Study area
The Tehuacán-Cuicatlán Biosphere Reserve covers 490 187 ha in the states of Puebla and Oaxaca, Mexico. The study area hosts 21 ecosystem types; more than 3000 species of vascular plants, of which 12% are endemics, and 180 families have been recorded there, roughly representing 10% of vascular plant types for Mexico (CONANP 2013). One of the most important vegetation types there is the xeric shrubland, which covers 25% of the area. Of the four types of shrublands existing in the reserve, the one that covers the largest area is the rosetophilous shrubland (14.3% of the reserve’s total area, CONANP 2013), which includes Dasylirion lucidum stands (Rivera and Solano 2012). The study area was located amongst these stands, and also included Echinocactus platyacanthus and Agave potatorum on the slopes. Juniperus deppeana dominated the lower flat areas.
The study area was in an area of low mountains aligned from southwest to northeast, reaching 2400 to 2500 m a.s.l., surrounded by flat areas (around 2350 m a.s.l.). The area was formed by Cretacic sedimentary rocks. On the hills, the soils were litosols (CONANP 2013). The data from the meteorological station at Cuesta Blanca, Municipality of Palmar de Bravo, showed a mean temperature of 11.5 °C and mean annual precipitation of 392 mm (for the period 2000 to 2014), falling mostly during the summer; the extreme annual precipitation levels were 133.2 mm (for 2005) and 569.5 mm (for 2008) (CICESE-CONACYT 2018). We visually estimated a surface fuel load of 2 to 4 t ha-1 (no succulent plants were included).
Fires are not common in the reserve. During the period 2013 to 2016, 26 forest fires were recorded, with an average affected area of 33.3 ha (ranging from 1 to 330 ha), all on shrubland. Although the records are incomplete, the main cause of fire was related to livestock raising, and fires occurred in spring (CONANP 2013; CONANP, Tehuacán, Puebla, Mexico, unpublished report: reporte de incendios forestales [2013–2018]). No data on fire regime were available for the study area but, according to the authors’ observations, the rosetophilous shrublands, dominated by Yucca L. species experienced relatively frequent spring surface fires (from every few years to every few decades). This regime is considered a base fire regime, because despite it includes natural and human-ignited fires it maintains the integrity of the ecosystem. Where Dasylirion species were present, the fire regime included passive crown fires. CONANP (2013) points out that there were no reports of fires of natural origin in the reserve. In any case, we could not discard the possibility of natural fires given the long return periods (decades to centuries) that typify this type of shrubland in the United States (USFS 2012a, 2012b).
Species
We considered four species: Dasylirion lucidum (Asparagaceae), Juniperus deppeana (Cupressaceae), Echinocactus platyacantus (Cactaceae), and Agave potatorum (Asparagaceae). The three former species populate rocky limestone slopes with shallow and sandy soils. J. deppeana is a common tree species that usually grows on flat, deep, alluvial soils or on low slopes (Lesur 2011). Except for Juniperus deppeana, the species are endemic. Echinocactus platyacanthus is at risk, and under special protection. All species provide some use to the local population (as ornamental and medicinal plants, for wood, or for making beverages).
Sampling
In April 2014, a human-ignited wildfire occurred in the community of San Sebastián Coacnopalan, Municipality of Palmar de Bravo, within the Tehuacán-Cuicatlán Biosphere Reserve, state of Puebla, Mexico, on rosetophilous shrubland of Dasylirion lucidum on the hillsides, and Juniperus deppeana on the lower and flat parts. The fire covered 330 ha, mostly with an intense, passive crown fire (by which each individual’s crown burns independently of the other crowns) in Dasylirion lucidum (M. Palma Martínez, Reserva de la Biosfera Tehuacán-Cuicatlán, Comisión Nacional de Áreas Naturales Protegidas, Tehuacán, Puebla, México, personal communication.)
Six months after the fire, a total of 32 plots were established on three transects in the affected area: two transects, each 240 m in length, with 12 plots sampled every 20 m; and one transect 160 m in length, with 8 plots also sampled every 20 m. Each plot consisted of a central point from which four quadrants (128 quadrants in total) were placed. From that point, the distance to the closest individual (of any of the four species) in each quadrant was measured. One of the transects (12 plots, 48 quadrants) was established in the stands of Dasylirion lucidum with Agave potatorum, between 2300 and 2400 m a.s.l., with SW and NE aspects; another transect was settled in stands of Juniperus deppeana (2250 m a.s.l., flat); and one more, selectively chosen for its patches of Echinocactus platyacanthus (with 8 plots, 32 quadrants, SW aspect, between 2300 and 2400 m a.s.l.). For each individual of the four species considered, the following variables were recorded, according to the plant life form: height, diameter at breast height and at the base, and crown diameter. In addition, for each individual, the severity of the fire (damage to trunk and crown) was evaluated. We recorded whether each was alive or dead, its stem char height, its lethal crown scorch height, the percentage of lethal crown scorch, and its scorch (necrosed) height. The scorched surface and green surface were recorded for Echinocactus platyacanthus. For A. potatorum, we measured: rosette diameter, proportion of leaves with >50% dehydration, number of dead leaves, number of leaves with <50% dehydration, and number of new leaves.
Data analysis
Means and intervals of density, diameter, and height were calculated, as well as mortality. To calculate the probability of mortality and resprouting, logistic mixed-model regression was used (Hosmer and Lemeshow 2000). The dasometric and fire severity (damage to trunk and crown) variables obtained in the sampling were used as independent variables individually and in combinations. These variables were considered with fixed effects, while the quadrants were considered with random effects. The R program (R Core Team 2017) was used for modeling. In addition, linear regression was used to relate recovery variables to dasometric and fire severity variables.