Study area

The study area is located in the ejido San Jerónimo in the municipality of Melchor Ocampo, where natural L. graveolens populations are exploited. It is located in the North of the state of Zacatecas (24°59′6.00″ N and 102°13′2.00″ W), at an average altitude of 1 450 m (Figure 1). The area is located within the Chihuahuan Desert and belongs to the RH36 Nazas Aguanaval hydrological region, in the basin of the Laguna de Mayrán y Viesca and the sub-basin of the Laguna de Viesca (Sistema Nacional de Información sobre ^{Biodiversidad [SNIB], 2025a}, ^2025b). The dominant vegetation is microphyll desert scrub and rosette desert scrub (^{SNIB, 2025c}). The soils are of the Lithosol (dominant) and Solonchak types (^{SNIB, 2025d}). The prevailing climate is very arid, semi-warm (BWhw), with temperatures ranging between 18 and 22 °C. The average annual precipitation is 200 to 400 mm with summer rains, equivalent to 5 and 10.2 % of annual rainfall (^{SNIB, 2025e}).

Melchor Ocampo = Municipality of Melchor Ocampo.

Figure 1 Location of the study area. 

Sampling and measurement of variables

In the L. graveolens collection area, a starting point was randomly selected, from which distances of 100 m were marked to define 22 sites (^{Martínez-Bencardino, 2019}; ^{Valdivieso-Serrano, 2021}). At each site, a 25 m² (5×5 m) quadrant was marked out and subdivided into four equal quadrants (^{Bonham, 2013}). The plants present at the site were counted; in the 25 m² quadrants, one plant was collected at random and its morphometric variables were measured: (a) plant height (h, cm); (b) smallest diameter (sd, cm), measured at ground level, considered the base of the plant; and (c) largest diameter (ld, cm) of the shrub cover, which corresponds to the widest part of the aerial cover of each plant. The above variables were obtained using a flexometer (model 5010 Arly^®); the mean diameter was estimated to calculate the average cover diameter (acd, cm). Subsequently, the entire plants from each quadrant were collected, placed in paper bags, and labeled for proper identification and transfer to a storage center.

The leaves of each individual collected were separated and weighed using digital scale (model L-EQ 5/10 Torrey ^® with a capacity of 20 kg); this represented the fresh leaf weight (flw, g). Next, the leaves from each plant were placed on a plastic sheet and exposed to the sun for a week to dehydrate them. The samples were considered dry when the leaves felt semi-brittle to the touch. Once the process was complete, the dry weight of the leaves (dlw, g) was recorded using the same digital scale (^{Conafor, 2011}; ^{Orona-Castillo et al., 2017}).

Statistical analysis

Twenty-two sites (m=22) were considered, and four plants were collected at each site, yielding a total of n=88 plants. The analyses were carried out in Excel^®. A data exploration analysis was performed to estimate the sample mean (x-) and sample standard deviation (s), the interval x-±s, and the minimum and maximum values for each variable evaluated. The normality of the data was tested using the Kolmogorov-Smirnov test (KS>P) (Table 1) (^{Zar, 2014}).

Pearson's simple linear correlation coefficient (R) (P<0.01) was calculated for the variable pairs comprised in the field data matrix (^{Zar, 2014}). Since the predictions of flw (g) and dlw (g) were an essential part of this research, the highest correlation was considered for the variables h (cm) and acd (cm). Once the highest value of the R correlation was identified, the 88 plants were grouped into intervals of that variable in the units of measurement (cm) (^{González-López et al., 2022}; ^{Zar, 2014}).

Twelve linear and nonlinear regression models were evaluated using the Least squares method (Table 2) (^{Zar, 2014}) to predict the leaf production in terms of fresh weight (flw, g) and leaf yield, given by its respective dry weight (dlw, g). The following criteria were used to select models: the probability of committing a type I error (P<0.01) in the regression, the highest Adjusted coefficient of determination value (Raj2), minimum values for the Root mean square error (RMSE), and the Coefficient of variation (CV) (^{Zar, 2014}).

In addition, the significance of the regression coefficients of the models was tested (Table 3) using Student's t-statistic. Homoscedasticity was verified in the selected models (^{Breusch & Pagan, 1979}; ^{Maldonado-Ortiz et al., 2022}). Subsequently, flw tables for leaves and leaf yield were generated in terms of their dlw.

Statistical analysis

The average flw was 149.4±62.9 g per plant^-1, and the average dlw was 31.1±13.8 g plant^-1 (Table 4). On average, the dry leaf yield was 22.18 %. The average density was 8 272 plants ha^-1. The average measurements of h, acd, and dlw of leaves were similar to those recorded in Zacatecas by ^{López-Serrano et al. (2021)}. The values obtained comply with the standards that recommend using plants taller than 30 cm (^{Semarnat, 1997}) and with coverage diameters of over 50 cm (^{Conafor, 2011}). The re/sults of this study meet these criteria.

During the drying process of L. graveolens leaves, a 79 % decrease in initial weight was quantified. The values cited for Zacatecas for dlw (^{López-Serrano et al., 2021}) indicate that the variation is low; however, similar studies do not record the flw (^{Flores-Garnica, 1994}; ^{Villavicencio-Gutiérrez et al., 2018}). The methodology utilized seeks to promote the sustainable use of L. graveolens; it was applied in accordance with current regulations, as only plants taller than 30 cm and with a diameter of over 50 cm were measured (^{Conafor, 2011}), and at least 20 % of the specimens present at the site were left uncollected (^{Semarnat, 1997}). This management allows plants to reach maturity, regenerate naturally, and help prevent soil erosion in their habitat. Together, these actions support the use of methods based on morphometric measurements and promote the sustainability and conservation of the resource (^{López-Serrano et al., 2021}; ^{Semarnat, 1997}).

Correlation between the evaluated variables

The variable pairs with statistical significance at 99 % (P<0.01) were for flw and acd with R=0.753, and dlw and acd with R=0.472, which allowed us to order them into six 25-50 cm intervals for the independent variable acd. Similar studies determined relationships between the dlw and the variables h and acd (^{Flores-Garnica, 1994}; ^{López-Serrano et al., 2021}; ^{Villavicencio-Gutiérrez et al., 2018}).

These results are important for estimating the biomass and dry leaf yield of the species (^{Flores-Garnica, 1994}; ^{Villavicencio-Gutiérrez et al., 2018}). Correlation values support the creation of performance tables; however, such analysis is not always carried out (^{Flores-Garnica, 1994}); however, their estimates prevent excessive harvesting of plants, ensuring their rational use (^{Ares et al., 2002}; ^{Semarnat, 1997}).

Table of the fresh leaf biomass yield (flw) and the dry fiber yield (dlw) for Lippia graveolens

Table 3 shows the adjusted values for predicting the flw and dlw performance; also included is the significance test result for the regression coefficients of the models with Student's t-statistic. Table 5 shows the values of the model selection criteria. For the flw, the best result corresponded to the nonlinear model 4 (flw=22.0631.050acd), while for the dlw, it corresponded to model 10 (dlw=10.9331.027acd). The selected models and their graphical behavior are shown in Figure 2, which shows the confidence bands (99 %) for the field data and the intervals for the average regression.

LCB = Lower-confidence band; HCB = Higher-confidence band; acd = Average cover diameter of the plant (cm); flw = Fresh leaf weight (g); dlw = Dry leaf weight (g); e _i = Residuals.

Figure 2 Selected models with confidence bands (P<0.01) and residual plots. 

The results are similar to a study of the dry leaf biomass of the same species in the state of Coahuila, in which regression models were adjusted without recording the coefficients, making it impossible to compare yields (^{Flores-Garnica, 1994}; ^{Villavicencio-Gutiérrez et al., 2018}). The similarity lies in the fact that the best fits are also with nonlinear models between the variable dlw of the leaf and the canopy diameter; with the variable h, the fit was poor (R=0.08) (^{López-Serrano et al., 2021}; ^{Villavicencio-Gutiérrez et al., 2018}). Furthermore, a good fit was obtained in the present study between the flw and the acd (R=0.979).

In the selected models, the residuals were shown to comply with homoscedasticity (P<0.01) (^{Breusch & Pagan, 1979}; ^{Maldonado-Ortiz et al., 2022}). In research conducted with L. graveolens in the state of Coahuila, other adjusted models exhibited heteroscedasticity (^{Villavicencio-Gutiérrez et al., 2018}).

The model selection criteria and the residuals’ compliance with homoscedasticity allowed us to obtain the flw and dlw yield tables of L. graveolens in grams (Table 6).