Domestic livestock influence patterns of secondary succession across forest ecosystems. However, the effects of cattle on the regeneration of tropical dry forests (TDF) in Mexico are poorly understood, largely because it is difficult to locate forests that are not grazed by cattle or other livestock. We describe changes in forest composition and structure along a successional chronosequence of TDF stands with and without cattle (chronic grazing or exclusion from grazing for ~8 yr). Forest stands were grouped into five successional stages, ranging from recently abandoned to mature forest, for a total of 2.7 ha of sampled area. The absence of cattle increased woody plant (tree and shrub) density and species richness, particularly in mid-successional and mature forest stands. Species diversity and evenness were generally greater in sites where cattle were removed and cattle grazing in early successional stands reduced establishment and/or recruitment of new individuals and species. Removal of cattle from forest stands undergoing succession appears to facilitate a progressive and non-linear change of forest structure and compositional attributes associated with rapid recovery, while cattle browsing acts as a chronic disturbance factor that compromises the resilience and structural and functional integrity of the TDF in northwestern Mexico. These results are important for the conservation, management, and restoration of Neotropical dry forests.

We stratified observation sites across two types of land use: (1) privately owned, active ranches within APFF-SARC where cattle production has been ongoing for at least 50 years (chronic grazing), and (2) sites in the ReMM where cattle have been permanently excluded (grazing exclusion) since 2008. In both the active ranches and the ReMM, we identified secondary forest stands across a successional gradient, representing different times since agricultural abandonment. We grouped forest stands into five successional stages: i) 2-5 years since abandonment (Recent), ii) 5-10 years (Young), iii) 10-20 years (Intermediate), iv) 30-35 years (Late) and v) mature old-growth forest not  cleared for at least ~100 years (OGF). Combined with the two cattle management regimes (cattle grazing and grazing exclusion), we evaluated 10 combinations of management and successional stage. Information on stand age and cattle production was obtained through interviews with ranch owners and forest rangers. Each treatment combination had three replicates, except for Recent and Young forests excluded from cattle. The availability of sites recently excluded from cattle in the ReMM was limited (Table S1).

We conducted vegetation sampling during the rainy season (July-October) of 2015 (~8 years after the exclusion of cattle). In each of the 27 sampling stands, we established ten 50 x 2 m (0.1-ha) transects (Gentry, 1982), for a total of 2.7 ha sampled. All woody species (trees and shrubs) of height ≥ 1.30 m and diameter at breast height (dbh) ≥ 1 cm were recorded and identified.

We calculated aboveground biomass and carbon storage for each stand using the general model developed by Chave et al. (2005) for tropical dry forests. This model converts biometric measurements of trees (dbh, height, wood density) into biomass values using empirical allometric equations (Chave et al., 2014). We obtained wood density estimates for the 38 most abundant species from Quisehuatl-Medina (2019) and García-Ramírez (2017). For the rest of the species, we obtained wood density values from the world wood density database (Chave et al., 2009) and Ordóñez et al. (2015). We imputed an average value (0.589 g cm³) for six species missing density data. We estimated carbon mass in the woody vegetation using a conversion factor of 0.47 g C / g dry wood (Chave et al., 2005).

2.3 Physical and chemical soil variables

To identify the effects of cattle removal and successional state on soil nutrients, we collected five soil samples in each of the 27 plots between 0-10 cm depth. The five soil samples were combined to make a single homogeneous sample per plot. Half of each sample was dried and was used to measure i) total C by coulometric detection (Nelson & Sommers, 1996), ii) total N using the micro Kjeldahl method (Bremmer, 1996) and, and iii) P using the molybdate colorimetric method following ascorbic acid reduction (Murphy & Riley, 1962). The portion of the soil samples that remained were used to extract NH₄ and NO₃ by the phenol-hypochlorite method. Soil compaction was measured following protocols developed by Blake (1965) (For details see Appendix S1).  Samples were analyzed at the Soil Laboratory of the Instituto nacional de Investigaciones Forestales, Agricolas y Pecuarias (INIFAP, Mexico).