Evaluation of CH 4 Emission in Two Paddy Field Areas, Khonkaen and Ayutthaya, in Thailand

Simple Summary: In anthropogenic activities, ﬂooded paddy ﬁeld is one of those events that emits CH 4 and that comes the necessity of leveling down CH 4 production from soil so the study on mechanisms of CH 4 production in paddy soil is vital. This study was conducted for an aim of investigating CH 4 inﬂuential factors in wide range of soil texture in paddy soil in Thailand and interrogating a model for predicted CH 4 production potential map. Our result showed that soil carbon and sandy soil are the important factors generating CH 4 production. To be exact, soil carbon works as a provider that implements an energy to soil microorganism to anoxically produce CH 4 and sandy soil, that carries less iron oxide, accelerates the CH 4 production. In conclusion, CH 4 production in paddy ﬁeld under ﬂooding period is triggered when large amount of soil carbon exists, and the reduction being processed quickly due to a smaller number of iron oxide in the soil, especially sandy soil. The knowledge of factors inﬂuencing CH 4 production brings about a better crop and water management. Further than that, predicted map of CH 4 production potential can be utilized to identify whereabout of highly emitted CH 4 production potential, which attention should be paid. Abstract: It is well known that submerged soils emit high levels of methane (CH 4 ) due to oxygen deprivation and free iron oxide causing a quick reduction. However, there are other soil properties that control the reduction processes in soil, especially the amount of soil organic carbon (SOC). This study aimed to investigate the major factors controlling CH 4 production potential (CH 4 PP) in Thai paddy ﬁelds. Two provinces, Ayutthaya, a clay soil region, and Khonkaen, a sandy soil region, were selected to represent a wide range of soil textures. Soil characteristic analysis pre- and post-incubation, and weekly gas detection in an incubation experiment over two months, was conducted. Stepwise multiple regression analysis was employed to analyze major soil factors controlling CH 4 PP. For the regional prediction of CH 4 PP, a map dataset of Ayutthaya and Khonkaen by the Land Development Department, Thailand, and a soil texture map (with intersected point data using the soil property map in


Introduction
In recent decades, causes and mitigation issues related to global warming have become controversial. Methane (CH 4 ) is a stronger greenhouse gas (GHG) than CO 2 because it has a higher radiative trapping ability [1]. Submerged soils emit high levels of CH 4 due to free iron oxides causing rapid reduction reactions. In agriculture, paddy fields are an important source of atmospheric CH 4 , as flooded conditions are preferable for proper rice growth [2,3]. Globally, rice cropping is considered to account for 5 to 20% of total CH 4 emission from anthropogenic actions [4]. Paddy fields in Thailand cover almost half the total agricultural land and are mostly located in the central and northeastern parts of the country. These regions have different soil types, with clayey soil in the central region, and sandy soil in the northeastern part [5,6]. Rice cultivation in Thailand was ranked as the fourth highest global CH 4 emitter, contributing an average of 1756.6 Gg CH 4 between 2010 and 2017 [7].
Several factors controlling CH 4 emissions have been studied, particularly in relation to plant species, cultivation practices, climate, and soil properties [8]. CH 4 production in the anaerobic environment of water-saturated soil depends on a variety of soil properties, both chemical and physical [8]. Under anaerobic soil conditions, methanogens process CH 4 through the reduction of soil oxides [2,3]. Following the depletion of oxygen, when soil microbes initiate anaerobic respiration, the order of electron acceptors use is NO 3 − , MnO 2 , Fe 2 O 3 , SO 4 2− , and CO 2 , while organic matter acts as the electron donor [9]. Inubushi et al. reported that hexose (an easily decomposable carbon component in soil organic matter) was positively correlated with the production of CH 4 in an experiment on 23 soil samples in paddy fields from four different, southeast Asian countries [10]. This result indicated the importance of soil organic carbon, and indeed, soil organic carbon is a significant factor controlling CH 4 production in paddy fields. Sass et al. showed that CH 4 emissions during the growing season in paddy fields with different soil textures were positively correlated to the percentage of sand (R 2 = 0.999) during a four-year study, and indicated that CH 4 production was higher in sandy soils than clayey soils when the same amount of organic carbon was applied to the field [11]. Wang et al. observed the relationship between CH 4 emission and soil physicochemical properties of 16 paddy fields from USA, India, Thailand, and Liberia in an incubation experiment, and revealed that the decrease in redox potential (Eh) was correlated with biologically reducible Fe, Mn, and soil pH [12]. Moreover, Mitra et al. showed that Eh was correlated with cation exchange capacity (CEC) and available potassium [13].
Even though the knowledge of processes that contributes to the CH 4 emission is well-reported academically, understanding of upscaling or spatial level of the emission is still inadequate [14]. Moreover, The U.S. Environmental Protection Agency (USEPA) reported in 2006 that the increasing world population affects the demand of rice which lays an impact on methane emission. Specifically, three-fourths of the emission is emitted from south east Asian countries [15]. This means the assessment of methane production potential is highly essential as it implies how soil reduction processes perform [16]. With the reliable CH 4 production equation that is put into potentiality and represented in spatial level, the result can imply the possibility of a future amount of CH 4 emitted from paddy fields in wide scale.
Therefore, by determining the major soil properties that dominate CH 4 production, a strategy for soil amendment can be determined to reduce CH 4 emissions from paddy fields. In Thailand, the soil texture of major rice paddies ranges widely from sandy to clayey. In this study, topsoil collected from paddy fields in the central and northeastern parts of Thailand were used to measure CH 4 production potential (CH 4 PP) in an anaerobic incubation experiment, and multiple regression models with several soil properties were Agriculture 2021, 11, 467 3 of 17 made. By interrogating the regression model, the regional difference in CH 4 PP between sandy and clayey soil areas was evaluated to investigate the effect of soil properties on CH 4 production in paddy soil.

Site Description
This study was conducted in the Ayutthaya province (14 • (Figure 1). Soil samples were randomly collected from a total of 44 paddy fields during June 2018, of which 20 samples (AY1 to AY20) were from Ayutthaya, and 24 samples (KK1 to KK24) were from Khonkaen and Mahasarakam (Table 1). Approximately 1 kg of composite soil from a depth of 0-15 cm was taken from five places in each field, mixed well, sealed in a plastic bag, brought back to the laboratory, air-dried, and used for the experiment. Separately, soil was sampled from a depth of 10 cm for the measurement of bulk density, and undisturbed core soil was sampled with a 100 mL stainless steel cylindrical tube.
Agriculture 2021, 11, x FOR PEER REVIEW 3 of 17 sandy and clayey soil areas was evaluated to investigate the effect of soil properties on CH4 production in paddy soil.

Site Description
This study was conducted in the Ayutthaya province (14°21′6″ N, 100°34′38.6″ E), and in the Khonkaen (16°26′22.6500″ N, 102°49′43.4208″ E) and Mahasarakam provinces (16°11′05″ N, 103°18′02″ E) (Figure 1). Soil samples were randomly collected from a total of 44 paddy fields during June 2018, of which 20 samples (AY1 to AY20) were from Ayutthaya, and 24 samples (KK1 to KK24) were from Khonkaen and Mahasarakam (Table 1). Approximately 1 kg of composite soil from a depth of 0-15 cm was taken from five places in each field, mixed well, sealed in a plastic bag, brought back to the laboratory, air-dried, and used for the experiment. Separately, soil was sampled from a depth of 10 cm for the measurement of bulk density, and undisturbed core soil was sampled with a 100 mL stainless steel cylindrical tube.  Ayutthaya province has an area of 2556.64 km 2 in total and is in the central plain of Thailand. It is surrounded by rivers (JICA, 2007) and has mean annual precipitation, annual raining days, and maximum rain amount of 1023.7 mm·year −1 , 107 days, and 60.6 mm, respectively [17]. Khonkaen province is in a mountainous area, covering a total area of 10,886 km 2 [18] with a mean annual precipitation of 1304 mm·year −1 , total raining days of 112 days, and 64.4 mm maximum rainfall [17]. Mahasarakam province is next to Khonkaen province and has an area of 5292 km 2 with 1225.1 mm of annual rainfall, 102 raining days per year, and a maximum rainfall of 65.7 mm [19]. Based on soil texture analysis of the samples using the pipette method [20], Ayutthaya has clayey soil, and Khonkaen and Mahasarakam have sandy soil. For the purposes of classification of soil texture and location, Khonkaen and Mahasarakam were amalgamated as one province named Khonkaen. Sampling was conducted in June 2018.

Anaerobic Incubation Experiment
A 15 g sample of air-dried soil and 30 mL of deionized water were placed in a 100 mL bottle, 4 cm in diameter and 12 cm in height, following which the headspace was purged using N 2 (Hokkaido Air Water Inc., Hokkaido, Japan) to create anaerobic conditions ( Figure 2). Two replicates were incubated during a period of 2 months at a temperature of 25 • C. During the incubation period, CH 4 was measured weekly using gas chromatography with a flame ionization detector (GC-14B, Shimadzu, Kyoto, Japan), and the vial bottles were flushed with N 2 after each gas detection to retain anoxic conditions. CH 4 production potential was calculated from fluctuation of gas concentration in the incubation bottle as follows in Formulas (1)-(3): where, F is the gas emission(mg·C·kg −1 ·day −1 ); ρ is the density of gas at the standard condition (CH 4 = 0.717 kg m −3 ); gas concentration (ppmv); V(m 3 ) is volume of the bottle; W(g) is dry soil weight; α = is the conversion of factor for CH 4 to C (12/16). To calculate average cumulative CH 4 emission, the gas emission was utilized in the formular below: where, C is the last cumulative gas result; D is the number of days in the sampling interval. where, F is the gas emission(mg·C·kg −1 ·day −1 ); ρ is the density of gas at the standard condition (CH4 = 0.717 kg m −3 ); gas concentration (ppmv); V(m 3 ) is volume of the bottle; W(g) is dry soil weight; α = is the conversion of factor for CH4 to C (12/16). To calculate average cumulative CH4 emission, the gas emission was utilized in the formular below: Average cumulative CH emission (gC · kg ) = F + (C D) where, C is the last cumulative gas result; D is the number of days in the sampling interval.

Soil Properties Analysis
Soil was analyzed before and after incubation. Before incubation (BI), air-dried soil and deionized water were placed in a plastic container for the recording of 1:10 pHBI using a pH meter (pH meter F-22, Horiba, Kyoto, Japan), and electrical conductivity (1:10 ECBI)

Regression Model for CH 4 PP
Significant factors controlling CH 4 PP were analyzed using the Pearson's correlation matrix and a step-wise multiple regression analysis in IBM SPSS Statistics for Windows, version 21 [23]. The comparison of variable means between the areas was determined using one-way ANOVA at 95% confidential level.
Multicollinearity usually occurs with a number of independent values that are highly correlated. To overcome this problem, one each of the highly correlated paired independent parameters (at R 2 > 0.7) was removed, and the variance inflation factor (VIF) was calculated for the remaining parameters. Multicollinearity was prevented in independent variables with a VIF of less than 10 [24]. Therefore, the variables with VIF less than 10 were included in the stepwise regression to predict CH 4 PP.

Evaluation of the Regional Differences in CH 4 Production Potential
In order to compare the CH 4 production potential between the Ayutthaya clayey soil area and the Khonkaen sandy soil area, spatial CH 4 production potential was predicted according to Figure 3. Two experimental area map datasets and one in vitro predicted soil dataset were employed in order to invent predicted regional CH 4 PP map: (1) the soil properties map in point data was provided by the Land Development Department (LDD) of Thailand [25] consisting of organic matter (%), 1:1 pH, organic carbon (%), available phosphorus (mg·kg −1 ), and exchangeable potassium (mg·kg −1 ), and (2) Online OpenLandMap from GitHub (© LandGIS contributors, n.d.) in a raster data structure that represented sand content (%), and clay content (%). Ayutthaya and Khonkaen map datasets consisted of 1490 points and 4253 points, respectively. Because the map dataset of GitHub was in a raster data structure, the datasets were interpreted in ArcGIS Desktop version 10.1 [26].
Soil properties after incubation were higher than before incubation in 1:10 EC (9.06 and 5.9 mS·m −1 , respectively), 1:10 pH, ExNH4. In contrast, anions after incubation were lower than before incubation in PO4 − , SO4 2− , NO3 − , and Cl − . The multiple regression analysis for CH 4 PP showed that CH 4 PP was significantly correlated with SOC, 1:10 EC BI , ExNH 4(BI) , and sand content. However, 1:10 EC BI and ExNH 4(BI) were not available in LDD. Therefore, regression equations were produced to predict those parameters using the available soil properties in LDD. The equations were obtained as follows in Formula (4) and (5): The CH 4 PP prediction equation was used to predict CH 4 production potential at the sampling points in each province. An inverse distance weighted technique (IDW) was run in ArcGIS 10.1 for regional CH 4 production potential in the provinces. IDW was also appointed to display the significant soil properties of the areas spatially.

Results
As shown in Figure 4a, there was no significant difference in the average CH 4 PP between Ayutthaya and Khonkaen (2012.31 and 1742.81 mg·C·kg −1 ·week −1 , respectively, p > 0.05). For the soil properties measured before incubation, Ayutthaya was significantly greater than Khonkaen (p ≤ 0.05) in 1:10 EC BI (10.69 and 1.97 mS.m −1 ), SOM (4.26% and 1.27%), SOC (2.46% and 0.74%), ExNH 4(BI) (17.07 and 12.38 mg·kg −1 ), ExK, ExCa, ExMg, ExNa, Total N, CEC, SO 4 2− -S BI , and silt. On the other hand, Ayutthaya was significantly lower than Khonkaen in 1:10 pH BI (6.08 and 6.75), sand (4.38% and 44.62%) and NO 3 − -N BI . For the soil characteristics measured after incubation, Ayutthaya was significantly higher than Khonkaen in 1:10 EC AI , SO 4 2− -S AI , ExNH 4(AI) and IC, while Mn 2+ was significantly lower in Ayutthaya than in Khonkaen ( Figure 5). There was no significant difference between Ayutthaya and Khonkaen in Fe 2+ and WSOC (Figure 5c,j).   Asterisk (*) represents a significant difference at the p < 0.05 level between the 2 areas. PO 4 3− -P is not shown.   Soil properties after incubation were higher than before incubation in 1:10 EC (9.06 and 5.9 mS·m −1 , respectively), 1:10 pH, ExNH 4 . In contrast, anions after incubation were lower than before incubation in PO 4 − , SO 4 2− , NO 3 − , and Cl − . Results of the simple regression analysis showed that although there was no significant correlation between CH 4 PP and each soil property measured before incubation, except for ExNH 4(BI) (Table S3), CH 4 PP was significantly correlated with 1:10 pH AI , Fe 2+ , and WSOC measured after incubation ( Table 2). Using the soil characteristics measured before incubation, the result of step-wise multiple regression analysis showed that CH 4 PP was significantly correlated with 1:10 EC BI , SOC, ExNH 4(BI) , and sand content (Figure 6), and the multiple regression equation indicated that all of the soil properties except for 1:10 EC BI were positively correlated with CH 4 PP without multicollinearity due to a VIF < 10 ( Table 3). After incubation, however, the result of multiple regression analysis showed that only 1:10 pH AI was significant (R 2 = 0.22, p < 0.01). Results of the simple regression analysis showed that although there was no signif cant correlation between CH4PP and each soil property measured before incubation, ex cept for ExNH4(BI) (Table S3), CH4PP was significantly correlated with 1:10 pHAI, Fe 2+ , an WSOC measured after incubation ( Table 2). Using the soil characteristics measured befor incubation, the result of step-wise multiple regression analysis showed that CH4PP wa significantly correlated with 1:10 ECBI, SOC, ExNH4(BI), and sand content (Figure 6), an the multiple regression equation indicated that all of the soil properties except for 1:1 ECBI were positively correlated with CH4PP without multicollinearity due to a VIF < 1 (Table 3). After incubation, however, the result of multiple regression analysis showe that only 1:10 pHAI was significant (R 2 = 0.22, p < 0.01). Figure 6. Multiple regression scatter plot between CH4 production potential (CH4PP) and major soil properties before incubation. Table 2. Correlation matrix among CH4 Production Potential (CH4PP) and soil properties (after incubation). Multiple regression scatter plot of measured and predicted CH 4 PP Figure 6. Multiple regression scatter plot between CH 4 production potential (CH 4 PP) and major soil properties before incubation. The regional CH 4 PP predicted by the multiple regression equation using the map parameter values was not different between Ayutthaya and Khonkaen ( Table 4). The differences in SOC, ExNH 4(BI) and 1:10 EC BI were significantly larger in Ayutthaya than in Khonkaen. Consistently, the sand content in Khonkaen was more than twice that of Ayutthaya (49.76 and 21.01%). In Ayutthaya, 1490 variables were used, and 4253 variables were used in Khonkaen. The distribution map of predicted CH 4 PP and the soil properties as the significant controlling factors of CH 4 PP in Ayutthaya and Khonkaen are shown in Figures 7 and 8, respectively. The distribution patterns of CH 4 PP corresponded well to that of SOC in both regions (Figure 7a,b and Figure 8a,b), but corresponded less to ExNH 4(BI) in Khonkaen than Ayutthaya (Figures 7c and 8c). There was an inverse relationship between 1:10 EC BI and CH 4 PP (Figures 7d and 8d), especially in the area of low sand content (Figures 7e and 8e).

Discussion
Sand content has been suggested as a significant factor controlling CH4 emissions because of the low oxide content [11,27]. However, SOC and other soil properties related to redox potential have also been reported as significant factors controlling CH4 emissions

Discussion
Sand content has been suggested as a significant factor controlling CH 4 emissions because of the low oxide content [11,27]. However, SOC and other soil properties re-lated to redox potential have also been reported as significant factors controlling CH 4 emissions [28,29]. In the current study, the mean value of CH 4 PP and more than half of the soil properties were statistically higher in Ayutthaya clayey soil than in Khonkaen sandy soil (Figures 4 and 5). Therefore, soil properties is significantly related to CH 4 emission [3]. Soil properties measured before incubation did not correlate with CH 4 PP in simple regression (Table S3). A regression equation was obtained by stepwise multiple regression analysis relating CH 4 PP to the soil properties (Table 3). In the regression equation, SOC, sand, and ExNH 4(BI) contributed positively to CH 4 PP, but 1:10 EC BI contributed negatively to CH 4 PP (Table 3). Within the soil properties measured after incubation, CH 4 PP showed significant negative correlations with 1:10 pH AI , and a significant positive correlation with Fe 2+ and WSOC ( Table 2). These observations are reasonable because the rise in pH indicates the presence of free iron, which suppresses CH 4 production. Increases in WSOC provide evidence of the decomposition of OC, leading to the production of CH 4 . Moreover, the incline of Fe 2+ implies the high reduction of Fe 3+ , which organic carbon is acting as electron donor causing CH 4 production. On average, of the properties 1:10 EC, 1:10 pH, ExNH 4 , and anions (PO 4 − , SO 4 2− , NO 3 − , Cl − ) measured before incubation, 1:10 EC, 1:10 pH, and ExNH 4 increased, but anions decreased after incubation.
The SOC in submerged soil acts as an electron donor in the reduction process, enabling methanogens to generate CH 4 by decomposing SOC [3]. It is well known that direct incorporation of rice straw into paddy field soils increases CH 4 significantly, and the application of cow manure contributes to higher CH 4 emission than inorganic fertilizer only [19,30,31]. Chidthaisong et al. observed the response in CH 4 production under anaerobic incubation to different fertilizer inputs, and found that CH 4 formation was the highest for plots with cow manure, following by rice straw, rice straw with chemical fertilizer, and chemical fertilizer alone [32]. These decomposed organic carbons can not only perform as electron donor to generate CH 4 production, but also can induce Fe 2+ production which can be speculated from increasing amount of Fe 3+ reduction [33].
Sass et al. observed CH 4 emissions in paddy fields in a wide range of soil textures and reported that an increase in sand content elevated CH 4 emission in anaerobic conditions [11]. In a comparison of flooded fields of clayey soil and sandy soils with the same organic carbon application, sandy soil emitted more CH 4 than a clayey soil [27]. Following oxygen depletion, when CH 4 is produced, the sequence of electron acceptor use is NO 3 − , MnO 2 , Fe 2 O 3 , SO 4 2− and CO 2, where organic matter acts as the electron donor [9]. Usually, sandy soil has a lower amount of free MnO 2 and Fe 2 O 3 as compared to clayey soil [34]. Thus, the reduction process occurs more quickly in sandy soil than in clayey soil. In the current study, CH 4 PP was similar between Ayutthaya and Khonkaen, and Fe 2+ production was not significantly different between Ayutthaya and Khonkaen (Figure 5c), and a higher amount of Mn 2+ production was shown in Khonkaen than in Ayutthaya (Figure 5d). On the other hand, SO 4 2− was reduced more in Ayutthaya than in Khonkaen. Therefore, SO 4 2− reduction might be a main suppressor of CH 4 PP in this study. The electrical conductivity of soil water extraction is a significant indicator of watersoluble ion concentrations [35]. In the present study, only SO 4 2− correlated significantly with 1:10 EC (Table 2). Figure 9 shows a positive relationship between 1:10 EC BI and SO 4 2− BI (R 2 = 0.34, p = 0.00003). As SO 4 2− reduction would occur immediately before the production of CH 4 [9], this clearly supports the negative influence that 1:10 EC BI has on CH 4 PP as shown in the multiple regression ( Table 3). The significant positive relationship between ExNH 4(BI) and CH 4 PP may be due to NH 4 + being the product of organic matter decomposition. This is supported by the significant positive relationship between WSOC and CH 4 PP (Table 2). over, in Ayutthaya, CH4PP tended to correspond to ExNH4, which could be because ExNH4 is a product of organic matter decomposition which organic matter is an important electron donor in soil anaerobic reduction providing energy for microbial metabolism. Elevated CH4PP occurred at low 1:10 ECBI, but 1:10 ECBI in Ayutthaya fluctuated more than Khonkaen. This is because 1:10 ECBI was influenced by SO4 2− reduction, one of the reduced substrates in the process indicating that lower electron contained in soil would accelerate soil reduction anaerobically. Ayutthaya soil contained a wide range of SO4 2− . A digital soil map (DSM) attributes soil properties and provides pedological knowledge to users [36]. Pedotransfer functions (PTFs) provide a means of using little soil information to create a meaningful and larger soil data image in DSM. A prediction equation from stepwise regression was determined in the current study, and the result is a form of PTF, but improved, in that it works on the prediction process and simultaneously enables users to learn pedologically fundamental knowledge.
Uncertainty in mapping prediction is a threat and leads to invalid conclusions. To avoid uncertainty, sufficient predictors, and a high density of data points are essential. A reliable regression prediction, which implies a statistically acceptable relationship between predictors and the predicted value, is required [36]. Referring to Table 2 and Figure  3, our study utilized various soil properties as well as sufficient points from the soil database. In addition, using multiple regression (a PTF) as a tool to predict unavailable soil properties in map data with a 99.99% confidence level (R 2 = 0.50), with VIF less than 10 at all predictors (Table 3), meant that half of sampling results confidently met the prediction made in the course of this study.

Conclusions
This study demonstrated the importance of soil properties on CH4 emission in Thai paddy fields. The spatial prediction result showed SOC, EC, ExNH4, and sand content are CH4 production influential factors in the study area. Moreover, our incubation experiment showed that CH4 production in this study is influenced by SOC, EC, ExNH4, sand content, Fe 2+ , SO4 2− , and pH. The above interaction of each soil property and CH4 production can be explained as follows: Because SOC is a derived form of organic matter, the increases in WSOC provides evidence of the decomposition of SOC which is important as it acts as electron donor in reduction process resulting the production of CH4. Referring to the role of SOC as electron donor, NH4 + indirectly supports the CH4 production because it is the product of organic matter decomposition. Furthermore, both CH4PP and Fe 2+ were found increasing simultaneously due to the increase of Fe 2+ which implies the increase of Fe As shown in Figures 7 and 8, the distribution pattern of the predicted CH 4 PP was strongly influenced by SOC in Ayutthaya, and sand content in Khonkaen. Higher CH 4 PP tended to be shown at higher SOC content in Ayutthaya because it acts as electron donor which supports anaerobic reduction process. A large amount of CH 4 PP found in areas with high sand content in Khonkaen indicates that sandy soil that contain lowest free ion among other soils can perform poorly at slowing down the reduction process [11]. Moreover, in Ayutthaya, CH 4 PP tended to correspond to ExNH 4 , which could be because ExNH 4 is a product of organic matter decomposition which organic matter is an important electron donor in soil anaerobic reduction providing energy for microbial metabolism. Elevated CH 4 PP occurred at low 1:10 EC BI , but 1:10 EC BI in Ayutthaya fluctuated more than Khonkaen. This is because 1:10 EC BI was influenced by SO 4 2− reduction, one of the reduced substrates in the process indicating that lower electron contained in soil would accelerate soil reduction anaerobically. Ayutthaya soil contained a wide range of SO 4 2− . A digital soil map (DSM) attributes soil properties and provides pedological knowledge to users [36]. Pedotransfer functions (PTFs) provide a means of using little soil information to create a meaningful and larger soil data image in DSM. A prediction equation from stepwise regression was determined in the current study, and the result is a form of PTF, but improved, in that it works on the prediction process and simultaneously enables users to learn pedologically fundamental knowledge.
Uncertainty in mapping prediction is a threat and leads to invalid conclusions. To avoid uncertainty, sufficient predictors, and a high density of data points are essential. A reliable regression prediction, which implies a statistically acceptable relationship between predictors and the predicted value, is required [36]. Referring to Table 2 and Figure 3, our study utilized various soil properties as well as sufficient points from the soil database. In addition, using multiple regression (a PTF) as a tool to predict unavailable soil properties in map data with a 99.99% confidence level (R 2 = 0.50), with VIF less than 10 at all predictors ( Table 3), meant that half of sampling results confidently met the prediction made in the course of this study.

Conclusions
This study demonstrated the importance of soil properties on CH 4 emission in Thai paddy fields. The spatial prediction result showed SOC, EC, ExNH 4 , and sand content are CH 4 production influential factors in the study area. Moreover, our incubation experiment showed that CH 4 production in this study is influenced by SOC, EC, ExNH 4 , sand content, Fe 2+ , SO 4 2− , and pH. The above interaction of each soil property and CH 4 production can be explained as follows: Because SOC is a derived form of organic matter, the increases in WSOC provides evidence of the decomposition of SOC which is important as it acts as electron donor in reduction process resulting the production of CH 4 . Referring to the role of SOC as electron donor, NH 4 + indirectly supports the CH 4 production because it is the product of organic matter decomposition. Furthermore, both CH 4 PP and Fe 2+ were found increasing simultaneously due to the increase of Fe 2+ which implies the increase of Fe 3+ reduction rate while additional SOC acting as an extra electron donor acceleratively causing CH 4 production. As sand content has the least iron oxide compared to other soil texture, it, therefore, leads to quick consumption of electron acceptors causing CH 4 production to increase. On the other hand, SO 4 2− is one of electron acceptors reduced in reduction process and the reduction of SO 4 2− implies the reduction of EC. This means the decrease of EC in soil increases CH 4 production. pH in soil implied the presence of iron oxide in soil. At higher pH, soil would attach more iron oxides which helps delaying reduction process which suppresses CH 4 production. To conclude, soil properties related to reduction reactions, as well as soil texture, are strongly influential in accounting for CH 4 PP in the paddy fields in the two regions in Thailand. Further than that, this study showed that the result of the multiple regression analysis exploring the soil properties that controls CH 4 production can be used as a pedotransfer function, which is able to predict a map of regional CH 4 production using the map data sets of soil properties.

Supplementary Materials:
The following are available online at https://www.mdpi.com/article/ 10.3390/agriculture11050467/s1, Table S1: Details of reagents used in the study, Table S2: Soil characteristics and analysis method before and after soil incubation, Table S3: Correlation matrix for CH 4 production potential (CH 4 PP) and soil properties (before incubation).