(**a**)

**Figure 5.** *Cont*.

**Figure 5.** Soil organic carbon (Mg·C·ha<sup>−</sup>1, bar area) and *K*-value (line) of under different toposequences, and trend analysis charts of soil organic carbon and *K*-value. (**a**) Upland rice area, (**b**) terraced paddy field, (**c**) trend analysis chart of *K*-value and (**d**) trend analysis chart of SOC stock. **Figure 5.** Soil organic carbon (Mg·C·ha−<sup>1</sup> , bar area) and *K*-value (line) of under different toposequences, and trend analysis charts of soil organic carbon and *K*-value. (**a**) Upland rice area, (**b**) terraced paddy field, (**c**) trend analysis chart of *K*-value and (**d**) trend analysis chart of SOC stock.

*3.3. Soil Erodibility under Terraced Paddy Field and Upland Rice*  Soil erodibility is used to calculate the *K*-value in the universal soil loss equation (USLE) and the revised universal soil loss equation (RUSLE), which is an important factor for soil erosion assessments as well as soil and water conservation planning [76,77]. In the present study, there was no significant difference in *K*-values, ranging from 0.2261–0.2893 t·h·MJ−1·mm−1 and 0.2238–0.2681 t·h·MJ−1·mm−1 between terraced paddy and upland rice soils (*p* > 0.05, Table 1), respectively. In upland rice soil, the difference in *K*-values across different toposequences was not significant (*p* > 0.05, Table 1). We found that the *K-*value was slightly higher when SOC content dropped, especially in U1 and U25 (Figure 5a). As noted previously, the finer particles tended to be dominantly exported by erosion [78]; thus, SOC was relatively eroded [79,80]. Conversely, soil erodibility was significantly variable under the terraced paddy soil (*p* < 0.01, Table 1). The highest *K*-value was, as expected, detected in T24 (lowermost location point), which contained the lowest SOC stock (Figure 5b). This can probably be explained by particle distribution differences being reduced due to the terracing technique. Moreover, SOC stock was possibly conserved under terracing cultivation by its contribution to plant root distribution [61], and its impact varied strongly among management practices [81]. This is because puddling in land preparation and transplanting method for rice planting caused a decrease in bulk density and enhanced root length density [82]. However, our findings merely reflect a tendency that was not a significant difference between upland rice and terrace paddy soils. As presented in Table 1, SOC stocks in the terraced paddy and upland rice fields were 21.84 and 21.61 Mg·C·ha−<sup>1</sup> , respectively, but a significant difference was not detected. Previous studies [37,66,67] have reported that terracing can reduce SOC loss by modifying hillslopes to small flat fields, which was supported by the findings of the present study. During the dry period after harvest, the rice straw remaining in the terraced field is a great practice to improve SOC sequestration as well as restoring soil nutrients with fewer losses from the field, which is conserved by the structure of terracing, such as stair-rice paddy. Moreover, flooding in the terraced paddy field supplies suspended particles and soluble nutrients to the fields [68], while puddling facilitates the incorporation of organic inputs into the soil and creates low breakdowns of OM [69]. On the other hand, the loss in upland rice may be higher than terraced paddy fields due to having no riser or wall to slow down erosion, together with fewer weeds and plants to cover the soil after harvest. Chen et al. [70] reported higher SOC in the surface layer (0–20 cm) than the deeper soil layer (20–100 cm), indicating that protection of surface soil of terraced field is the key to enhancing SOC [33,71]. As the meta-analysis by Chen et al. [72] points out, terraces in China increased 32.4% of SOC sequestration compared with sloping areas. Tadesse et al. [73], Arunrat et al. [74] and Arunrat et al. [75] suggested that the application of manure, crop residues and soil conservation could increase SOC. However, SOC stock in the present study reflects a visual tendency, but no significant difference was found, highlighting that continuous investigation is necessary to conserve soil nutrients and SOC sequestration.

#### As a significant difference of soil erodibility under terraced paddy soil was detected *3.3. Soil Erodibility under Terraced Paddy Field and Upland Rice*

(*p* < 0.01) (Table 1), it indicated the high fluctuation of soil erodibility among flat sections across the terraced paddy field. Moreover, the trend analysis chart of *K*-values along the topographical gradients between terraced paddy field and upland rice can be observed in Figure 5c. This is because each flat section of the terraced paddy field contains a different proportion of sand, silt, clay and OC contents, and vice versa under upland rice (Table 1). Thus, maintaining the fractions of sand, silt and clay contents as well as improving OC contents is the primary important factor to control the erodibility of both cropping systems. Among these factors, increasing OC content seems to be the most possible strategy and does not disturb current farmer's management practices by retaining rice straw and stubbles, applying animal manure and reducing tillage. Once SOC is increased, the benefits can result in the stability of soil aggregates and enhanced soil structure, resulting in Soil erodibility is used to calculate the *K*-value in the universal soil loss equation (USLE) and the revised universal soil loss equation (RUSLE), which is an important factor for soil erosion assessments as well as soil and water conservation planning [76,77]. In the present study, there was no significant difference in *K*-values, ranging from 0.2261–0.2893 t·h·MJ−<sup>1</sup> ·mm−<sup>1</sup> and 0.2238–0.2681 t·h·MJ−<sup>1</sup> ·mm−<sup>1</sup> between terraced paddy and upland rice soils (*p* > 0.05, Table 1), respectively. In upland rice soil, the difference in *K*-values across different toposequences was not significant (*p* > 0.05, Table 1). We found that the *K*-value was slightly higher when SOC content dropped, especially in U1 and U25 (Figure 5a). As noted previously, the finer particles tended to be dominantly exported by erosion [78]; thus, SOC was relatively eroded [79,80]. Conversely, soil erodibility was significantly variable under the terraced paddy soil (*p* < 0.01, Table 1). The highest *K*-value

was, as expected, detected in T24 (lowermost location point), which contained the lowest SOC stock (Figure 5b). This can probably be explained by particle distribution differences being reduced due to the terracing technique. Moreover, SOC stock was possibly conserved under terracing cultivation by its contribution to plant root distribution [61], and its impact varied strongly among management practices [81]. This is because puddling in land preparation and transplanting method for rice planting caused a decrease in bulk density and enhanced root length density [82]. However, our findings merely reflect a tendency that was not a significant difference between upland rice and terrace paddy soils.

As a significant difference of soil erodibility under terraced paddy soil was detected (*p* < 0.01) (Table 1), it indicated the high fluctuation of soil erodibility among flat sections across the terraced paddy field. Moreover, the trend analysis chart of *K*-values along the topographical gradients between terraced paddy field and upland rice can be observed in Figure 5c. This is because each flat section of the terraced paddy field contains a different proportion of sand, silt, clay and OC contents, and vice versa under upland rice (Table 1). Thus, maintaining the fractions of sand, silt and clay contents as well as improving OC contents is the primary important factor to control the erodibility of both cropping systems. Among these factors, increasing OC content seems to be the most possible strategy and does not disturb current farmer's management practices by retaining rice straw and stubbles, applying animal manure and reducing tillage. Once SOC is increased, the benefits can result in the stability of soil aggregates and enhanced soil structure, resulting in resistance to erosion [83]. Moreover, terraces have the potential to reduce sediment yield and runoff [84] as well as increase water infiltration, soil moisture and soil water holding capacities in several areas [21,71]. It can be observed in Figure 5d that SOC stocks decreased from the uppermost toposequence to the downward slope.

It should be noted that the *K*-value of the EPIC model is dependent on soil particle size and organic carbon, which may not be sufficient to estimate soil erodibility with the different climate and cropping systems. For example, the studies of Zhang et al. [85], Chen et al. [20] and Zhang et al. [86] have developed a database of K factors for China's agricultural soils to reduce the biases of soil loss estimation. Therefore, the study of the feasibility of combining methods (e.g., *K*-value from nomographs method [49], *K*-value of EPIC model method [50] and the soil geometric mean diameter method [51]) to provide accurate estimations of *K*-values in Thailand is required for future studies. Moreover, it should be constructed as Thailand's database of K factor and K factor maps.

#### *3.4. Correlation Coefficient Matrix and PCA Analysis*

The Pearson's correlation matrix among soil physicochemical properties is presented in Table 2. In terraced paddy soils, SOC showed a high positive correlation with both CEC (0.88, *p* < 0.05) and available Ca (0.85, *p* < 0.05). Available Ca had significant positive correlation with pH (r = 0.80, *p* < 0.05). Meanwhile, it was found that the available Ca positively correlated with CEC (r = 0.78, *p* < 0.05) and pH (r = 0.75, *p* < 0.05) in upland rice soils. A significant negative correlation was also found between clay and sand contents in upland rice soils and terraced paddy soils (r = –0.96 and –0.78, respectively). Interestingly, the negative values of correlation coefficients of the relationship between SOC or CEC and clay content were found in the upland rice field, indicating that clay minerals might not be active (Table 2).

These results, together with principal component analysis, explained 62.6% of the total variance (Figure 6, PC1: 46.3% and PC2: 16.3%) and allowed a better understanding of the correlation between the physicochemical properties of the soils collected in different cropping systems. Additionally, factor loading analyses showed that the first 6 of 16 PCs can explain 90.5% of the total variance with eigenvalue greater than 0.7. The significant loading factors with 10% of the highest factor loading in each significant PCs are underlined in Table 3. Three variables, available Mg, CEC and sand content, were obviously weighted in PC1. Meanwhile, in PC2, SOC, soil erodibility and toposequence were considered significant. In contrast, PC3 was strongly related to EC, NH4-N and NO3-N. Both EC and NO3-N were significantly

included in PC4 and PC5. While pH and EC were considered in PC4 and PC5, respectively. In PC6, soil erodibility, clay content and bulk density were significantly weighted.

**Table 2.** Pearson's correlation matrix of soil properties in terraced paddy field samples (*n* = 72, white background) and upland rice (*n* = 75, grey background).


\* Correlation is significant at 0.05 probability level (*p* < 0.05); BD = bulk density; SOC = soil organic carbon; ECe = electrical conductivity; CEC = cation exchange capacity; P = available P; K = available K; Ca = available Ca; Mg = available Mg.

**Figure 6.** Principal component analysis (PCA) and the loading values of soil properties and toposequences for terraced paddy field samples (red area) and upland rice samples (blue area). **Figure 6.** Principal component analysis (PCA) and the loading values of soil properties and toposequences for terraced paddy field samples (red area) and upland rice samples (blue area).

**Table 3.** Principal components (PCs), eigenvalues, percentage of variance explained by the PCs (% Var.) and cumulative percentage of variance explained by PCs of soil properties and toposequences

Eigenvalue 7.408 2.607 1.502 1.196 0.909 0.866 Var. (%) 46.301 16.296 9.386 7.473 5.681 5.415 Cum. var. (%) 46.301 62.598 71.984 79.457 85.138 90.553

Toposequence 0.022 **0.378** 0.266 **0.442 −0.409** 0.310 pH −0.103 −0.337 0.276 **0.493** −0.186 **−0.330**  Bulk density −0.325 0.201 0.051 −0.124 −0.053 −0.009 SOC −0.050 **−0.541** 0.158 −0.169 −0.141 0.044 ECe 0.129 −0.005 **−0.537** 0.263 **−0.552** 0.003 CEC **−0.348** −0.125 0.065 −0.007 0.011 0.049 NH4-N −0.170 0.024 **−0.506** −0.286 −0.327 −0.291 NO3-N −0.125 0.001 **−0.396 0.445 0.494** −0.239 P 0.289 −0.273 −0.120 −0.107 0.015 0.175 K −0.329 0.116 0.087 0.032 −0.139 −0.003 Ca −0.287 −0.273 0.055 0.198 −0.034 −0.205

for terraced paddy field and upland rice samples.

Factor loading/eigenvector


**Table 3.** Principal components (PCs), eigenvalues, percentage of variance explained by the PCs (% Var.) and cumulative percentage of variance explained by PCs of soil properties and toposequences for terraced paddy field and upland rice samples.

BD = bulk density; SOC = soil organic carbon; ECe = electrical conductivity; CEC = cation exchange capacity; P = available P; K = available K; Ca = available Ca; Mg = available Mg.

As shown in Figure 6, the result of PC1 shows the direct correlation among some properties (bulk density, available Mg, CEC, available K, available Ca, percent clay, percent silt, NH4-N and NO3-N), which were inversely correlated with percent sand, available P and ECe. A negative relationship between bulk density and sand content was detected. It indicated that upland rice soils were in higher bulk density, while the terraced paddy field had more sand content. It was clear that the upland rice and terraced paddy fields were different in the relationship between bulk density and sand content. Lower sand content and higher bulk density can make soil more fertile, while in the higher sand content area, more nutrients are expected to be gathered by irrigation. These results suggested that terraced paddies might be constructed to increase soil fertility rather than to reduce soil erosion in low upland crop production fields on natural hilly slopes. A strong correlation between the soil erodibility and toposequence was detected, indicating that higher soil erodibility occurred at lower toposequence points. As expected, by using the *K*-values of the EPIC model method, the correlation between soil erodibility and SOC was the opposite, meaning that a decrease in SOC stock was related with increasing soil erodibility (Figure 6). This is in line with the studies of Shabani et al. [87] and Ostovari et al. [88], who found a significant negative correlation between soil erodibility and OM.

#### *3.5. Recommendations for Further Study*

The findings of our study were investigated from a specific area, which may not be stated with high confidence about the differences between terraced paddy and upland rice fields, especially SOC stock and soil erodibility. Our study can be simply stated that SOC stocks and soil erodibility were not significantly different between the terraced paddy and upland rice fields in our study area. Therefore, more research should be conducted to validate the results in our study for providing appropriate management practices for these two systems. In future studies, experimental measurements should be conducted by measuring nutrient movement characteristics in terraced paddy and upland rice fields as well as soil erosion, water runoff and infiltration.

#### **4. Conclusions**

This study investigated hillslope cultivation fields that have been continuously managed as terrace paddy fields and upland rice cultivation, with the aim to explain how SOC sequestration, erodibility and physiochemical properties of topsoils are affected by terracing management. More variation in soil properties for each toposequence was found in terraced paddy soils rather than upland rice soils. Most soil nutrients (NH4-N, NO3-N, available K, available Ca and available Mg) in the terraced paddy field were lower than in the upland rice field. SOC stocks in the terraced paddy and upland rice fields were 21.84 and 21.61 Mg·C·ha−<sup>1</sup> , respectively, but a significant difference was not detected. Similarly, there was no significantly difference in soil erodibility between terraced paddies (range 0.2261–0.2893 t·h·MJ−<sup>1</sup> ·mm−<sup>1</sup> ) and upland rice (range 0.2238–0.2681 t·h·MJ−<sup>1</sup> ·mm−<sup>1</sup> ). Higher soil erodibility and lower SOC stock were found at the lower toposequence points of both cropping systems.

**Author Contributions:** Conceptualization, N.A., S.S. and R.H.; methodology, N.A., S.S. and R.H.; investigation, N.A., S.S. and P.K.; writing—original draft preparation, N.A. and P.K.; writing—review and editing, N.A.; supervision, R.H. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research project is supported by Thailand Science Research and Innovation (TSRI) and Office of the Higher Education Commission (OHEC) (MRG 6280146).

**Institutional Review Board Statement:** The study was conducted according to the guidelines of the Declaration of Helsinki and approved by the Institutional Review Board of Institute for Population and Social Research, Mahidol University (IPSR-IRB) (COA. No. 2019/05–154).

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Acknowledgments:** The authors extend their appreciation to Thailand Science Research and Innovation (TSRI) and Office of the Higher Education Commission (OHEC) (MRG 6280146) for supporting this research project. Furthermore, the authors would like to thank the reviewers for their helpful comments to improve the manuscript.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**

