*2.5. Field Work, Soil Sampling and Laboratory Tests*

Selection of 17 sites for the comparison of soil samples collected from fallow land with samples from neighboring utilized agricultural fields was carried out according to the following assumptions:


**Figure 7.** Example of soil sampling strategy—site 17, where in the close vicinity were sampled: Two arable plots and three fallowed plots covered by goldenrod, wood, and bushes.

In total, 72 soil sampling points were identified for the 17 selected sites. Soil samples were taken from the top soil layer, 0–20 cm, where the reaction to the change in land use should be clearly visible. For each sample, the following characteristics were determined: Physical and chemical properties of soil, type of land use (arable land, grassland, and fallow type), and site affiliation.

The valuation class of the soil sampling site was preliminary determined based on the utilization class map available in the Spatial Information System of the Puławy poviat (SIS Pulawy) [34] and then verified by laboratory methods. National soil classification used in this work focuses on soil suitability for agricultural purposes, taking into account the morphological features and physicochemical properties of the soil, such as location and structure of the soil profile, water relations/conditions, and pH [35]. For better soil characteristics, each sample was described by its granulometric properties: pl (loose sand), pg (clay sands), ps (weak loamy sand), gp (sandy loam), and pyg (clay dust) [36,37]. Relation between the (Polish) classification used in this study and the international USDA classification was shown on Figure S1 (in the Supplementary part).

The scope of the performed analyses of soil samples included the basic physicochemical properties, including: pH in a 1-molar KCl solution, granulometric composition according to the norm: BN-78/9180-11 and PTG (2008) [38], which were determined by the laser method. The Egner–Riehm method was used to determine the content of assimilable forms of nutrients (phosphorus, potassium). The Egner–Riehm method is a chemical laboratory method. It involves extraction of the available forms of nutrients from the soil by means of special solutions, usually a buffer one. The same extraction solution is used for phosphorus and potassium. It is lactic acid buffered with calcium lactate, pH = 3.6. This solution is obtained by dissolving calcium lactate with hydrochloric acid. It is well buffered against both hydrogen ions and calcium ions—two factors significantly affecting the solubility of phosphorus compounds in the soil. Organic carbon (Corg) and Humus were determined using the modified Tiurin method [39].

#### *2.6. Statistical Analyses*

In order to verify the working hypothesis, the results of determination of the chemical properties of soils between the adjacent fallowed and used plots were compared. In the first step, corresponding pairs were selected (classes of pairs):


For all classes of pairs, differences in carbon content in soil, pH, and potassium and phosphorus content were assessed in comparison to arable plot. If there were more than one arable plot sampled on the site, then higher values were taken into account. Choosing the highest value in the significance test introduces a more restrictive approach because the working hypothesis is always rejected when the upper-tailed critical region is exceeded. In the first step of statistical analyses, the significance level (α) at which the criteria of the research hypothesis are met was estimated.

Principal component analysis (PCA based on correlations) was also used to capture the correlation between particular parameters in different types of land use [40]. This approach was to test whether the change in land use may affect the relationship between the parameters. Principal components analysis creates new artificial variables (principal components) based on the variables (features) that we analyze. Its main assumption was the possibility of visualizing the relationships of individual variables on a two-dimensional graph, which shows the coordinate system representing the first two principal components. Based on the position of the vectors in space, it can be determined which features are correlated with each other. The smaller the angle between the vectors, the stronger the positive correlation. When vectors are aligned on the same line but in opposite directions, there is a strong negative correlation between the variables. However, when the vectors are at an angle close to 90 degrees, no correlation occurs. Statistical analyses were performed using Statistica software package Statistica v13.1 (TIBCO Software Inc., Palo Alto, CA, USA).

#### **3. Result**

The summarized results of field work, remote sensing and laboratory tests are listed on Figure S2 (see Supplementary). For each site, it is specified: Chemical and physical properties for each tested land cover, soil granulometric type and fallow period. The conducted research confirmed the influence of the change in the use of agricultural land on its physicochemical properties. It should be noted that the specific granulometric composition and the valuation class within each of the sites were similar, which made it possible to attempt a comparison of the results of chemical analyses at the sites level. The vast majority of the tested samples belonged to the granulometric group of sandy loam—gp (31 samples) and clay sands—pg (25 samples). Only in sites no. 1, 12, 14, and 15 were the sampled soils classified as lighter groups such as: Loose sand—pl, weak loamy sand—ps and heavy group such as clay dust—pyg in site no. 4 (Figure S2). Detailed characteristics of the granulometric fractions are presented on the Ferret's triangle (see Supplementary, Figure S1).

### *3.1. Changes in Carbon Content*

The conducted research showed differences in the content of organic carbon within the sites in relation to different types of land use. The content of organic carbon in arable land ranged from 0.63% to 1.42%. It can be seen that a relatively high percentage of Corg was determined for site No. 12 despite its poor valuation soil class, which is associated with straw management in this field and manure fertilization. Newly abandoned land (FGL) with a predominance of grassy vegetation and several goldenrod instances showed an increased % Corg content, compared to arable land, on average by 32%, even up to 46.5%. Similar results were obtained by comparing arable land and permanent grassland (GL), where % of carbon content increased by max. 71.2%. In these soils, similar to meadow soils, the increase in carbon content is associated with the year-round cover of vegetation forming a compacted turf, which promotes binding of carbon from the atmosphere in the process of CO2 assimilation by these plants [41]. Different tendencies are observed in the

case of abandoned land dominated by species of later succession, in this case by goldenrod (*Solidago* L.). FGL species, for in this type of abandoned land one can see a clear decrease in carbon content even by 23.7% compared to arable land. On abandoned lands of the later succession, overgrown with FB bushes or trees, in sites similar to the FW forest, the carbon content in the studied samples is quite diversified. The highest increase, by 103.6%, was recorded for sites no. 14 and no. 7 and 10, by 95.2% and 90.4%, respectively. However, in addition to the increase in carbon content in this type of fallow land, we can also notice a decrease in carbon content in comparison with arable soils, which applies to sites no. 1, 3, and 17. In the case of site no. 1, carbon loss amounted at 52.3%, and the sample from this area was taken six months after the removal of bushy vegetation with single trees.

In general, it can be stated that in all researched sites, no case was observed in which the carbon content determined in the fallow fields, in relation to arable land, fell below the adopted critical value (1.7%). However, in most cases (except for fallow goldenrod—FG) an increase in the content of organic matter was noted. It should also be noted that in each considered case of the fallow type, values of standard deviations in the sample set (SD C%—presented in Table 1) are significantly smaller than the adopted critical value (1.7). A low variance in this case confirms that the observed relationships are not accidental.

**Table 1.** Standard deviations (SD) of chemical soil properties in various types of land use and various types of fallow land.


#### *3.2. Changes in pH*

The pH indicator in the top layer of the investigated arable lands was mostly very acidic < 4.6 and acidic 4.6–5.5, only in sites 3, 7, 12, and 17 the soil pH was more favorable slightly acid 5.6–6.5 (Figure S2). Comparing the average values of this parameter in individual types of use (Figure 8), we notice that the most favorable values were found in grasslands and grasslands fallow (pH = 5.3 and 5.1), besides with increasing pH, the percentage of organic carbon also increased.

**Figure 8.** Average chemical soil properties in various types of land use and various types of fallow land.

In the case of the remaining types of fallow land, the mean pH is lower in relation to arable land, and the lowest for perennial fallow land covered with older trees, where the beginning of natural succession is determined for the years 1996, 2006. Despite the lower pH in these lands, the average organic carbon content is about 30% higher than in arable land. The amount of organic carbon was restored through the annual deposition of organic matter from tree leaves.

In general, it can be stated that in all researched sites, four cases were observed in which the pH value determined in the fallow fields, in relation to arable land, fell below the adopted critical value (2). In three cases it was the site no. 17, which may indicate the specificity of this site and its lack of representativeness in relation to other sampling sites. If we reject these outliers, we can conclude that the research hypothesis has been rejected for 1.8% of cases (α < 0.05). Similar to the carbon content, in each considered case of the fallow type, values of standard deviations in the sample set (SD pH—presented in Table 1) are smaller than the adopted critical value (2.0). A low variance confirms that the observed relationships are not accidental.
