2.1 Study zone

The research was carried out in the micro watershed Centerra (Figure 1), located in the high zone of the Dagua river watershed, west Valle del Cauca, in the Colombian Pacific Center with 3º38´45” North Latitude and 76º41´30” West Longitude. The micro watershed counts with an area of 1,814.1 ha, with average temperature of 20° C, average annual rainfall of 2,000 mm. Its main drainage network subdivides in three slopes: Centella, Aguas Calientes and La Virgen; which flow into the Dagua river ⁷. It is characterized for having a relief moderately uneven to moderately steep, with slight erosion, uneven and short gradients. Geomorphologically located on the hills in the mountain landscape, with altitudes between 1,050 msnm and 1,700 msnm approximately, with taxonomic classification of the soils belonging to Typic Hapludands ⁸.

Figure 1 Geographical distribution of selected farms in La Centella micro watershed 

The main crops presented in the micro watershed Centella are coffee (Coffea), plantain (Musa sp), coffee-plantain association (Coffea - Musa sp Association), bean (Phaseolus vulgaris) and sugar cane (Saccharum officinarum) ⁷.

2.2 Selection and description of study farms

Nine farms were studied (Table 1) distributed along the three slopes of the micro watershed Centella, defining as selection criteria the soil taxonomy, changes in the field elevation that divided the micro watershed in three zones: high, medium and low and the established crops of coffee-plantain and coffee-guamo associations, sugar cane and grass for livestock.

2.3 Determination of soil hydrophysical properties

The experimental design allowed to obtain 36 samples, two per farm, between altered and non-altered at two depths of 0 - 10 cm and 10-20 cm. The extracted samples were analyzed in the Water and Agricultural Soils Laboratory (LASA) of the Universidad del Valle with the methods presented in the Table 2.

2.4 Experimental estimation of the humidity retention curves

The humidity retention curves were estimated once the laboratory work finished, based on the study farms soil samples, using the beveled cylinder and the pressure plate corresponding to the pressures of 0, 0.33, 3, 7, 10 and 15 Bar. Then, the tabulation of humidity contents versus the pressures, represented in a curve through a spreadsheet.

2.5 Hydrus 2D estimation of the humidity retention curves

The humidity retention curves were estimated from the hydrophysical properties using the Hydrus 2D software, applying the Van Genuchten (VG), Brooks and Corey (B-C), Van Genuchten modified (VGM) and Kosugi (K) hydraulic models that involve two variable entrance options: The determined soils catalogue considering the textural class and the Rosetta module that allows to predict hydraulic parameters of the soil using pedotransfer functions (PTF) associated to the soil texture distribution (% Sand, % Silt and % Clay), the apparent density and two water retention point corresponding to 0.33 Bar and 15 Bar ¹². The models results were compared through the medium error and the cuadratic mean error square root (RMSE) considering the two sampling depths.

Software Hydrus 2D individual licenced, is a finite elements model to simulate the water flow through porous mediums partially saturated, resolves numerically the Richard’s equation ¹³.

The equation (Eq.1) that rules the flow in porous mediums non-saturated in two dimensions, considering the movement of the air does not produce a noticeable effect in the water movement ¹⁴ will be:

Where:

θ: Volumetric water content [L³/L³]

h: Pressure altitude (L)

𝑥_𝑖: Spatial coordinates (i=1,2) (L)

t : Time (T)

KijA : Anisotropy tensor components

Z: Vertical coordinate

K =: Non-saturated hydraulic conductivity (L/T).

This equation is not lineal and for its solution the definition of the soil hydraulic functions is needed.

2.6 Verification of the correlation degree in the obtained data

The hydraulic model used in the humidity retention curves determination was compared with the obtained data in laboratory, through the medium error (ME), the cuadratic mean error square root (RMSE), the Pearson’s correlation coefficient (r) and the Lin concordance correlation coefficient.

3.a Field capacity behavior

The farm soils with higher humidity value to field capacity (FC) (0.33 Bar), correspond to those presenting higher clay content, like La Camelia and Villa Victoria, followed by El Brillante and Bellavista with sandy-loam texture. This behavior is atypical in thick textures and can be related to the high OM contents and mostly, to the low apparent density values in these soils ²⁷.

In general, there are differences in the humidity content at FC between the plant covers of study, being the farm soils with coffee - plantain and coffee - guamo associations the highest values of FC, followed by the soils cultivated with sugar cane and grass. These differences can be attributed to the fact that the guamo trees provide biomass that gives the soil the water regulation capacity under the coffee association, besides, it decreases temperature and water evaporation from soil. The farms with grass plant cover reported the lowest FC values at both depths, which can be related to highest apparent density values and low - medium OM content, influenced by the cattle trampling, and the lands being susceptible to superficial compaction, complicating water infiltration ²⁷. For the sugar cane cultivated farms, it was found the FC values are adequate for the crop development.

3.b Available water content

The difference in the humidity contents between 0.33 and 15 Bar, reflects that the study soils have a great water retention capacity, which can be available for agricultural crops production. The largest usable available water content (AWC) have values between 17.33 to 23.41 mm for 0 - 10 cm depth and from 16.38 to 31.14 mm for 10 - 20 cm depth, presenting the lowest values Villa María and El Potrero farms (< 8.12 mm), as a consequence of the applied practices in pasture.

3.1 Humidity retention curves estimation in Hydrus 2D

According to the Table 6, the Van Genuchten and Van Genuchten hydraulic model modified with the Rosetta method presented the lowest medium error (ME) for most of the farms at both evaluated depths, with a value no higher than -0.11 cm³/cm³. On the contrary, the models Van Genuchten, Van Genuchten modified, Brooks and Corey and Kosugi for the soils catalogue present the highest values. The greatest water content underestimations are presented at the 10 - 20 cm depth, reaching values of -0.27 cm³/cm³ in Van Genuchten, Van Genuchten modified and Kosugi (with catalogue) and -0.22 cm³/cm³ in Brooks and Corey. The greatest dispersal were observed in the models Van Genuchten, Van Genuchten modified (with catalogue), Brooks and Corey and Kosugi for the soils catalogue with 0.28 cm³/cm³, 0.23 cm³/cm³ and 0.28 cm³/cm³ values, respectively. However, in the hydraulic models of Van Genuchten and Van Genuchten modified with Rosetta module a less dispersal no superior to 0.11 cm³/cm³ was presented. The obtained values in the ME and RMSE calculations for the hydraulic models that use the soils catalogue present the greatest underestimations and dispersals in comparison to the models of Van Genuchten and Van Genuchten modified with Rosetta module, since these involve each farm soil properties, generating a water retention model similar to the real conditions. Furthermore, Hincapie ¹⁸ reports that the Van Genuchten hydraulic model based on the soils catalogue appears to be limited for volcanic ashes-derived soils, since these present humidity values at saturations higher than 48.6 % (cm³/cm³), being these model obtained values not valid. In turn, Rubio ⁴ describes that the Brooks and Corey and Kosugi models present slightly acceptable values in the retention curve.

Therefore, these models were used to continue the humidity retention behavior analysis in the selected farm soils, through the evaluation of the uncertainty obtained in the experimental data and Hydrus 2D data.

3.2 Comparison between the experimental and the Hydrus 2D humidity retention curves

The Hydrus 2D retention curves obtained present underestimation for most of the humidity points in comparison to the experimental values obtained from the table 5. However, in the farms Villa María (Figure 2) and El Potrero (both depths), La Yiyita and La Esperanza (0 - 10 cm) and El Encanto (10 - 20 cm) was showed an overestimation in the saturation point. Although the values of the experimental and Hydrus 2D curves present variations between the farms, a similar behavior was noticed.

Between the 0.33 Bar and 4 Bar pressures it was observed that the curves of both methods show a pronounced slope change. However, the corresponding to Villa María and El Potrero (both depths) do not present this behavior from the saturation, since these soils have a smaller total porous space and a high apparent density, product of the applied agricultural labors. Subsequently, gradual changes in the humidity content were observed when the pressure increased. This result is similar to the research developed by Dirksen in 1991, which attributes to the characteristic properties of the Andisols related to high water retention capacity ¹⁸.

Figure 2 Experimental and Hydrus 2D humidity retention curves for Villa María farm at 0 to 10 cm depth. 

3.3 Correlation between the experimental and the Hydrus 2D humidity retention curves

The farms that presented overestimation in the saturation point (Table 7) were Villa María and El Potrero (both depths), La Yiyita and La Esperanza (0 - 10 cm) and El Encanto (10 - 20 cm), showing humidity differences from 0.05 to 0.08 cm³/ cm³, values that invert in the near FC zone.

In the field capacity point was found the majority presented underestimation in the humidity content between -0.01 and -0.06 cm³/ cm³, except from La Esperanza (10-20 cm), which presented the greatest underestimation of 0.13 cm³/ cm³. For 3 Bar pressures between -0.02 and -0.13 cm³/ cm³, 7 Bar between -0.05 and -0.17 cm³/ cm³, 10 Bar between -0.04 and -0.16 cm³/ cm³ and between 0.06 and 0.16 cm³/ cm³ for permanent wilting point. The previous values are considerate slightly acceptable for being superior to 2.5% in accordance with Rajkai (1996) mentioned by Rubio ⁴.

The dispersal found in the humidity values is between 0.04 and 0.12 cm³/ cm³, whose values are high (>5%) indicating there are great errors in the prediction, which can be not only due to the estimation procedure by the Hydrus 2D software but also to the error that carries the application of an experimental method.

The humidity retention curves obtained with both methods were compared through the Pearson correlation coefficient and the Lin concordance coefficient (Table 8). In general, a Pearson coefficient of 0.96 (p<0.05) was found, indicating an elevated correlation degree, which can be result of its sensibility to opposite values, while the LIN coefficient of 0.80 corresponds to a poor concordance correlation.

Finally, according to the uncertainty analysis it is inferred that Hydrus 2D software presents humidity content values slightly acceptable, tending to an increasing underestimation as the pressures increase, for the study soil. This underestimation in the pressure range from 0 Bar to 0.33 Bar can be related to the particular physical and hydraulic properties of the Andisol soil like a low apparent density, high porosity and high organic matter content that increase the water retention capacity. However, at a 0.33 Bar pressure the program presents a better estimation, since the values are closer to the laboratory measurements. As the pressures increase, the humidity retention is less related to the structure and is more influenced by the clays texture and mineralogy, so they can be related to the greatest underestimations due to the high microporosity, characteristic of the volcanic ashes-derived soils and due to the high clay percentage that presents the largest specific area, allowing to form organic-mineral complex that increase the humidity retention.