1. Introduction
Existing increasing demands for agricultural products are driven by the growing World population and economic growth. Intensification of agricultural production causes several potential risks to water supplies. Nutrients from chemical fertilizers are more immediately available for plant uptake than in manure, but they may also be more easily leached into groundwater if used in excess [
1]. The use of chemical fertilizers in Thailand started to increase exponentially, increasing more than 100-fold from 18,000 tons in 1961 to 2,000,000 tons in 2004, but in spite of this massive increase in chemical fertilizer use, the yields of rice and maize have hardly increased. This suggests a tremendous loss of fertilizers into the environment due to their imbalanced use and poor management [
2].
Nitrate-nitrogen (NO
3−-N), which is an essential source of nitrogen (N) for plant growth, is now also considered a potential pollutant by the U.S. Environmental Protection Agency (EPA) [
3]. This is because excess applied amounts of NO
3−-N can move into streams by runoff and into groundwater by leaching, thereby becoming an environmental hazard [
4]. Many of the major sources of nitrate come from the use and production of fertilizers and waste materials, which are anthropogenic sources of nitrate contamination of groundwater [
5]. Although nitrogen exists in many forms, nitrate is the most available form to plants [
6]. Nitrate is very soluble in water, and it is readily carried to plant roots as the crop uses water. Soil nitrate unused during the growing season is free to move with water that percolates through the soil. This nitrate has the potential to contaminate groundwater if the water percolates beyond the root zone [
7]. NO
3−-N is a problem as a contaminant in drinking water, primarily from groundwater and wells, due to its harmful biological effects [
8]. The determining factor in the WHO’s decision to set the Maximum Contaminant Levels (MCLs) at 10 mg/L for the safety limit of drinking water [
9], and it was the occurrence of methemoglobinemia in infants under six months [
1]. Although the Maximum Contaminant Level for nitrogen was set at 10 mg/L nitrate-nitrogen, in 1976, EPA suggested that water having concentrations above 1 mg/L should not be used for infant feeding [
10].
Groundwater in Thailand is a source of household drinking water and supplements surface water for agriculture and livestock uses. Groundwater is used for the public water supply in 20% of the nation’s towns and cities and for half of the sanitary districts. It is estimated that 75% of domestic water is obtained from groundwater sources, and that they serve some 35 million people in villages and in urban areas [
1]. The nitrate concentration in the groundwater increased more than three times after fertilization, from 1.8 mg/L to 7.2 mg/L of NO
3−-N, contamination groundwater due to the application of excess fertilizers which have been applied for long time [
11].
Soils have varied retentive properties depending on their texture and organic matter content. Soil texture refers to the relative proportion of particles of various sizes in a given soil. Sandy soils have less retention than finer clay soils because sandy soils have less silt and clay [
12], which give rise to a lower cation exchange capacity (CEC) [
13], in the case of NO
3−-N allows it to leach into groundwater faster, so there is positive relationship between the percentage of sands and NO
3−-N concentration in groundwater wells [
14]. CEC is the sum total of exchangeable cations that a soil can absorb. Although NO
3−-N is an anion that can readily leach through the soil profile, soils with significant quantities of silt, clay, and organic matter will also retain more NO
3−-N than soils without much silt and clay. Soil texture also affects water permeability or percolation rate of a soil. Smaller amounts of silt and clay have higher water permeability rates than loamy sands or sandy loams [
4].
Remote sensing has the general advantage of providing spatially distributed measurements on a temporal basis. However, it mostly observes the surface of the Earth. There are also below ground remote sensing applications in fields such as geology, hydrogeology, mineralogy,
etc. Therefore, a link must be established between the surface observation and the subsurface (groundwater) phenomena. Physical features of the landscape such as alignments are detected in satellite images, and these provide valuable information for groundwater investigations [
15]. Aerial photography and visible and near-infrared satellite observations are widely used in groundwater exploration [
16–
18]. Spectral reflectance will be related with nitrate content in groundwater by indirect relationships.
The hypothesis presented herein is the fact that the quantity of nitrate nitrogen accumulated in groundwater through leaching by loam textured soils is higher than that of clay textured ones. This study aimed to estimate the effect of soil texture on NO3−-N content in groundwater. Optical reflectance data obtained by remote sensing was used in this study.
2. Materials and Methods
2.1. Data Preparation
The study site is Nakhon Pathom province in Thailand (
Figure 1) which is one of the monitoring provinces for land subsidence of Bangkok and vicinities. The province is divided into seven administrative districts. Most of the area are plains with no mountainous land. Plateaus are found in the west of Amphoe Muang and Amphoe Kamphaeng Saen. The plains are found along the Tha Cheen River. The study covers an area of 2,168 square kilometers, and an area of about 50% of this fertile land is mainly devoted to rice and fruit crops, and 8.16% is sugarcane[
19], thus most of the residents earn their living from agriculture.
LANDSAT TM-5 images acquired on 14 May 2010, scene center location (lat/long: 13.917, 100.100), were used to classify for landuse and extract the spectral reflectance. Groundwater analysis data was collected in May 2010 from 32 monitoring stations operated by the Department of Groundwater Resources (DGR). The spatial data layers such as soil unit and geology were provided by the Geo-Informatics and Space Technology Development Agency (GISTDA). The analysis of groundwater properties were derived from secondary data obtained from DGR. Several interpolations were implemented and then the most suitable results was used in the study. Soil texture and soil pH layers were reclassified from soil properties of the top soil layer.
2.2. Data Interpolation
Nitrate content (NO3−-N) from 32 monitor wells were converted to raster data by several interpolation methods in ArcView GIS (extension: Spatial analyst by ESRI and Kriging interpolation), such as IDW, Spline, and Kriging. The most suitable result for this study was selected by accuracy checking from each interpolation method. The extraction of grid value under the point includes checking points were provided. The details of each criterion are Inverse Distance Weighted (IDW) with p = 2; known as the inverse distance squared weighted interpolation, Spline: Weight (W) = 0.1, 0.5; Regular (R), Tension (T), Kriging: Spherical, Circular, Exponential, Gaussian, Linear, Universal1 (linear with linear drift), and Universal2 (linear with quadratic drift).
2.3. Spatial Autocorrelation Analysis
The interpolation results were analyzed by Spatial Autocorrelation Analysis. Initially, Moran’s I was used to assess global autocorrelation of the nitrate concentrations for the methods of analysis. Values of Moran’s I less than 0 indicated a negative spatial autocorrelation,
i.e., clustering of dissimilar values, while those greater than 0 indicated positive spatial clustering, that is, clustering of similar values in similar areas [
20]. For the interpolated concentration of NO
3−-N result which is a freeform polygon analysis, a binary weight matrix was created, using Queen Contiguity, identifying which areas were considered neighbors. This method takes into account those areas that share edges to the immediate left, right, up, and down as well as taking diagonal edges into account (reflecting how a queen moves in a game of chess). In this matrix, a “1” was assigned if location
i was the neighboring location
j, otherwise a zero was assigned [
21]. All spatial autocorrelation analyses were performed using the GeoDa software.
2.5. Statistical Analysis
By interpolation method, statistical analysis (ANOVA and DMRT) were applied to the process of best fit selection based on interpolation technique.
2.6. Software
Various software packages, namely ArcView GIS, Spatial analyst by ESRI, Kriging Interpolation Extension 2.01 [
22], ENVI, SPSS statistics, and GeoDa, were used in this study.
4. Discussion
The study approach followed spectral analysis of relationships between agricultural crops and nitrate concentrations in ground water by comparing two different spatial soil textures. The mean nitrate concentration of 1.0969 mg/L indicates that there is some human influence on nitrate concentrations in groundwater. Therefore, the results of this study should assist in the determination of significant sources of nitrate, helping in the estimation of fertilization practices to keep the levels within acceptable limits, lower than 1 mg/L.
4.1. Data Preparation
4.1.1. Landuse Class
Since most of the land in the study site was located in irrigated areas, an individual crop calendar was present in the variety. The main crops in the study site are rice paddy and sugarcane. As seen from
Figure 2, the adjacent crop area represents the different stages of planting in both rice paddy and sugarcane fields. These land uses frequently have nitrogen-based fertilizers applied to improve crop yield. Rice paddies are generally heavily fertilized, with a practical average of 40.63 ton/km
2 of 16-20-0 and 46-0-0 NPK, and sugarcane fertilizer is applied at a rate of is 46.88 ton/km
2 [
21]. Visual interpretation allows determination of the landuse cover: rice paddy field cover 29.4% (629.88 km
2) and sugarcane cover 10.4% (224.23 km
2) of whole study area.
4.1.2. Soil Texture
Although there are nine units of soil covered in the area, the top soil presented only two types of texture. The clay texture covers 1,264.56 km2 and loam covers 740.84 km2, which are related with the rice paddy and sugarcane.
4.1.3. Soil pH
Most of the area is found to be Class 4, which is the highest range of pH class in this study (pH 6.0–6.5) and it covered 862.49 km2, or 39.78 % of the provincial area. The lower pH areas cover 554.83 km2, 451.94 km2, and 125.32 km2 for pH 5.5, 6.0, and 4.5, respectively.
4.1.4. Groundwater Pond
Groundwater monitoring stations were distributed in the whole study area (
Figure 11). The maximum content of NO
3−-N in groundwater is 6.1 mg/L from the station PD0102, located in the urban area. Mean value is 1.0969, minimum is 0.45 mg/L (less than 0.9 mg/L according to the measurement by ion-selective electrode methods—Department of Groundwater Resources, Bangkok, Thailand, 2009, and the standard deviation is 1.1887.
4.2. NO3−-N Interpolation
The results from
Figure 7 show that several methods such as IDW1, spline3, spline4, krig1, krig3, and krig5 have a “bull’s eye effect”.
Table 3 shows the observed group of means of the original measured NO
3−-N content from DGR. The comparisons of mean value of nitrate concentrations among interpolation methods were assessed using the fixed-effect models analysis of variance (ANOVA). Each value point was extracted from the grid value under the point, including five check points. There were significant differences in nitrate concentrations among interpolation methods (P > 0.05), as revealed by Duncan’s Multiple Range Test (DMRT). The closest means from their residual when compared with the observed value was selected. Hence, Kriging with Gaussian criteria (KRIG4) was selected to be the most suitable result, as shown in
Figure 8.
4.3. Spatial Autocorrelation Analysis
The nitrate-nitrogen layer was classified by spatial autocorrelation analysis, which was compared by the mean of local and neighborhood of each other, and then classified into four classes, giving a high local and high neighborhood (HH), high local and low neighborhood (HL), low local and high neighborhood (LL), and low local and high neighborhood (LH). This was represented in the 1st, 2nd, 3rd, and 4th quadrant of the graph, respectively. Spatial autocorrelation analysis of NO
3−-N from Kriging interpolation with Gaussian for whole study area as shown in the
Figure 9(a), where the map is clustered into two big paths, up and down. However, the reclassifications of the clusters from separated High-High and Low-Low values in
Figures 9(b) and (c) were shown to be more detailed and clustered, with Moran values of 0.8316 and 0.9548, respectively. The local NO
3−-N clusters from HH and LL were divided into four classes of nitrate. Soil pH and soil texture were reclassified into four and two classes, respectively, from a soil unit of the study area from soil map scale 1:100,000 and use only the top layer of each soil unit considering the root zone of the typical crops (30 cm). Landuse class focused on agricultural crop with high nitrate-nitrogen practice which gave two classes of layers. The multiple classes of layer which will be input for the intersection analysis and the number of data layer classes are shown in
Table 4.
Although dissolved nitrogen will have the highest concentrations in soil with pH 6–8, the scatter plot combination of NO
3−-N (Kriging-Gaussian) and soil pH (
Figure 12), had no significant correlation between soil pH and NO
3−-N content in groundwater (correlation slope = −0.0125). Hence, the soil pH class was not implemented in the intersection and spectral extraction process.
The average of difference method was used to compare an average of nitrate concentration from five points with the observed value. Kriging interpolation provided good results for all criteria, but the most suitable method was selected by the minimum difference from mean of an observed value.
The experiment of the combination of NO
3−-N and landuse crops as shown in
Figure 13, where high nitrate-nitrogen content in groundwater of the study area were mostly found in the south to east of the study area, which is related to soil texture and landuse crops. However, the National Statistical Office has reported an increasing trend of fertilizer use in Thailand [
29]. The use of nitrogen fertilizer (16-20-0 and 46-0-0, NPK) in the study area is also very high. The spectral reflectance extraction was processed from LANDSAT imagery data, which was separated into difference groups by the combination of NO
3−-N data with soil texture and crops.
On the lower right photo is sugarcane, and upper right shows a rice paddy photo dated 20 May 2010. The black point of each crop is the point where the photo was shot and the color gradient represents the concentration of NO3−-N from groundwater, interpolated from KRIG4, processed by the spatial overlay with landuse crops (rice paddy and sugarcane).
5. Conclusions
The nitrate was classified into four classes by spatial autocorrelation analysis (Moran’s I and Local Moran statistics; LISA), by a means comparison. The cluster map legend contains five categories: Not significant (Areas that are not significant at a default pseudo significance level of 0.05), High-High (High values surrounded by high values), Low-Low (Low values surrounded by low values), Low-High (Low values surrounded by high values), and High-Low (High values surrounded by low values). There were 2 classes of the global clustered of NO3−-N with values of Moran’s I of 0.8316 (p = 0.05) for HH, and I = 0.9548 (p = 0.01). However, the local Moran’s statistics shown HH-hh (I = 0.8182, p < 0.01), HH-ll (I = 0.3486, p < 0.01), LL-hh (I = 0.6534, p = 0.01), and LL-LL (I = 0.4065, p < 0.5). The effect of soil texture on nitrate-nitrogen content in groundwater was directly observed by its reflectance values through remote sensing. It was found that NO3−-N measured through the loam in sugarcane (I = 0.0054, p < 0.05) was lower than clay represented in paddy (I = 0.0305, p < 0.05). This had a significant negative impact on the assumption, the quantity of nitrogen leached into groundwater through loam was higher than through clay.
According to the research [
2] and local statistical data [
19], farmers always apply excess fertilizer to paddy fields. This is a main reason for the higher quantity of NO
3−-N found in clay than in loam in this study. This case might be an exceptional study in terms of the quantity of fertilizers applied to agricultural fields. There was high level of NO
3−-N contaminants in urban areas, showing that there are other sources of contaminants. Therefore, there is a need to investigate the combined and multiple sources of contamination in urban areas that can cause hazard to urban populations.
