Impact of Land Use On The Variability of Soil Attributes Within Distinct Amazonian Environments

. Changes in soil attributes caused by the conversion of native forest for agricultural use 4 in the Amazon region is an area of research because of current uncertainties regarding land use and 5 occupation processes. These uncertainties are significant for tropical soils. Understanding changes in 6 soil attributes is vital for developing strategies to mitigate greenhouse gas emissions in the Amazon 7 region. The objective of this study was to investigate the impact of land use on soil attribute variability 8 occurring in distinctly Amazonian environments. This study was conducted using five meshes in 9 Southern Amazonas: Forest 1, cassava, sugarcane, Forest 2, and Archeological Dark Earth (ADE). 10 Descriptive statistical, geostatistical, and multivariate analyses were performed on data obtained from 11 local measurements of CO 2 emissions and data obtained from physical and chemical analysis of soil 12 layers up to a depth of 20 cm. Most physical, chemical, and biological attributes of the soil were 13 related to land use classifications. The similarity between cultivated and forested areas yielded no 14 evidence of land degradation resulting from land use. Increasing certain physical attributes total 15 porosity (PT), soil moisture (SM), and Macroporosity (Macro) yielded a greater increase in the CO 2 16 efflux for ADE and Amazon forest environments than for cultivated regions. 17


Introduction 21
Native forests are of fundamental importance for local biodiversity and the protection of soil 22 and water (Shorupa 2003). Human activities are changing soil characteristics. Such changes are 23 credited to several factors, such as agricultural exploitation, ranching, and forestry management 24 (Cohen et al. 2007). Significant increases in the degradation of natural resources are being observed 25 as a result of no or little consideration being given to soil management during aforementioned human 26 activities (Lima et al. 2009;Rockstrom 2009).
In forest environments, the removal of cover vegetation (e.g., deforestation) may disrupt the 28 natural state of the forest by modifying the chemical, physical, and biological attributes of the soil 29 (Wendling et al. 2012). This disruption is especially significant for low-fertility soils such as those 30 found in most of the Amazon region (Loureiro 2002;Cardoso et al. 2009). As a result, recent studies 31 have reported increases in soil density (Cunha et al. 2018). Ranching increases these impacts. 32 Additionally, nutrient leaching, increased CO2 efflux, and an abrupt reduction of biological activity 33 result from the scarcity of organic material input to soil. The loss of feedback in Amazonian 34 environments is demonstrated by these soil quality indicators. 35 Some studies have adopted geostatistical methods (

Statistics and mean test 116
The data were subject to a variance analysis (ANOVA) and the means were compared by the 117 Tukey test at a 5% probability level. This analysis was performed with the Minitab statistical software 118 (Minitab 2000) and had the purpose to detect the possible differences among the studied attributes in 119 varying environments. 120 Exploratory data analysis was performed by calculating the mean, median, minimum, 121 maximum, and coefficients of variation. The coefficient of variation (CV) was calculated based on 122 the criterion of Warrick and Nielsen (1980), who categorized the CV as low (< 12%), medium (from 123 12-24%) or high (> 24%). 124 125

Geostatistical analyses 126
We used Geostatistical analysis to characterize the spatial variability of our data. Under the 127 theory of the intrinsic hypothesis, the experimental semivariogram was estimated by the following The nugget effect is the semi-variance value for non-zero distances smaller than the smallest distance 136 between samples. It represents the random-variational component. The sill is the semi-variance value 137 for which the curve stabilizes to a constant value. The range is the distance from the source to the sill, 138 and is interpreted as the distance to which the samples start to be are uncorrelated (Trangmar et al. 139 1985). 140 To determine the significance of spatial dependence (SD) within the data, the semivariogram 141 exam was utilized using GS+ software (Robertson 1998). Multiple models were utilized to calculate 142 the same semivariogram for verification. The R 2 (coefficient of determination) criteria was employed 143 to determine the most accurate model. The Cambardella et al. (1994) classification was adopted to 144 analyze the correlation of soil attributes with spatial dependence; a strong spatial dependence is 145 indicated by comparing the nugget effect, resulting from semivariograms, to the sill value. Nugget 146 effects that are less than or equal to 25% of the sill indicate a strong spatial dependence, a nugget 147 effect between 25% and 75% of the sill indicates moderate dependence, and a nugget greater than 148 75% is indicates marginal dependence. 149 150

Multivariate statistical analysis 151
The data were subjected to principal component analysis (PCA) with the objective of reducing 152 the number of variables required to describe the variation in the attributes studied within the various 153 environments. As a result, most of the variance in the data was assigned to the first and second 154 principal components (PCs). The criteria used in the choice of PCs to be interpreted was the 155 percentage of attribute variance explained by each. 156 For this purpose, the initial set of 13 variables is reduced to two new latent variables (CP1 and 157 CP2). This greatly simplifies graphic illustration of the data, enabling two-dimensional figures 158 (ordering of indices by principal components). The suitability of this analysis is verified by 159 comparison of the transformed data to the data from the original variables;. The main components 160 contain eigenvalues superior to the unit and contain data that contribute significantly to the variance, 161 while eigenvalues inferior to the unit do not contain data that contribute significantly to the variance. 3 Results and discussion 166

Descriptive statistics and mean test
Results regarding the attributes evaluated are presented in Table 1. CO2 emissions in the soil 168 (FCO2) did not reflect any biological attribute for forest 2 or ADE. In contrast, a few differences were 169 observed in forest 1, cassava, and sugarcane crops. In forest 2, the FCO2 value was 7.00 µmol m -2 s -170 1 ; for ADE, it was 7.59 µmol m -2 s -1 . The mean values found in forest 2 were higher than those in data 171 presented by Xu and Qi (2001), who found high CO2 emissions   In this study, the CVs for FCO2 were 46.47%, 31.79%, and 59.61% for forest 2, ADE, and 241 forest 1, respectively. These CV values are high. Conversely, cassava and sugarcane areas displayed 12 CVs of 23.15% and 23.07 respectively; these are average (Warrick and Nielsen, 1980 The SM had a high CV value in all environments except forest 1, which displayed an average CV 245 value. As for St, the CV value obtained was low in all studied areas, indicating a low variability of 246 ST therein (Table 1). 247 Regarding the CV variations in anthropogenic environments compared to forest 2, ADE, 248 which has an anthropogenic heritage had a 47% reduction in variability for FCO2, In relation to soil 249 physical properties, expressed by BD, Macro, TP and clay content, anthropization promoted an 250 increase of 5 to 75% for variability. This variation can be related to the higher sensitivity of physical 251 attributes when soil is subjected to intense use (such as agriculture). Such intense use promotes 252 inversion of the soil surface layer and/or compaction. Both ADE and forest 1 environments presented 253 the greatest CV for physical attributes (17 a 71%). 254 Interestingly, the highest CV variation was observed in chemical attributes comparing forest 255 and anthropogenic environments. For pH values, the variation promoted by anthropization ranged 256 between 2% and 213%. OM ranged from 72% to 128% and available P from 29% to 121%. The 257 greatest variations occurred for SB, CEC, and V%, which varied from 31% to 1564%. Following the 258 same trend found for physical attributes, the ADE and forest 1 environments showed the greatest 259 variation. Chemical attributes are more sensitive to changes in environment; therefore, favoring the 260 greatest variation of these attributes. 261

Geostatistical analysis 262
A pure nugget effect (PNE) was observed during the geostatistical analysis for FCO2 in forest 263 2 (Table 2). It was also observed for SM, pH, and V% in ADE; for SM, clay content, pH, OM, SB, 264 and V% in forest 1; for clay content, CEC, and V% in cassava; and finally, . PNE may indicate that the utilized sample spacing is larger than necessary to detect spatial dependence for a given attribute (Cambardella et al. 1994). 269 In general, the evaluated attributes presented a spatial dependency, which was quantitatively Exponential models are capable of describing erratic phenomena on small scales, while spherical 273 models describe properties with high-spatial continuity which are less erratic at small scales (Isaaks 274 and Srivastava 1989). According to Trangmar et al. (1985) and Cambardella et al. (1994), these are 275 the most common theoretical models used to express soil and plant attributes. 276 The degree of spatial dependence (DSD) was modeled according to Cambardella et al. (1994).   Regarding the chemical attributes, forest 2 presented a large CV value, with exceptions for 288 pH and V%, which had a moderate CV, while the other environments presented large CV values. In 289 general, the large CV of soil properties is given due to intrinsic factors, while a weak dependency is 290 attributed to extrinsic factors (Cambardella et al. 1994). Therefore, a moderate or low CV might occur 291 due to the soil homogenization among the varying systems and adopted management in the areas 292 Due to the observed spatial variability given by descriptive statistical (Table 1) and 310 geostatistical (Table 2)

Multivariate statistical analysis 318
The multivariate structure contained in the original data set was evaluated in the "scree-plot" 319 graph and in the principal components analysis (PCA). The "scree-plot" graph ( Fig. 2) can be used 320 to verify the importance and contribution of each variable to explain the total variance of the data set.  Eigenvalue number PCA was used to identify the ability of the variables to account for the variance observed. In 337 this study, the two first principal components (PC1 and PC2) were considered. The eigenvalues for 338 each was greater than 1 (Kaiser 1958), and together the two variables were responsible for 82.14% 339 of the variability of soil properties. In this case, these components explain at least 80% of the total 340 variance to be accounted in decision-making. PC1 explained 64.36% of the total variance, whilst 341 17.78% of this variance was explained by PC2. The eigenvalue for PC1 and PC2 were 8.36 and 2.31 342 respectively, reinforcing the choice of using these two components. 343 The biplot graphical representation, which expresses existing correlations among variables 344 with the principal components, defined four well-differentiated groups as shown in   The variables BD (-0.865776) and clay (-0.826436) displayed correlations with PC1 and were 355 responsible for the differentiation of group II, cassava, and sugarcane. This characteristic relates to 356 the use of agricultural machinery used for soil preparation, fertilization, harvest. The use of 357 agricultural machinery increases BD, within a given group, provided the soil management is the same. 358 The clay content associated with groups II and III was high for these environments, particularly forest 359 1. The areas that formed group II are the most closely approximated by forest 1, due to the lower 360 variability provided by soil management, which occurs in the agricultural cultivation of cassava and 361 sugarcane. 362 In group IV, which presented correlations with both PC1 and PC2, the variables responsible 363 for its differentiation were TP (0.867287), Given the forest soils as references of soil quality, the similarity of group II to group I and III 374 support the statement that, for the conditions of the present study, there is no reason for concern in 375 relation to soil degradation processes. This is an important finding for sustainable agricultural 376 management which is recommended for the natural physical-chemical maintenance of Amazonian 377 soils. Thus, the continuation of studies similar to the current study are necessary. If soil degradation is a silent and slow process, this work is of importance of the Amazonas region, and consequently, 379 the world ecosystem. 380

Conclusions 381
Physical-chemical and biological soil attributes generally presented a structure of spatial 382 dependency that was coordinated by the type of land use. 383 The similarity among cultivated and forest areas indicated no evidence of land degradation 384 imposed by land use under subsistence activities. 385 The influence of high content organic matter in the soil upon the efflux of CO2 in ADE and 386 forest areas is significant compared to that of cultivated sites; consequently, this variable is useful as 387 an indicator of the preservation of Amazonian environments.  Schematic representation of the experimental areas located in the south of Amazonas state and sample collection Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors. Biplot graphical representation of the major components PC1 and PC2 resulting from the principal component analysis of physical, chemical, and biological properties of soil at a depth up to 0.20 m, in different southern-Amazonian environments. F1 = forest 1; F2 = forest 2; C = cassava; S = sugarcane; ADE = archaeological dark earth