Production and quality of Urochloa decumbens (stapf) r.d.webster forage co-related to the physical and chemical properties of the soil

Frequently degraded pastureland characterized by low soil fertility and compacted surface is the basic environment of Brazilian livestock. The physical and chemical characterization of soil and its co-relationship with forage production are determining factors for performance of animals raised on pasture. The objective was to analyze the forage production of Urochloa decumbens grass correlated, linearly and spatially, with physical and chemical attributes of a savannah soil in Selvíria - MS, Brazil. A geostatistical web was introduced for the collection of soil and plant data, with 120 sampling sites within an area of 56.09 ha. The descriptive analysis of the data was undertaken and linear co-relationships, both simple and multiple, were established between plant and soil properties. Semivariograms were modeled and their respective krigings and cross-validations obtained, coupled to co-krigings (plant and soil). Production of dry matter and crude protein rates of U. decumbens may be estimated by regressions and the mechanical resistance to penetration and gravimetric humidity of the soil evaluated. Since organic matter rate and the gravimetric humidity of the soil are co-related spatially with the rate of crude protein of U. decumbens , they are the best factors to calculate or increase the forage crude protein rate.


INTRODUCTION
Brazil has the most numerous commercial cattle herd in the world and the state of Mato Grosso do Sul contributes largely for such ranking. Although the country is known to have large areas for extensive planting of forages (Ribeiro-Júnior et al., 2017), low productivity and unsatisfactory scientific indexes are verified when compared with rates from other countries which also export beef worldwide (Cavallini et al., 2010;Pariz et al., 2011;Montanari et al., 2013a). In spite of low productivity rates, the Brazilian commercial cattle herd amounts to 209,541,109 units. It has been estimated that 89% are raised exclusively on pasture in native or cultivated pasturelands covering some 172 million ha and approximately 85% of this area is pasture of the genus Urochloa (Leal et al., 2016).
Factors such as species, climate (water, light, and temperature), fertility and physical texture of soil, forage cycle, harvest management, animal monitoring, and others may affect the chemical composition and thus the availability of food energy (Leonel et al., 2009;Lisbôa et al., 2016). The production of dry matter and satisfactory rates of crude protein and acceptability by the animals are important factors in the selection of a cultivar for pasture sowing (Maranhão et al., 2009).
According to Cavallini et al. (2010) and Montanari et al. (2010), studies on the spatial variability of the soil physical and chemical properties are highly relevant in agriculture and livestock-raising since they determine specific regions for management. The use of semivariograms at this phase helps in planning soil sample collection to analyze the spatial variability of the most interesting properties by geostatistical techniques that would help the farmer in determining and improving the productive, physical, and chemical capacity of the soil.
In fact, Pariz et al. (2011) investigated the spatial variability of forage productivity and the physical properties of soil planted with U. decumbens pasture in the Brazilian savannah and discovered that the geostatistical use could be underscored as an important tool to understand the interactions in a pasture ecosystem, reduce possible causes of degradation, and demonstrate the best alternative management for the soil-plant-animal complex. In an analogous way, Montanari et al. (2013b) evaluated the production of dry matter of U. decumbens according to the chemical attributes of a Latosol in Selvíria, MS, Brazil and verified that the organic matter content was a good index of soil quality when co-related to the productivity of Urochloa grass. Further, the rate of crude protein was also a good quality index of the forage plant with regard to the production of dry matter.
Taking into account the Brazilian relevance in livestock, the current paper evaluated the forage production of U. decumbens grass correlated, linearly and spatially, with physical-chemical attributes of a savannah soil.

MATERIAL AND METHODS
The experiment was performed In the Teaching, Research and Extension Farm (FEPE) in Selvíria, MS, Brazil, of the Engineering Faculty of the Universidade Estadual Paulista (UNESP), 51°24'46'' W and 20°21'44'' S ( Figure 1); mean altitude 342 m. The soil in the area is a typical mediumtexture Red-Yellow Latosol (Oxisol), a moderate, hypodystrophic, alic, caulinitic, hypoferric, very deep, moderately acid, with 410 g kg -1 of sand, 70 g kg -1 of silt, and 520 g kg -1 of clay at 0-0.20 m depth. The initial chemical attributes of the soil, assessed respectively at the depths of 0.0-0.10 and 0.10-0.20 m, were: pH (CaCl2) = 4.9 and 4.7; H+Al = 26 and 22 mmol c dm 3 ; P (resin) = 10 and 5 mg dm 3 ; K + = 0.6 and 0.1; Ca 2+ = 9 and 7; Mg 2+ = 4 and 3 mmol c dm 3 ; and V = 34 and 31%. The Urochloa decumbens pasture was implemented in 1978 and has been managed under continuous stocking (mean stocking rate of 1.5 UA in that period) for Guzerá cattle raising, receiving liming and fertilization during implementation and rehabilitation in 1988. It is currently in a stage of moderate degradation, with losses in productivity and quality, broadleaf weeds, and soil compaction.
According to Köppen, climate is Aw, featuring megathermal humid climate, with dry winters and hot rainy summers. Mean annual rainfall is approximately 1232 mm; mean annual temperature is close to 24.5 0 C and relative humidity between 70 and 80%.
Urochloa decumbens (cultivar Basilisk) was planted about 25 years ago and has been extensively employed, with liming when required, to upgrade base saturation at 60%. For data collection in March 2009, U. decumbens was in its tenth reformation year with surface scarification and liming.
Directions x and y of the cartesian coordinate system were defined in the allocation of the experimental design, with 120 sampling sites distributed in 56.09 ha (1060.23 × 529.00 m). Spacing of sites varied between 35.18 and 279.93 m. Soil and plant properties, individually collected around each sampling site of the experimental design were analyzed. Soil properties comprised mechanical resistance to penetration (RP) and gravimetric humidity (GH) at layers 0-0.10 m (RP1 and GH1); 0.10-0.20 m (RP2 and GH2); 0.20-0.30 m (RP3 and GH3); 0.30-0.40 m (RP4 and GH4); 0.40-0.50 m (RP5 and GH5); and 0.50-0.60 m (RP6 and GH6). Resistance to penetration was evaluated once by impact penetration meter at each site and calculated according to the following expression (Dalchiavon et al., 2011): (1) in which RP is the soil mechanical resistance to penetration (MPa); N is the number of impacts by the penetration-meter hammer for reading; and A and P are, respectively, penetration readings in the soil before and after impacts (cm). Deformed samples of soil were collected by shovel drill, with 0.10 m diameter and 0.20 m height, to determine GH (kg kg -1 ). The hydrogenionic potential (pH CaCl 2 ) of the soil and organic matter (OM) rates by organic carbon were determined from the same samples. The latter was obtained by the humid combustion method through calorimetry, by the following formula: in which OM is the rate of organic matter (g dm -3 ) and C is the rate of organic carbon. Soil samples were analyzed in the Physics and Soil Fertility Laboratories of the Universidade Estadual Paulista, Ilha Solteira, SP, Brazil. Plant factors comprised production of forage green matter (GM); production of forage dry matter (DM); rate of crude protein (CP); neutral detergent fiber rate (NDF); total digestible nutrient rates (TDN); and ash rates (ASH), collected above pasture height (0.20 m). Samples for the evaluation of the production of green matter of the Urochloa grass were collected by a 2 × 2 m metal square. The material within the square was harvested at approximately 0.20 m from the ground. A sample was then retrieved from the material, conditioned in a paper package, weighted and placed in a forced-air buffer at 55 0 C for 72 h. The production of the forage dry matter was determined. Neutral detergent fiber and TDN rates were determined according to methodology by Werner et al. (1997), whereas rates of crude protein and ashes were determined following Silva & Queiróz (2002).
Classical description analysis was performed for each factor under analysis by SAS (Statistical Analysis System, version 8.2.), in which mean, median, minimum and maximum rates, standard deviation, coefficient of variance, kurtosis, asymmetry, and distribution of frequency were calculated. Co-relation matrix was prepared to determine the simple linear co-relationships for combination, two by two, between the factors under analysis. The modeling of simple linear regressions was henceforth performed on Excel sheets. On the other hand, multiple linear regression of the dependent variables (plant factors) was performed by computer package of SAS ("stepwise") at 10% probability for the inclusion and exclusion of the variables in the model and compared to the independent variables (plants or soil factors). Consequently, those that provided the best equation to estimate the respective dependent variable would be selected.
Geostatistical analysis was undertaken by Gamma Design Software 7.0, following procedures by Dalchiavon and Carvalho (2012). Spatial dependence was analyzed for each factor by calculating the semivariograms, in which adjustments were performed in the first place by the initial selection of the lowest sum of the squares of the deviants (SSD); the highest coefficient of determination (r 2 ); and the highest evaluator of spatial dependence (ESD). Interpretation of ESD also followed method by Dalchiavon and Carvalho (2012): ESD < 20% = spatial variable with very low dependence (VL); 20% d" ESD < 40% = low dependence (L); 40% d" ESD < 60% = fair dependence (F); 60% d" ESD < 80% = high dependence (H); and 80% d" ESD < 100% = very high dependence (VH).
Co-krigings between plant factors and between the later and soil factors were performed.

RESULTS AND DISCUSSION
The descriptive analysis of the plant data and the soil chemical properties showed low variability for NDF, TDN, pH1, and pH2; medium for ASH; high for CP, OM1, and OM2; and very high for GM and DM (Table 1). As a rule, data were highly similar to those reported by Dalchiavon et al. (2013a) and Montanari et al. (2013b,c) when they investigated the variability of the chemical factors of soils cultivated with rice, Urochloa, and beans, respectively. The soil physical factors were medium (RP4 to RP6, UG1 to UG6) and high (RP1 to RP3) data variability has been registered ( Table 2). High variability for RP data up to 0.30 m is mainly due to management of reform and incorporation of corrective measures in the pasture and to the effect of   trampling by animals on these layers. The above indicates the soil heterogeneity evaluated by the high coefficients of variations and with the consequent interference in the forage production of the region. Souza et al. (2006), Dalchiavon et al. (2011), and Montanari et al. (2013a) also investigated the variability of the physical properties in managed soils with sugarcane, soybean, and Urochloa in their respective geostatistical studies and reported coefficients of variation similar to those of the current analysis.
With the exception of GM, DM, pH2, OM1, and OM2, which had a non-determined frequency distribution, and RP1 with log normal frequency distribution, all the other factors provided a normal frequency distribution (Tables 1 and 2). Probability varied between 0.050 (RP2) and 0.664 (UG3) and showed that centrally trend measures did not represent atypical rates in the distribution, which is typical of plant data, as underscored by Dalchiavon et al. (2013b).
Green matter (GM) and DM productivity were 5.39 and 1.50 t ha -1 , respectively (Table 1). Dry matter rate was actually low when compared with productivity reported by Montanari et al. (2013b,a). Low DM productivity was high due to the fact that the area was a continuous pasture at that moment, with more than 10 years after reformation and without any fertilization, with only one animal unit per hectare. In fact, it simulated the normal conditions used by the farmers in the reagion. However, there was a great decrease in mean annual productivity of U. decumbens in low fertility soils and rates between 1 and 2 t ha -1 of DM were normal (Serrão and Simão Neto, 1971). In fact, tropical forages provide dry matter according to clear-cut seasonality, or rather, between the rainy and dry seasons, and thus determine the lack of uniformity in the distribution of production throughout the year (Evangelista et al., 2004;Crusciol et al., 2012;Pariz et al., 2017).
The 7.59 percentage of CP in the DM (Table 1) showed that it was 15% higher than the rate reported by Montanari et al. (2013b), when these authors studied the same forage species in a Latosol in Selvíria, MS, Brazil. According to Milford and Minson (1966), CP rate in DM is adequate since only rates lower than 7% of DP in the forage actually show any reduction in the intake of DM by the animals. In fact, it reduces the digestibility of DM caused by nitrogen deficiency to the ruminal bacteria.
Means of NDF, TDN, and ASH were 67.18%, 55.75%, and 7.96%, respectively. It should be emphasized that the higher the ADF, the less is the digestibility. On the other hand, NDF has a negative co-relationship with forage intake, with 40% of ADF and 60% of NDF as the respective limits of digestibility and intake. As a rule, TDN rates were over 55%, which was considered the best by Van Soest (1994) in tropical forages.
In the case of the soil chemical properties, pH provided high acidity (pH between 4.4 and 5.0) with low OM availability, following classification by Raij et al. (1997). In fact, these rates were lower than 15 g dm -3 (Table 1) and partially explained the low production of the forage DM.
It should be highlighted that RP provided the lowest rates, considered medium (1.0 ≤ RP < 2.0 MPa) by Arshad et al. (1996), up to 0.10-m depth and between 0.30 and 0.60 m (Table 2). However, RP rates were considered high (2.0 ≤ RP < 4.0 MPa) between 0.10 and 0.30 m, since rates above 2.0 MPa in most cultures start showing limitations to the normal development of the root system (Tormena et al., 1998). The above rates clearly show the effects of natural thickening of the neighboring layers and the effects of animal trampling of the upper layers. The phenomenon mainly depends on soil class, humidity rate, animal density rates, mass of forage produced (deficient soil covering), and the forage species in the system (Marchão et al., 2007).
It is not only important to underscore RP rates but also the soil humidity conditions (GH) from which the data were obtained, since these qualities are normally inversely proportional (Dalchiavon et al., 2011). It must be emphasized that, at the instance soil data collection was done, GH was between 0.099 (UG3) and 0.108 kg kg -1 (UG1) ( Table 2) Studies on Pearson's linear co-relationships of DM and CP with the soil properties revealed the following, featuring relevant agronomy interest: DM had positive and significant co-relationships (n = 120) with GH3 (r = 0.221*), UG5 (r = 0.247**), and UG6 (r = 0.371**), agreeing with rates by Cavallini et al. (2010), who reported positive corelationship between DM and GH. Analogically, CP corelated negatively with RP3 (r = -0.192*), UG1 (r = -0.183*), UG2 (r = -0.308**), UG3 (r = -0.313**), UG4 (r = -0.328**), UG5 (r = -0.317**), and UG6 (r = -0.330**). The analyses of such co-relationships demonstrated that the higher the humidity rate in the soil, the greater the production of dry matter by forage. In fact, an increase in hydric availability would ensue and, therefore, a probable improvement in the absorption of the soil nutrients, with better vegetal growth. However, the forage CP would be lower since the co-relationship between DM and CP was significantly high (r = "0.331**), probably due to the effect of high N leaching from the mineralization of the soil OM. Potential (Equations 3 to 6) and linear (Equations 7 to 12) equations were modeled, which demonstrated a positive co-relationship between DM and GH, showing that increase in GH would have a positive effect on DM (Equations 3 to 5) due to a greater biomass production. Although tolerant to water deficits, U. decumbens still revealed seasonality in reproduction caused by drought or low temperatures.
Model (Equation 13) for the analysis of multiple linear regression of CP due to the soil physical and chemical properties explains approximately 21.9% of CP variation rate of U. decumbens (R 2 = 0.219**). The coefficients of the independent variables RP4 and RP5 were significant at 1%, whereas coefficients of GH3 and GH6 were significant at 5% probability. By means of the above regression and taking into account the mean rates of the independent factors (Table  2), CP rate of 7.60% may be estimated. Since this rate is close to the mean CP rate in current analysis (7.59%), the exactness of Equation 13 and its agronomic importance in estimating the parameter are confirmed.
1 # = attribute with data residue; parenthesis succeeding model = number of pairs in the first lag; exp = exponential, gau = gaussian, sph = spheric; pne = pure nugget effect. The coefficient of spatial determination (r 2 ) of simple semivariograms lay between 0.361 (#GH1) and 0.994 (OM2). It was the best semivariograph adjustment when analyzed by r 2 . Since the geostatistical extremes (A 0 ) lay between 107.0 (RP1) and 763.0 m (GH3), extreme rates for specific and localized managements that will be employed as reference in future research should not be lower than 107 m, since they are the distance within which the rates of a specific quality are equal. For spatial dependence evaluation (SDE), classification varied between low (#TDN; SDE = 28.6%) and very high spatial dependence (RP1; SDE = 95.8%).
The crude protein kriging map showed a high direct spatial co-relationship with OM1 and an inverse relationship with GH3 (Figure 2), especially within the central region where the highest CP rates occurred (7.2 to > 9.8%) (Figure 2b). The above coincided spatially with the highest OM1 rates (12.0 to > 16.8 g dm -3 ; Figure 2d) and with the lowest GH3 rates (0.080 to 0.096 kg -1 ; Figure 2e). On the other hand, lowest CP rates (4.7 to 7.2%) may be observed on the extremities of the maps, whereas OM1 rates were between 7.3 and 12.0 g dm -3 . Similarly, GH3 rates were included between 0.096 and > 0.112 kg -1 , with a slight compact in the zones. This fact is common on the margins of the area where trampling of animals and traffic of agricultural machines are highest. It should be emphasized that OM mineralization is the main N source since the relationship with CP is direct and, due to excess of humidity (highest GH), N leaching could occur and, therefore, an inverse response with forage CP rates.
Since co-kriging between CP and OM1 by the Gaussian model provided a limit of 318.7 m and high SDE (99.9%), it became clear that 72.2% of CP spatial variability may be   explained by the spatial variability of OM1 (Table 5, Figure  3a,b). In fact, the highest OM1 rates (Figure 3d) provided the highest CP rates (Figures 2b and 3b). The inverse is true. Consequently, from the spatial point of view of the researched area, expected CP rates in the sites where OM1 varied between 12.0 and 16.8 g dm -3 , lies between 6.9 and 10.4%. On the other hand, in the sites where OM1 lies between 7.3 and 12.0 g dm -3 , CP will be between 3.5 and 6.9%. This reveals the relevance of agricultural practices that aim at raising OM rates in the soil, since its benefits for the raising of CP rates for U. decumbens becomes evident through improvements in the physical (aggregation) or chemical (fertility) characteristics of the soil (Souza et al., 2004;Boeni et al., 2014;Rocha Junior et al., 2014).
Cross semivariogram (spherical model) and the cokriging map of CP due to GH3 proved that spatial variability of GH3 accounted for 84.8% of the spatial variability of CB (Table 5; Figure 3c,d). The sites where the highest GH3 occurred (0.096-0.112 kg kg -1 ) were precisely and inversely those in which CP provided the lowest rates (4.8 -7.7%), whereas the sites where the lowest GH3 rates (0.080-0.096 kg kg -1 ) occurred were the sites in which CP had the highest rates (7.7-10.7%). This shows the negative influence of the excessive humidity of the soil on the development of U. decumbens in sandy soil and confirms that the species analyzed is not tolerant to environments with high humidity rates, even though it showed low requirements with regard to soil fertility.
Nevertheless, from the point of view of space and soil management, it may be seen that the organic matter and humidity rates of the soil may be perfectly used as indexes of CP rates in the dry matter of U. decumbens. CP -crude protein; OM -organic matter; GH -gravimetric humidity.