Assessing the Sensitivity of Riparian Algarrobo Dulce (Prosopis flexuosa DC) Radial Growth to Hydrological Changes

The study area is located in the Central Monte Desert district belonging to the Mendoza Province, central-west Argentina (Fig. 1). Climate is arid to semi-arid, characterized by a mean annual precipitation of 155 mm, and by large seasonal and daily temperature variability (Morello, 1958; Abraham et al., 2009) (Fig. 1).

We selected one site (El Mateo) of P. flexuosa forest growing near the Desaguadero river (32°14′31.4″S; 67°41′21.8″W; Fig. 1). At the analyzed forest stand, vegetation is composed of P. flexuosa with shrub and low arboreal associates such as Larrea divaricata Cav., Geoffroea decorticans (Hook. & Arn.), Capparis atamisquea Kuntze, and Bulnesia retama (Hook.) Griseb. Sampled site presents evidence of relatively recent human pressure (stumps), as most of the P. flexuosa woodlands distributed in the Central Monte Desert (Villagra et al., 2009).

During the austral winter of 2011, 2 rectangular plots of 1000 m² (50 × 20 m), differing in hydrological settings (EMN: El Mateo near; EMF: El Mateo far; Fig. 1), were established, with the longest plot side parallel to the streambank (Table 1). Within the 2 plots, dendrochrono-logical sampling was conducted on all trees regardless of their one- or multi-stemmed form. One to two cores per tree were extracted at about 50 cm above the ground with a gas-powered drill (TED_262R, Tanaka Kogyo Co. Ltd, Chiba, Japan) from a total of 62 trees.

Standard dendrochronological procedures followed Stokes and Smiley (1968): wood samples were glued on wooden supports and polished with progressively finer sandpaper (60–1200 grit), then tree-ring were identified and dated, and widths were measured from pith to bark to the nearest 0.01mm resolution through the facilities of the IPWin4 Image Analysis software (v4.5, Media Cybernetics, USA). The obtained tree-ring series were crossdated and their data quality validated by statistical controls (COFECHA program; Holmes, 1983).

The age-related trend and the influence of exogenous disturbance recorded in raw ring width measurements were removed by standardization procedures through the program ARSTAN40c (Cook and Krusic, 2006). We chose a double de-trending method (Cook, 1985): first, a negative exponential function was fit to each raw tree-ring series, and then we applied a 32-year cubic spline function with 50% cut-off. To reduce the autocorrelation present in ring-width series, autoregressive model of order determined by the Akaike Information Criterion (AIC; Sakamoto et al., 1986) was applied. Tree-ring indices were computed by dividing the measured ring widths by the expected value. Finally, individual standardized series were averaged to produce a mean chronology for each site. We selected the “Residual” version standardized chronology obtained from the ARSTAN40c program, which avoids any possible overestimation of the climate-growth relation (Villalba and Veblen, 1997).

The internal quality of each chronology was examined according to the following statistical indexes: MS (Mean Sensitivity), a measurement of the relative year-to-year change in tree-ring width; EPS (Expressed Population Signal), an estimation of how well a finite-sample chronology can represent the theoretical population chronology based on an infinite number of trees; and RBAR, a measure of the common variance between the single series in a chronology (Fritts, 1976; Wigley et al., 1984). The two latter indexes were calculated for a 15-year win-dow with a 14-year overlap. EPS, following a previous study performed for P. flexuosa woodlands distributed in the Central Monte Desert (Piraino et al., 2015), was considered satisfactory when surpassing a threshold of 0.75 instead of the original 0.85 value proposed by Wigley et al. (1984).

Environmental-growth relations were calculated through a combination of different dendroecological and statistical methods, considering streamflow and precipitation as the abiotic drivers of the species stem growth.

First, since rapid environmental changes are supposed to cause the formation of extremely narrow or wide tree-ring as response to particularly adverse or favorable conditions, raw annual growth series were examined through pointer year analysis (Schweingruber et al., 1990). Pointer years were calculated following the “normalization in a moving window” method, by selecting a 5-years-long window (Cropper, 1979). A given year was considered as a pointer one when at least 67% of the analyzed trees exhibited a change in the radial growth greater than 50%. Calculations were performed through the WEISER software (Gonzalez, 2001). Once established, pointer years were compared to annual streamflow and rainfall averages, respectively defined according to hydrological and biological criteria. Mean annual streamflow, coincident with previous year July-current year June period, was computed following Vich et al. (2014). On the other hand, annual precipitation average corresponded to previous year April-current year March period, representing the time window between the end of preceding and current growing seasons (Giantomasi et al., 2012).

Then, to assess the most important hydrological factors influencing radial growth, correlation functions (Blasing et al., 1984) were performed matching residual ring growth against monthly streamflow and precipitation amount. In order to directly compare the influence of both hydrological variables upon the ring development, we focused on the common time period from preceding year April to current year March, when cambium activity ends (Giantomasi et al., 2012). Correlations were computed with the software DENDROCLIM2002 and tests for significance were determined through the bootstrap method (Guiot, 1991; Biondi and Waikul, 2004). Possible differences in radial growth-environmental variability relation as function of tree spatial distribution were analyzed by the means of ANOVA computed on the r/s coefficients (r being the Pearson correlation moment, and s its standard deviation), focusing only on the current growing season period (October to March; Giantomasi et al., 2012).

Finally, regression analyses were computed to model the relation between radial growth and environmental variables. For model selection, we considered an information-theoretic approach, using Akaike Information Criterion corrected for small samples (AICc; Burnham and Anderson, 2002) to compare a suite of competing models. The best model was selected using the lowest AICc value from a set of candidates. AICc was calculated through the “AICcmodavg” package present in the R software (R Core Development Team 2011; Mazerolle and Mazerolle, 2015).

In all the abovementioned analyses, calculations were performed for the common time period 1991–2008 (n = 18), selected due to the short extension of stream-flow data (available for 1991–2012), and to the last year (2008; data not shown) of both EMN and EMF tree-ring chronologies showing an EPS value higher than the 0.75 threshold established in this research. Although precipitation data covers a larger period (1950–2011), through the established common time window we obtained a clearer picture of the dendrohydrological potential of the studied species. Streamflow data, belonging to the Encón gauge station (32°13′30.10″ S; 67°48′24.00″ W), located at approximately 12 Km from the analyzed forest stand, were downloaded from the webpage of the “Subsecretatia de Recursos Hídricos” (available at http://www.hidricosargentina.gov.ar/acceso_bd.php). Precipitation values were obtained from the San Juan Airport gauge station (31° 34’ S; 68° 25’ W), placed at 100 Km from the study area and which have been successfully used in a previous dendroclimatological analysis of P. flexuosa (Piraino et al., 2015).

As reported in previous researches performed in the Central Monte Desert, the P. flexuosa wood is particularly heavy and hard, and difficult to cut (Villalba, 1985; Iglesias, 2010). Annual rings are semi-ring-porous and present small to large-sized vessels with simple perforation plates (Villalba, 1985). Tree-ring structure is more evident in the radial section close to the pith, since closer to the bark rings become thinner and more difficult to recognize (Giantomasi et al., 2009). Nevertheless, the transition from early- to late-wood, characterized by a decrease in average pore diameter, along with the presence of abundant axial parenchyma, allowed a reliable determination of the annual species radial growth (Fig. 2).

38 samples from 32 trees contributed to the construction of the two tree-ring chronologies, spanning from 47 to 51 years (Table 2). During the chronologies development, two trees belonging to the EMF plot were excluded, in order to obtain radial growth data from individuals of similar age and therefore reduce the ‘noise’ caused by the variation of other factors than spatial position in the riparian zone. Mean annual growth rate, mean correlation (MC), mean sensitivity (MS), as well as mean EPS and RBAR values present similar results among the developed chronologies, being slightly higher for the EMN plot (Table 2).

Pointer years analysis detected respectively three (years 1996, 2000 and 2007) and two (1996 and 2002) characteristics narrow tree-rings for the EMN and the EMF chronologies (Fig. 3). During 1996, at both EMN and EMF plots the totality of individual chronologies showed extremely narrow rings. Regarding the two remaining pointer years emerged at the EMN, 69% (2000) and 73% (2007) of the sampled trees experienced a strong decrease in their radial growth, while at the EMF plot, pointer year during 2002 where registered by 69% of the sampled individuals.

Narrow rings emerged for both the EMN and the EMF chronologies during 1996 corresponded to a year of zero-streamflow value. Concerning the EMN plot, annual streamflow values during the two remaining pointer years were lower than mean (years 2000 and 2007, respectively 34% and 89% of annual streamflow average), while annual precipitation was higher than historical mean (respectively 112% and 199%). Finally, the negative pointer year emerged during 2002 for the EMF chronology corresponded to a year characterized by lower than mean annual precipitation (94%) and higher than mean annual streamflow amounts (304%).

Correlation function analysis performed between standardized chronologies and monthly environmental data revealed different dendrohydrological signals among the selected P. flexuosa populations. EMN radial growth is favored by above-mean spring-summer streamflow values (September to March), while abundant precipitation during May negatively influenced the ring development (Fig. 4). Regarding the EMF chronology, ring growth is positively associated to high streamflow values during autumn and winter months (Fig. 4). Furthermore, at this plot radial growth is benefited by higher than mean streamflow values during spring period (September-October) and by late-summer (March) abundant precipitation (Fig. 4).

Variability in the hydrological settings among the analyzed plots modulated the dendrohydrological signals regarding streamflow influence on the ring growth (oneway ANOVA, F = 17.08; p = 0.0002; data not shown), while no significant differences emerged concerning the precipitation-radial growth relation (one-way ANOVA, F = 4.68; p = 0.0557; data not shown).

Model selection using different combination of dependent and explanatory variables showed that the set of models for residual growth values depending on stream-flow and precipitation were stronger than the null model (no predictors; data not shown). EMN ring growth is best explained by a quadratic regression with mean stream-flow values of spring-summer months (September, December, January, February and March) as explanatory variable (Table 3). On the other hand, the EMF radial growth is best explained by a linear model with August streamflow and March precipitation as explanatory variables (being the last variable marginally statistical significant at the p<0.05 level; Table 3).