Study area

The study was carried out in the Toreo El Alto ejido, in Uruapan municipality, Michoacán; located in the Transversal Neovolcanic Axis, between 19°28’22.2’’ N and 102°00’19.7’’ W. This area belongs to the Balsas river region, at an altitude of 1 890 m; the climate is temperate humid, with summer rains (Cw) (^{García, 2004}), with an annual mean temperature range of 10 and 27 °C and has an annual mean precipitation above 1 500 mm. The soil type is volcanic Andosol (^{Alcalá et al., 2001}).

Soil sampling

Four plots on an Andosol with different land use were selected: one forest plot and three agricultural plots (for the culture of conventional avocado, organic avocado and macadamia). Four samples consisting of approximately 2 kg were collected from each, at two different depths (0-20 and 20-40 cm) and transported to the Soils Laboratory of the Graduate School of Forest Sciences of the of the Universidad Autónoma de Nuevo León for further analysis.

Determining the moisture retention capacity

The 32 collected samples were dried in the open air and sifted in a 2 mm sieve. CRH was determined by the pressure plate and pressure membrane method with disturbed samples (^{Yáñez, 2017}).

Rings with a 10 cm diameter were placed on the plate; the samples were subsequently saturated in water during 12 hours and were subjected to different pressures: 33, 50, 150, 500, 1 000 and 1 500 kPa in pressure plate extractors (Soil Moisture Equipment Corp.). They were subsequently dried during 24 hours at a temperature of 105 °C, and the soil moisture was determined using the AS-05 method of the NOM-021-RECNAT-2000 (^{Semarnat, 2002}).

Where:

θg = Moisture content (%)

PB = Weight of the container with a lid (g)

Psh = Moist soil weight

Pss = Dry soil waste (g)

PB+Psh = Weight of the container plus weight of the moist soil (g)

PB+Pss = Weight of the container plus weight of the dry soil (g)

CRH, CC (33 kPa), PMP (1 500 kPa) and AD were estimated using the values thus obtained (difference between CC and PMP).

Complementary variables for determining FPTs

Complementary variables are generally utilized for estimating the FPTs, based on a true, acceptable foundation; in this case, the MO content was determined using the method of Walkley and Black modified by ^{Woerner (1989)}; the texture was determined by using the Bouyoucos hydrometer method, based on the AS-09 technique of the NOM-021-RECNAT-2000 (^{Semarnat, 2002}); and the DA (g cm^-3) was determined by using the gravimetric method (^{Woerner, 1989}).

Pedotransfer functions (FPT)

There are numerous pedotransfer equations generated from edaphological variables, with which mathematical models have been developed in order to determine the CC and PMP.

Water retention was estimated by a collection of mathematical functions from various researches (Tables 1 and 2), in order to obtain the CC (33 kPa) and the PMP (1 500 kPa).

Statistical analyses

Normality tests were applied to the moisture retention capacity. Since the data did not comply with the normality, the Kruskal-Wallis non-parametric test was used (^{Berlanga and Rubio, 2012}) in order to verify the existence of significant differences in the water content at different tensions between land uses.

Validation of pedotransfer functions

The mean squared error (CME) was estimated in order to validate the fitness of the models, based on the following expression (^{Patil and Singh, 2016}):

Where:

Z*i= Moisture retention value estimated for the validation sample i

Zi=Real value for the same sample

n=Total number of validation samples (n=32)

CME allows measuring the error when calculating the moisture content with the regression equation. In addition, the determination coefficient (R ² ) -which establishes the quality of the model for replicating the results- was estimated (^{Patil and Singh, 2016}).

Moisture retention capacity (CRH)

The Kruskal-Wallis test shows significant differences between the land uses assessed at different tensions and depths. According to ^{Daza et al. (2014)}, the changes in the cover generate loss of soil moisture; this impact is reflected by the results obtained in the present research (Table 3). In general, all types of agricultural practices cause a loss of cover in a given time period, with the resulting exposure of the soil to the air and the increased evaporation on its surface.

^{Musso and Suazo (2019)} relate it to the distribution of mineral particles (size), specifying that sandy soils are more permeable, unlike the silts and clays, which retain a larger amount of water.

The moisture retention curves show that the land use for the culture of macadamia exhibited the lowest moister content at both depths, unlike the use for the culture of conventional avocado (Figure 1), which depends on the texture. The results indicated 50 to 70 % of sands in macadamia, and 30 to 40 % in conventional avocado (Table 4).

Humedad volumétrica = Volumetric moisture; Forestal = Forestal; A. convencional = Conventional avocado; A. orgánico = Organic avocado.

Figure 1 Moisture retention curves at two different depths (0-20 and 20-40 cm) in the different uses of the soil. 

The forest system registered high water contents at low pressures, with 65 % and 60 % moisture at depths of 0-20 and 20-40 cm, respectively. The organic avocado system started with a 55 and 50 % at depths of 0-20 and 20-40 cm, respectively, and lost 15 % of the moisture content in both conditions.

As depth increases, the moisture retention capacity decreases as a result of the presence of a larger amount of sands. ^{Volverás et al. (2016)} determined the moisture retention curves and the impact of agricultural management; the authors stated a negative effect on the moisture retention capacity due to the influence of the depth and slope of the assessed areas. In general, the amount of water available in the soil is associated with the MO content. ^{Minasny and Mcbratney (2018)} document that the increase in the amount of water available for the plants due to an increase in MO is uncertain and may be overestimated.

The moisture retention capacity in the forest at CC at a depth of 0-20cm was 65 %; for the culture of organic avocado and macadamia, it decreased by 7 and 13 %, respectively, and for the culture of conventional avocado, it increased by 5 %. The same tendency was observed for PMP (40 % moisture): in the culture of organic avocado, it diminished by 9 %, in macadamia 12 %, and in the conventional avocado, it increased by 3 % (Figure 2).

Humedad = Moisture; Uso de suelo = Land use; Humedad volumétrica = Volumetric moisture; Forestal = Forestal; A. convencional = Conventional avocado; A. orgánico = Organic avocado; CC = Field capacity; PMP = Permanent wilting point; AD = Available water.

Figure 2 Moisture constants at two different depths (A: 0-20 cm, and B: 20-40 cm) in an Andosol with different land uses. 

The amount of available water is in terms of the CC and PMP; at the first 20 cm of depth, it was estimated in approximately 20 % for all land uses; at a depth of 20-40 cm, the forest soil use registered an increase up to 30 % (+10 %) in retention capacity; for the rest of the land uses, there was a slight increase of 1 to 2 %, except for the culture of macadamia, which exhibited 15 % reduction in relation to forest use. ^{La Manna et al. (2018)} researched the water retention capacity in Andosols with different textures and indicated AD values of 7 to 15 % in loamy sandy textures, and of 13 to 18 % in loamy silty textures.

^{Daza et al. (2014)} obtained a CC of 80 % and a PMP of 50 % in forests with native vegetation on volcanic soils in Colombia. ^{Meza and Geissert (2006)} assessed the stability of Andosols with forest and agricultural use; their results indicate a moisture retention of 50 % at CC and 20 % at the PMP in plots for forest use, and of 30 and 10 %, respectively, in plots for agricultural use. ^{Chicas et al. (2014)} report a CC of 13 to 40 % and a PMP of 8 to 30 % in volcanic soils; these values are considered to be low with respect to the values estimated in the present study for forest use, which were 60 to 70 % at CC, and 30 to 45 % at the PMP (Figure 2).

Complementary variables for determining FPTs

Other variables generally associated to the capacity of the soil to retain water had to be assessed in order to determine the FPT. The high MO contents at the first depth (0-20 cm), which ranged between 9 and 16 %, stood out; the DA was low, with a high content of sands and silts, resulting in a loamy sandy texture for most of the assessed soils (Table 4).

At a depth of 20-40 cm, the MO contents diminished noticeably, ranging between 5 and 12 %; the DA increased slightly, while the content of sands and the reduction of the content of silts were noticeable. The content of clays was similar at both depths, with values ranging between 5 and 11 %.

Pedotransfer functions (FPT)

Table 5 shows the CME of the models used for determining the CC and the PMP. It may be seen that the model of ^{Rawls and Brakensiek (1985)} exhibited a low CME (CC = 241.10 and PMP = 210.25), and a high R ² (0.99), which reflects the fact that this model has the best fit with the slightest bias; the rest of the models had higher CMEs: ^{Malavé (1991)}, followed by ^{Delgado and Barreto (1988)} and ^{Chicas et al. (2014)}, to cite just a few.

^{Peralta and Barrios (2006)} point out that the best prediction is represented by the coefficient of determination, which shows that the variations are accounted for by the variables defined in the equation.

The coefficient of determination was high for all models, except for that of ^{Peralta and Barrios (2006)}, with an R ² of 0.39, which accounts for goodness of fit; therefore, the fit of this model is the least reliable, compared to that of the rest of the models (Table 5).

^{Pineda and Viloria (1997)} used the equations of ^{Rawls and Brakensiek (1985)} and ^{Malavé (1915)} and obtained CMEs of 99.73 and 360.21, respectively, at CC, while the values at the PMP were 27.19 and 195.09, respectively. Notably, the model of ^{Rawls and Brakensiek (1985)} exhibited a low CME, just as in the present study. For this reason, it is considered to be one of the best predictors for CC and PMP estimation in Andosols.

According to ^{Pineda and Viloria (1997)}, the regression models represent an inexpensive alternative for estimating the moisture retention in the soil; however, their precision depends on the number of samples and variables utilized for determining it. For example, based on only 57 samples, ^{Chicas et al. (2014)} estimate a coefficient of determination of 0.59 % at CC and 0.68 % at the PMP, which accounts for the variability of the data observed using the generated models. On the other hand, ^{Peralta and Barrios (2006)} obtained an R ² of 0.83 % at CC and 0.70 % at the PMP using only 22 samples; in both cases, the variation in field capacity and permanent wilting point is accounted for by the variables used in the generated models (Tables 1 and 2).