipitation δD estimates were based on an algorithm from only two GNIP stations in Mexico and both of these are located on the Atlantic drainage (Bowen et al., 2005). Thus, our estimated δDp values for those months may not match very well with actual values. Precipitation δD values for these months are expected to contribute the most to the groundwater δD signal but the strength of this linkage may vary spatially depending on when and how water is utilized by plants and animals on the landscape. Our analyses suggest that it was also appropriate to consider isotopic sub-regions of Mexico when delineating correlates of feather δD values. At elevations below 300 m, corresponding largely to the coastal plain areas, we observed the most scatter in our data. This likely was associated with more complex and variable patterns of δD in precipitation and water available to terrestrial foodwebs near oceanic coastlines (Dutton et al., 2005). Similarly, we found better overall regression results between δDf and δDgw when we removed Baja samples from our analyses, again suggesting potential isotopic variance associated with coastal regions. Variance in rainfall amount was also identified as a factor in our top models explaining variance in δDf , although this may be partially related to convective water recycling particularly in the coastal lowland regions (Wassenaar et al., this volume). The lack of clear trend in δDf on the west coast follows that found for δDgw and is likely due to the complicating effect of extreme differences in rainfall amounts with latitude that range from b100 mm/yr (Baja) in the north to N2500 mm/yr in the south (Wassenaar et al., this volume). We selected House Sparrows as our test species simply because they were readily available and easily caught, especially around sites of human habitation where groundwater samples were also available. This close connection with human habitation, however, also provided opportunity for these birds to consume imported human foodstuffs. If foods were not 100% locally derived, then this could add to the isotopic variance in the relationship between δDf and δDgw. There was no way to estimate this effect, and in much of rural Mexico we expected human foods to be primarily locally produced. Again, we were encouraged by isotopic studies in the USA showing a very strong correlation between human hair δD and drinking water which suggested strong local water signals are transferred to human tissues despite the complicating factor of a “global supermarket”. Use of more specialist avian feeders like foliage insect gleaners may indeed reduce the inter-individual variance in δDf we found in this species and the chance of non-local foods entering the foodweb. Specialist avian feeders would be considerably more difficult to catch across all of Mexico given the diverse ecosystems ranging from desert to tropical habitat. In addition, tissues from species that represent more longterm average diet, such as bone collagen, may provide a stronger connection between their δD values and those in groundwater (Hobson and Clark 1992). We found that sex was a significant factor appearing in the top model explaining isotopic variance among House Sparrows, with females tending to show stronger correlations between δDf and δDgw. It was not clear why this was the case. Possibly, physiological differences between the sexes resulted in differences in feather isotope values linked in turn to workload and heat balance (McKechnie et al., 2004). Alternatively, if females showed greater site philopatry than males, then we would expect them to better reflect local δDgw values. Clearly, factors contributing to within-site variance in δDf values 173 remain poorly understood and beyond explanation here. In addition to potential physiological and ecological differences among individuals, there is a possibility that some birds sampled were non-local individuals. Our analysis suggested that the groundwater δD isoscape (precipitation proxy) for Mexico may not be useful to generate a useable predicted feather isoscape. Further studies designed to evaluate which hydrogen flow contributes to foodwebs of interest in Mexico is now required (e.g. Dugger et al., 2004). Our study further emphasizes that it may be unrealistic to expect a single model to predict δDf for countries like Mexico whose topography involves a continental divide separating markedly different drainages. Similarly, in their description of δDf values of raptors in Canada and the United States, Lott and Smith (2006) found different relationships between δDf and predicted growing season δDp for various regions. However, ultimately we are most interested in describing tissue-specific δD isoscapes that can be used to infer origins of unknown individuals and here we simply require our best estimate of such isoscapes. In that sense, our feather isoscape map for Mexico stands for House Sparrows. Differences in explained isotopic variance between drainages suggest that attempts to assign individuals of unknown origin in isotopically complex situations (such as Mexico) may require novel approaches. In particular, stochastic likelihood-based assignments tests (e.g. Wunder and Norris 2008) using different error variance for each drainage may be useful. However, we first need to determine how this isoscape and the derived δDgw isoscape (Wassenaar et al., this volume) relates to δDf isoscapes for other species. Acknowledgments This project was funded by Environment Canada operating grants to KAH and LIW. Field sampling was conducted by KL with the assistance of J. E. Martinez-Leyva, A. Schiller, and L. Cruz-Paredes. We are especially grateful to N. Ferriz and staff of Pronatura (Vera Cruz) for their logistical support in Mexico. We thank G. Bowen and three anonymous reviewers for helpful comments on an earlier draft of this manuscript. References Bowen, G.J., Wassenaar, L.I., Hobson, K.A., 2005. Application of stable hydrogen and oxygen isotopes to wildlife forensic investigations at global scales. Oecologia 143, 337–348. Burnham, K.P., Anderson, D.R., 1998. Model Selection and Inference: A Practical Information-theoretic Approach. Springer Verlag, Berlin. Clark, I.D., Fritz, P., 1997. Environmental Isotopes in Hydrogeology. Lewis Publishers, New York, p. 328 pp. Dugger, K.M., Faaborg, J., Arendt, W.J., Hobson, K.A., 2004. Understanding survival and abundance of overwintering warblers: does rainfall matter? Condor 106, 744–760. Dutton, A., Wilkinson, B.H., Welker, J.M., Bowen, G.J., Lohman, K.C., 2005. Spatial distribution and seasonal variation in O-18/O-16 of modern precipitation and river water across the coterminous USA. Hydrological Processes 19 (20), 4121–4146. Ehleringer, J.R., Bowen, G.J., Chesson, L.A., West, A.G., Podlesak, D.W., Cerling, T.E., 2008. Hydrogen and oxygen isotope ratios in human hair are related to geography. Proceedings of the National Academy of Sciences 105, 2788–2793. Hobson, K.A., 2008. Applying isotopic methods to tracking animal movements. In: Hobson, K.A., Wassenaar, L.I. (Eds.), Tracking Animal Migration Using Stable Isotopes. Academic Press, London, pp. 45–78. Hobson, K.A., Clark, R.W., 1992. Assessing avian diets using stable isotopes. I. Turnover of carbon-13 in tissues. Condor 94, 181–188. Hobson, K.A., Wassenaar, L.I. (Eds.), 2008. Tracking Animal Migration Using Stable Isotopes. Academic Press, London. Hobson, K.A., Wassenaar, L.I., Milá, B., Lovette, I., Dingle, C., Smith, T.B., 2003. Stable isotopes as indicators of altitudinal distributions and movements in an Ecuadorean hummingbird community. Oecologia 136, 302–308. Hobson, K.A., Van Wilgenburg, S., Wassenaar, L.I., Moore, F., Farrington, J., 2007. Estimating origins of three species of Neotropical migrants at a Gulf coast stopover site: combining stable isotope and GIS tools. Condor 109, 256–267. Hobson, K.A., Van Wilgenburg, S., Wassenaar, L.I., Hands, H., Johnson, W., O Melia, M., Taylor, P., 2006. Using stable-hydrogen isotopes to delineate origins of Sandhill Cranes harvested in the Central Flyway of North America. Waterbirds 29, 137–147. Kelly, J.F., Atudorei, V., Sharp, Z.D., Finch, D.M., 2002. Insights into Wilson's Warbler migration from analyses of hydrogen stable-isotope ratios. Oecologia 130, 216–221. 174 K.A. Hobson et al. / Journal of Geochemical Exploration 102 (2009) 167–174 Lott, C.A., Smith, J.P., 2006. A GIS approach to estimating the origins of migratory raptors in North America using hydrogen stable isotope ratios in feathers. Auk 118, 16–23. Lowther, P.E., Cink, C.L., 2006. House Sparrow (Passer domesticus), the birds of North America online. In: Poole, A. (Ed.), Ithaca: Cornell Lab of Ornithology; Retrieved from the Birds of North America Online. http://bna.birds.cornell.edu/bna/species/ 012doi:10.2173/bna.12. McKechnie, A.E., Wolf, B.O., Martinez del Rio, C., 2004. Deuterium stable isotope ratios as tracers of water resource use: an experimental test with rock dove. Oecologia 140, 191–200. Pan, W., 2001. Akaike's information criterion in generalized estimating equations. Biometrics 57, 120–125. Perez, G., Hobson, K.A., 2007. Feather deuterium measurements reveal origins of migratory western Loggerhead Shrikes (Lanius ludovicianus excubitorides) wintering in Mexico. Diversity and Distributions 13, 166–171. Petit, D.R., Lynch, J.F., Hutto, R.L., Blake, J.G., Wade, R.B., 1995. Habitat use and conservation in the Neotropics. In: Martin, T.E., Finch, D.M. (Eds.), Ecology and Management of Neotropical Migratory Birds. Oxford University Press, Oxford, pp. 145–197. Rubenstein, D.R., Chamberlain, C.P., Holmes, R.T., Ayres, M.P., Waldbauer, J.R., Graves, G.R., Tuross, N.C., 2002. Linking breeding and wintering ranges of a migratory songbird using stable isotopes. Science 295, 1062–1065. Wassenaar, L.I., Hobson, K.A., 2003. Comparative equilibration and online technique for determination of non-exchangeable hydrogen of keratins for use in animal migration studies. Isotopes in Environmental and Health Studies 39, 1–7. Wunder, M.B., Norris, D.R., 2008. Improved estimates of certainty in stable-isotope based methods for tracking migratory animals. Ecological Applications 18, 549–559.