Study on pH in water and potassium chloride for Bulgarian soils

Soil pH is commonly measured in water pH(H 2 O) or pH(KCl). The relationship between pH(H 2 O) and pH(KCl) across all Bulgarian soils were investigated and results examining the effect of soil type on the relationship were presented. Several functions were used to estimate dependence between the two measures. For all soils and depths, a linear regression accounted for 95.32% of the variation, which predicts pH(KCl) very well. From the analysis of data follows that they were differentiated into three clusters.


Introduction
Study on pH measured by different methods are in progress in different countries such as: Romania (Gavriloaei, 2012), Australia (Ahem et al. 1995;Minasny et al., 2011), Poland (Kabała et al., 2016). Unfortunately, there are several measures of soil reaction used worldwide. The most common extracts are distilled water (H2O), 1 mol L -1 KCl (KCl), and 0.01 mol L -1 CaCl2 (CaCl2). Different measurement methods lead, however, to incompatibility of data from various countries and disturb data integration in the international soil databases. In this article the relationship between pH(H2O) and pH(KCl) across all Bulgarian soils were investigated and results examining the effect of soil type on this relationship were presented. For all soils and depths, several regression equations were calculated, which predicts pH(KCl) in dependence of pH(H2O). The study was intended to help scientists and practitioners in using both methods of pH when dealing with problems of liming of acidic soils.

Material and Methods
Since 1956 the data from the large-scale soil survey have been used to compile soil maps of Bulgarian geographical regions at different scale. Thematic maps of the whole of Bulgaria have been prepared also to facilitate the soil agro-ecological partition at a scale of 1:600,000 (Yolevski et al., 1980), land evaluation for crop production at 1:1,000,000 scale (Kabakchiev et al., 1985). Until that time the so-called agro-ecological grouping of soils was adopted for the needs of agriculture. In Table 1 the total areas and arable areas are presented. The materials of the study are the values of pH(H2O) and pH(KCl) given in Reference database for soils in Bulgaria (Teoharov et al., 2009). This valuable source contains 306 data from different soils, namely: Chernozems (64), Gray Forest soils (33), Pseudopodzolic Forest soils (54), Cinnamonic Forest soils (33), Zheltozem soils (30), Leached Smolnitsa (15), Brown Forest soils (11), Mountainous Meadow soils (3), Alluvial Meadow soils (21), Peat-gley soils (11), Rankers (21), Regosols (9), Rendzinas (2), and Technogenic soils (17). Table 1. Agro-ecological groups of the total and arable area (Yolevski, 1986 Preliminary check of the data shows that three points were erroneous (Chernozem profile No. 32/3/C1k and C2k; Gray Forest profile No. 8 B2t) and were excluded from the analysis. The Technogenic soils also have been excluded. So we had 304 pairs for analysis. First step was performing descriptive statistical analysis. Results are given in Table 2 and 3. There is another collection of data with the properties of soils in Bulgaria (Ninov et al., 1975). Unfortunately, it is too limited and not all pH analyzes by both methods are included simultaneously for different soils.
Next step was performing regression analysis to describe the link between pH(H2O) and pH(KCl). The aim was to find the most appropriate function, which accurately describes the relationship between the values of both pH analyzes. It was also interesting to investigate the distribution of pH and its differentiation across the different soil groups. For this purpose cluster analysis was applied.
We are looking for a regression of the type where y = pH(KCl), x = pH(H2O) and f is a selected regression model.
We consider the following types of equations: Selected function have no more than three parameters to be estimated. The principle of Ocam is followed: "Of two competing theories, the simpler explanation of an entity is to be preferred" (Duignan, 2017). If you have a few hypotheses that could explain an observation, it is usually best to start with the simplest one.

Results and Discussion
The United States Department of Agriculture, formerly Soil Conservation Service (Soil Survey Staff, 1993) classifies soil pH in water ranges as follows in Table 4. FAO classification applicable in Bulgaria is given in (Gyurov and Artinova, 2015). Moderately alkaline Strongly alkaline 8.5-9.0 8.7-8.9 Alkaline Very strongly alkaline > 9.0 9.0-10.0 Strongly alkaline 10.1-11.0 Very strongly alkaline First, for each of the soils all models (a), (b), (c) and (d) are calculated. Results of corresponding correlation coefficients R 2 are presented in Table 5. Because of their small numbers of data Mountainous Meadow soils and Rendzinas are excluded from separate consideration, but they are included in the combined analysis. The R-Squared statistic indicates that the model as fitted explains 95.33% of the variability in pH(KCl). The adjusted R-squared statistic, which is more suitable for comparing models with different numbers of independent variables, is 95.32%. The standard error of the estimate shows the standard deviation of the residuals to be 0.33029. This value can be used to construct prediction limits for new observations.
If available data for pH are analyzed with KCl, it can be used the reverse analysis, where y = pH(H2O), x = pH(KCl) and gives the equation: pH(H2O) = 1.53466 + 0.88787 × pH(KCl) with R 2 = 0.9533.
It should not be forgotten that the analyzes of both methods -in water and potassium chloride, produce results with a certain error. It is logical to use the orthogonal regression method (Total least squares), which is appropriate in such case. That gives the equation: pH(KCl) = -0.076319 + 0.86878 × pH(H2O) with R 2 = 0.9536.
It should be noted that this method gives the same average value of the dataset, but with a smaller standard error equal to 1.2057. Figure 1 shows the straight lines of linear and orthogonal regression.  Obviously, it differs significantly from Normal (Gaussian) distribution and is a mixture of two or more different distributions. This justifies the use of cluster analysis to identify the groups that determine the differences in analyses. As a result of this analysis, three clusters were obtained, the statistical characteristics of which are given in Table 6.  Figure 4 shows a distinct differentiation between the three clusters. The presence of objects from the same soil group as members of different clusters can be explained by the large soil diversity in Bulgaria and some inaccuracies in the identification of soil profiles. The simple linear model equation (a) seems to perform very well and is almost accurate as the nonlinear models equations (b, c, d) across all the datasets. As a result, values of pH in KCl can be predict as a function of pH in water and vice versa. From a purely statistical point of view, it is advisable to use orthogonal linear regression, which takes into account the fact that in both methods the results are obtained with a certain error.

Conclusion
The analysis of 304 pairs results from soil samples is the basis for reliable conclusions. The results of this study allow soil reaction data obtained from different methods -with distilled water (H 2 O) and 1 mol L -1 KCl (KCl), to be converted and integrate into national and international soil databases. The difference between pH in water and pH in KCl is that the first refers to the acidity of the soil solution, while the pH in KCl refers to the acidity of the soil solution plus the reserve acidity in the colloids and therefore it is always more acid than pH in water. Regression between pH(H 2 O) and pH(KCl) is important because it gives the possibility for soil scientists to directly compare own values with the data already existing in literature from other country.
Monitoring pH changes over time is an important management tool. By comparing past and present soil tests, it is possible to see if the soil acidity is increasing over time and, if it is, to alter management methods to prevent this trend from continuing. Analysis of the results from the soil survey in Bulgaria shows that almost half of the soil resources are vulnerable to anthropogenic acidification. Special attention must be paid to genetically acid soils under cultivation. Their additional acid loading has to be controlled to avoid anthropogenic soil degradation.