Introduction

Pinus pinceana Gordon is an endemic pine of México. Its distribution is restricted to 17 small and isolated populations, in semiarid mountains of the Sierra Madre Oriental, in soils of calcareous origin, between 1400 and 2300 m above sea level. These populations are distributed in the states of Coahuila, Zacatecas, San Luis Potosi, Nuevo Leon, Queretaro and Hidalgo (^{Ledig et al., 2001}; ^{Favela et al., 2009}; ^{Villarreal-Quintanilla et al., 2009}). Overgrazing by goats and collecting for firewood and cones have placed the species at extinction risk (^{Richardson and Rundel, 1998}; ^{Ledig et al., 2001}). This pine tree is subject to special protection by the Mexican government. For these reasons conservation and management strategies are needed (^{Ledig et al., 2001}; ^{Molina-Freaner et al., 2001}).

The southern limit of the natural distribution of P. pinceana is located at the state of Hidalgo, with two populations, the Arenalito and San Cristóbal, in the El Cardonal municipality (^{Ramírez-Herrera et al., 2010}). Populations that inhabit the margins of the natural distribution of a species are fragmented and isolated from others, with low reproductive capacity due to unviable seed production (^{Westemeier et al., 1998}; ^{Mápula-Larreta et al., 2007}). Species in these areas are limited by the environment exerting strong selection pressure (^{Mosseler et al., 2000}). According to ^{Ledig et al. (2002}), climate change is the primary factor in the gradual decline of many conifers to its current levels, since the last glaciation. Therefore, this pine could have moved north within the Sierra Madre Oriental, leaving small populations at the state of Hidalgo, isolated from populations distributed in greater areas located in the states of Coahuila, Nuevo Leon and Zacatecas (Figure 1). Also, if climate change forecasts come to be true in the short term, species would have to move north, with the associated disappearance of their southern populations (^{Sáenz-Romero et al., 2010}), given that the species natural migration is limited by the natural seed dispersal mechanisms.

Figure 1 Natural population distribution of Pinus pinceana Gordon in México (based on ^{Favela et al., 2009}; ^{Ledig et al., 2001}; ^{Molina-Freaner et al., 2001}, and personal data). 

Seed production is irregular and varying throughout periods of years (^{Cain and Shelton, 2000}). In order to monitor the reproductive capacity of a population, reproduction analyses are used. These are calculated using the potential seed production, the total full and empty seeds, and the relationship between filled seed weight and the total number of seeds per cone (^{Bramlett et al., 1977}; ^{Mosseler et al., 2000}). Saplings are rare in P. pinceana populations south of its natural range. Therefore, the aim of this study was determine the production of cones per tree and the proportion of full and empty seeds in P. pinceana at harvest, during three consecutive production cycles at two southern locations of this pine in the state of Hidalgo, in order to identify possible problems in the conservation of this pine nut in those locations.

Materials and Methods

The two studied populations are located in the El Cardonal municipality, Hidalgo, separated by 8 km (Table 1). The San Cristobal population is located within the Tolantongo canyon, with a low forest timberline of Juniperus flaccida Schltdl. (^{Hiriart and González, 1983}). The Arenalito population is located in the same name canyon, with vegetation in the arboreal stratum dominated by Pinus cembroides Zucc. and Yucca filifera Chabaud (^{Villarreal-Quintanilla et al., 2009}).

Table 1 Characterization of Pinus pinceana Gordon populations located south of its natural range. 

++Normal diameter. *Average and extreme values (in parentheses) of sampled trees.

At each population, 25 healthy trees with cones were selected. These were separated a minimum of 50 m, to cover a larger area in each locality. They were identified for further harvesting. The total height of each tree was measured with a Haga^® altimeter, normal diameter with a diameter tape and the age was measured with an increment borer. During September 2001, 2002 and 2003 the same tree was evaluated for its total cones production, from which five cones were randomly selected per tree to determine reproductive capacity via the cone analysis methodology and their seed production (^{Bramlett et al., 1977}; ^{Mosseler et al., 2000}). The cones length was measured and then placed in a greenhouse (45 °C maximum temperature) in order to induce the scales opening to retrieve the seeds. Each cone was dissected in order to count the full, vain and abortive seeds and their fertile scales, which are located in the central portion of the cone with the capacity to produce fertile ovules. Cones were placed in a drying oven at 70 °C for 48 h to obtain their dry weight. For each cone, the total filled seeds were weighed and 20 seeds per tree were chosen to measure their width and length. These two characteristics and the cone length were only obtained in the first evaluated year.

The cones analysis method provide information to identify seed production and its failures (^{Bramlett et al., 1977}), calculations were performed by cone as follows:

To detect differences between sites and sampling dates we used the repeated measures analysis of variance (^{Gumpertz and Brownie, 1993}), considering these two as fixed effects, except for the error which was random in the model:

Y_ijk= μ+S_i+F_j+SF_ij+ε_ijk

where Y_ijk: observed value of the variable in the k^-th sample in the j^-th date in the i^-th site; µ: general mean; S_i: effect of the i^-th site; F_j: effect of the j^-th date; SF_ij: effect of the interaction of the i-th site with the j^-th sampling date; ε_ijk: error associated with the j^-th date k^-th random sample (tree) of the i^-th site (sampling error).

The analyses were performed using the PROC MIXED and the REPEATED option (^{Littel et al., 1998}) for which averages data per tree were used. Correlation analyses were performed between morphological and trees variables. Statistical analyses were performed using the SAS software (SAS Institute, 1998). The data was transformed by arcsine (ratio)^0.5 before their analysis to improve its normality (^{Sokal and Rohlf, 2012}).

Results and Discussion

There were no significant differences between the two locations, but there was among years and the populations by years interaction (Table 2). Cone production in the 50 trees (25 for each population) was variable during the three years of this study (Figure 2). Cone production was significantly higher during 2001, with 2,188 compared to 747 and 695 cones produced in 2002 and 2003, accounting for 50 collected trees.

Table 2 Probability values obtained from the repeated measures analusis of variance with different seed production characteristics in two populations of Pinus pinceana Gordon at the state of Hidalgo, during three consecutive sampling years (2001, 2002 and 2003). 

Figure 2 Number of trees per cone production category at two locations of Pinus pinceana Gordon. Populations at the state of Hidalgo, during three sampling years (N=50, each population of 25; category 120 represents one tree with 121 and another with 297 cones in 2001 and a tree with 117 cones in 2003). 

Pine nut yield is erratic from one year to another in the same tree (^{Flores and Diaz, 1989}; ^{Richardson and Rundel, 1998}). This depends on the resources availability and proper environmental conditions for reproduction (^{Romero et al., 1996}). There were significant differences between years at the San Cristobal and El Arenalito locations, with a reduction in the number of cones in the second and third year (Table 3), and wide differences between the sampled trees (Figure 2). These differences were most evident in the second population. During abundant production years, seeds have a higher germination capacity and will retain their viability for longer; furthermore, the proportional insect attack impact is lower compared to other years (^{FAO, 1991}; ^{Barner and Olsen, 1994}; ^{Kelly, 1994}).

Table 3 Mean values of different seed production characteristics in two populations of Pinus Pinceana Gordon at the state of Hidalgo, sampled during three consecutive years (2001, 2002 and 2003) 

Mean values with different letters are statistically different (p≤0.05) between years within each population per line. ⁺Calculated on the number of potential seeds. ^§Efficiency=number of filled seeds / potential seed. ^¶ Empty - developed seeds ratio (vain+full). ^¶¶ seed weight / dry cone wight.

In San Cristobal, an average of 34 full seeds was observed in 2001 (54% x 63.4 potential seeds; Table 3), down to 26 and 21 filled seeds per cone in 2002 and 2003. At the Arenalito, 35, 19 and 10 seeds per cone were found during 2001, 2002 and 2003. This indicates a dramatic difference in the Arenalito between years, similar to that of the cones per tree production. The empty - developed seeds relationship (endogamy index) ranged from 0.13 to 0.15 in the first year at the two locations, that is, five to six empty seeds per cone (7.6% x 63.4 potential in San Cristobal and 9.5% x 63.3 potential in Arenalito; Table 3). As not a single plagued seed was detected, and the population density is low, it is assumed that certain level of self-pollination or related crossbreeding occurs. During 2002 and 2003 the empty - developed seed proportion increased from 0.24 and 0.43 to 0.33 to 0.59, at San Cristobal and El Arenalito. These represent 8 and 12.7 empty seeds in 2002 and 8.5 and 11.4 empty seeds per cone in 2003 in San Cristobal and El Arenalito, respectively. The largest proportion of empty seeds at the Arenalito in 2003 is due to lower potential seed in 2002.

The number of aborted ovules was higher in 2002 and 2003 (from 27 to 32 aborted ovules per cone). This suggests a lack of pollen in spring 2001, which increased in 2002, particularly at the Arenalito, which resulted in more unfertilized eggs in the third harvest.

When a cumulative production cone curve was generated, we found that few trees produced the largest seed proportion and that the more pronounced the arc is, the fewer trees produce a higher proportion of the cones. Ideally, each tree should produce the same amount of cones and seeds (balanced production at 45°). Ten trees or less produced 50% of the cones in each of the three evaluated years (Figure 3). Since few individuals produce the greatest amount of germplasm in a production cycle, a lower genetic variability regards the possible maximum is expected in the progeny, which should be more drastic in the second and third year, when fewer trees produce 50% of total cones. In addition, a significant correlation between the cone per tree production and 2002 and 2003 was found (r=0.53, p=0.003), but not during 2001. This indicates that not all individuals produce the same amount of cones each year and that a small percentage of trees contributes to most of the seedlings establishment in a given year, but different trees can provide seeds over time (hence the low repeatability between years).

Figure 3 Contribution of individual trees to the total production of cones (%) in each of the populations of Pinus pinceana Gordon during three sampling year. 

Although there were significant differences in some characteristics, little variation was attributed to differences between the two natural populations (Table 3). There were no significant differences between populations for the number of cones, seed potential, abortive ovules, the number of seeds per kilogram and reproductive efficiency. The interaction between the populations and the year of samplings was significant in all characteristics; thus, the filled seeds production was significantly higher in 2001 and was lower for 2002 and 2003, but variable between these two years depending on the population. The seed weight was higher in 2001 in San Cristobal, with no differences between years at the Arenalito.

The largest contribution to the total variance in 10 of the 11 analyzed characteristics was that of the error, that is, the variance between trees within populations and years. Regarding the cone length, the largest contribution to the total variance was the tree; the tree effect has a greater contribution to the variance in length (54%) and seed width (37%). The variation between trees in the genus Pinus is high and often, the cones and seeds characteristics show greater variation among trees within populations than between them (^{Mosseler, 1992}; ^{López-Upton and Donahue, 1995}; ^{Flores-López et al. 2005}).

During the first year both populations had high seed (54 and 55%) and reproductive efficiency (0.96 and 1.11 g g^-1). These represent a greater biomass proportion allocated to the seeds. The following crops report reduced values, which are between 41% to 18% efficiency less (Table 3, filled seeds proportion). According to ^{Flores-López et al. (2012}), conifers in arid regions have lower seed production than those in better environmental conditions. In Pinus orizabensis D. K. Bailey at Altzayanca, Tlaxcala, efficiency was 29% in an assessment year (^{Sánchez-Tamayo et al., 2005}) and 42 to 81% in 12 populations of P. greggii Engelm. ex Parl. (^{López-Upton and Donahue, 1995}). In seed orchards of P. banksiana Lamb., P. taeda L., P. elliottii Engelm., P. echinata Mill. and P. palustris Mill. Seed efficiency was up to 60% under pest controls (^{Bramlett, 1987}; ^{De Groot and Schnekenburger, 1996}). On the contrary, it is common to find seed production problems in small populations: 37% in P. arizonica Engelm. (^{Narváez, 2000}) and 1.97% in P. leiophylla Schiede ex Schltdl. et Cham. (^{Morales et al., 2010}). These values improved up to 17% with the establishment of a sexual seed orchard of this species (^{Gómez et al., 2010}). Therefore, we conclude that these two evaluated P. pinceana locations produce acceptable seed amounts when many cones are produced, but will be relatively small when there are few cones, which will have few filled seeds.

Other fragmented conifers populations have a less adequate behavior; in Picea Mexican Martínez the efficiency value was of 7% and the inbreeding rate of 0.73 to 0.84 (^{Flores-López et al. 2005}; ^{Flores-López et al. (2012}). Pseudotsuga menziesii (Mirb.) Franco had a seed production efficiency of 25.5% in nine Mexican localities, which relates to the small number of trees in these stands (^{Mápula-Larreta et al., 2007}). According to ^{Owens et al. (2005}), self-pollination and the scarcity and inviability of the pollen are the main causes for the abortion of conifer seeds. In the two evaluated P. pinceana populations the index values ranged between 0.13 and 0.15 in 2001, an abundant cone production year, but increased up to 0.33 in San Cristobal and 0.59 in El Arenalito, both in 2003.

The percentage of filled seeds is determined by the abundance and quality of the pollen, which is variable from year to year (^{Sorensen, 1973}; ^{Todhunter and Polk, 1981}). The average abortive ovules in the third evaluation year were higher than in the early years and were higher at the Arenalito. The highest percentage of empty seeds was also obtained. Their production has been linked to self-pollination, insect and fungi induced damage (^{Bramlett et al., 1977}). By using cones without external damage, no seeds with insect-damage were detected; therefore, self-pollination should be the main cause for the empty seeds found in this study.

The reduction in filled seed production per cone in the second and third harvest is consistent with the expected inbreeding depression effect in nonproductive seed years and in small, isolated populations, where the quantity and quality of pollen are highly variable (^{Mosseler et al., 2000}). In a study of five populations of P. pinceana in the states of Queretaro, San Luis Potosi and Coahuila, the population size was of 1000 to 3000 total trees (^{Molina-Freaner et al., 2001}). Hidalgo stands are smaller than those listed above. In our study, the density was of 117 trees ha^-1 at the Arenalito and of 107 trees ha^-1 in San Cristobal. In small, low density populations an increase of the inbreeding and homozygosity can be expected, so deleterious allele expression and reduced viability and reproductive capacity (^{Mosseler, 2000}).

In the correlation analysis, some of the relationships such as length and cone weight, and the number of filled seeds with the total weight of seeds per cone, were expected to be significant due to the characters interdependence, but it was not the case. Others, such as the seed reproductive efficiency and effectiveness of did not show a significant correlation; only some variables were correlated (Table 4).

Table 4 Significant Pearson correlation coefficients (p≤0.05) among production characteristics of cones seeds of Pinus pinceana Gordon (n=50) 

N.S. Not significant. +Tree elevation within each population in meters.

Cone length was positively correlated with the size (length and width), but not with the amount of filled seed and also cone dry weight to the total weight of the seed. For this reason, if large seeds want to be collected, larger cones should be selected. For the establishment and regeneration, it is important because larger seeds produce a higher number of vigorous seedlings during the first year of life (^{Castro, 1999}; ^{Mueller et al., 2005}). Seed potential has a positive correlation with the weight of the cone (r=0.41), indicating that a greater amount of energy invested in reproductive structures is favorable for the production fertile ovules.

The population at the Arenalito is very sensitive to water stress. This was evident during a drought and high-temperature test where two-year-old plants from this locality were more sensitive than those from San Cristobal and other 10 locations of P. pinceana (^{Martiñón-Martínez et al., 2010}). The two studied stands of P. pinceana in the state of Hidalgo are susceptible to climate change, given their location in the southern periphery of their distribution and because of continuing increase in temperature and higher water deficit (^{Sáenz-Romero et al., 2010}). The most likely scenario is the removal of the populations to lower altitudinal and south ends in the Northern Hemisphere (^{Aitken and Withlock, 2013}), as has been predicted for extreme locations of Pseudotsuga menziesii from Oaxaca (^{Rehfeldt et al., 2014}), or the southernmost of Pinus leiophylla (^{Sáenz-Romero et al., 2015}). Along with reduced phenotypic plasticity at the southern distribution areas of P. pinceana (^{Martiñón-Martínez et al., 2010}). Therefore, in situ and ex situ conservation actions should be conducted (^{Ledig et al., 2001}), making the most out of those years of high cone production, when the seed quality is higher and obtaining seeds is cheaper. Besides, it is necessary to carry out restoration activities such as tree plantings from both populations in a reciprocal manner, in order to promote gene flow between them (^{Aitken and Withlock, 2013}). Even more, the populations of the state of Queretaro should be considered for this purpose, because of their ecological and morphological similarities, similar growth and their quantities of wax in their needles (^{Ramírez-Herrera et al., 2010}), which help to increase their genetic variability by promoting gene flow.

Conclusions

The first year was of high cones and seeds production, with efficiency values of seed production and an adequate amount of empty seeds. These were better than those recorded in other conifers. The second and third year had decreased cones production, their reproductive characteristics value was lower, including less filled seeds and more empty seeds. There is an imbalance in the cones production amongst the evaluated trees; few of these produce most of the harvest, and between years different trees were higher producers. The observed disparity between crops coincides with the documented frequency of seeds production in the pinyon pines. There are no indications that these populations are going through inbreeding depression effects, which is more related to natural variation between years of high or low production, and the disparity between trees with high and low cone production