In this paper, we present a novel approach, enabling the measurement of nitrate concentrations in natural soil porewater containing natural soil dissolved organic carbon (DOC). The method is based on UV absorbance spectroscopy, combined with fluorescence spectroscopy, for simultaneous analysis of DOC and nitrate concentrations. The analytical procedure involves deduction of the absorption caused by the DOC from the total absorbance in the UV range that is attributed to both DOC and nitrate in the water solution. The analytical concept has been successfully tested in soil water samples obtained from five agricultural sites, as well as in water samples obtained from a commercial humus soil mixture. We believe that the new analytical concept can provide a scientific foundation for developing a sensor for real-time nitrate concentration measurements in agricultural soils. As such, it can play a significant role in reducing nitrate pollution in water resources, optimizing input application in agriculture, and decreasing food production costs.
During the last half-century, clear trends of rising nitrate concentrations
in groundwater have been observed in aquifers all around the globe (Jin
et al., 2012; Kourakos et al., 2012; Liao et al., 2012). The World Health
Organization (WHO) has determined that nitrate levels in drinking water
should not exceed 50 mg L
Currently, agricultural fertilizer application relies primarily on farmers' experience, expert recommendations, and sporadic soil testing. None of these provide information that is in line with the timescale of N-fertilizer mobilization, consumption, and transformation dynamics in the soil. As such, continuous in situ nitrate measurement in the soil is essential for optimizing fertilizer application and reducing groundwater pollution potential (Šimůnek and Hopmans, 2009).
Presently, monitoring of soil chemical parameters is performed in water samples that may be obtained using suction cups installed in the soil or through water extraction of soil samples (Kabala et al., 2017; Reck et al., 2019). Water samples collected using these methods need to be transferred to a laboratory for further chemical analyses or direct analysis on-site using analytical kits (Golicz et al., 2020). Nitrate in the soil is highly soluble, mobile, consumed by plants, and subject to various biochemical transformations. Accordingly, nitrate concentration in the soil may fluctuate dramatically on timescales of hours to days in response to irrigation and precipitation patterns, fertilization, root uptake, and different plant growth phases (Mmolawa and Or, 2000; Turkeltaub et al., 2015). Thus, monitoring nitrate concentrations with conventional tools often does not meet the required time resolution for optimizing fertilizer application and preventing groundwater pollution from excess use. The use of highly sophisticated soil sampling methods can be expected from a dedicated research team but not from farmers who are focused on food production in large-scale agricultural setups (Thompson et al., 2007). Therefore, automated, real-time monitoring of soil nitrate concentrations may provide growers with vital information on nutrient availability. Such data are critical for maximized productivity, in addition to water resource protection from nitrate pollution.
Ultraviolet (UV) absorption spectroscopy is one of the most common methods for nitrate analysis in aqueous solutions (Mayerstein and Treinin, 1961; Moorcroft, 2001). Light absorption by nitrate in an aqueous solution takes place at two main wavelength bands on the UV spectrum: (1) the high absorbance band from 200–240 nm and (2) the low absorbance band from 280–340 nm. Tuly et al. (2009) suggested a continuous soil solution monitoring setup to measure nitrate concentration by applying absorbance spectroscopy techniques in a porous cup installed in the soil. This setup consisted of a porous stainless-steel cup filled with deionized water that was placed in a reservoir of a potassium nitrate solution. Once the solution inside the porous cup achieved chemical equilibrium by diffusion between the cup and the surrounding medium, the absorption spectrum of the solution was measured by a UV dip probe. The dip probe was connected to a spectrophotometer and a UV light source through optical fibers, which enabled continuous measurement of the solution within the suction cup. However, this method was limited by two main factors: (1) achieving chemical equilibrium between the porous cup and the surrounding medium is a slow process, especially in unsaturated sediment that has limited water storage. The time lag between the actual variation in soil nitrate concentration and its measurement in the porous cap precludes the real-time measurement of the rapid changes in the soil nitrate concentration, and (2) the presence of natural soil dissolved organic carbon (DOC) limits the UV absorption spectroscopy analysis since both nitrate and soil DOC absorb UV light in a similar wavelength range (Causse et al., 2017; Ferree and Shannon, 2001; Shaw et al., 2014).
DOC interference to the nitrate absorbance spectrum can be reduced by applying a combination of the UV–VIS absorption spectrum and multivariant statistical models such as partial least squares regression (PLSR) (Avagyan et al., 2014; Etheridge et al., 2014; Rieger et al., 2006). PLSR is similar to principal component regression (PCR), when dealing with multidimensional data sets. However, PLSR has the additional advantage of maximizing the correlation between the variables and a parameter of interest (Wehrens, 2011). Vaughan et al. (2017) had demonstrated the use of a PLSR-based method along with an in situ optical apparatus to estimate nitrate concentration in watersheds of urban, agricultural, and forested lands. However, the general DOC and nitrogen concentration in their research had ranged between 0–10 ppm, 1 order of magnitude less than the nitrate and DOC levels found in the shallow subsurface porewater of agricultural soils (Dahan et al., 2014; Yeshno et al., 2019).
Recently, a monitoring setup for real-time soil nitrate concentration measurement was developed, which is based on continuous spectral analysis of soil porewater (Yeshno et al., 2019; Dahan et al., 2020) in an optical flow cell that is connected to a specially designed porous cup (suction lysimeter). This enables the creation of a continuous low rate flux of soil porewater flowing from the soil through the cell. The absorption spectrum of the soil porewater in the optical flow cell is continuously recorded and analyzed to achieve real-time nitrate concentration measurements. The analysis involves an algorithm that scans the entire absorption spectrum of the soil porewater to identify an optimal wavelength where DOC interference with nitrate measurement is minimal. Use of the method showed that the optimal wavelength is site-specific, and the calibration equations for nitrate concentration are consistent over a long-time duration. Measurements conducted for 2 years in four different agricultural fields and through a series of column experiments offered detailed information on time variations in soil nitrate concentrations (Yeshno et al., 2019). Despite this significant breakthrough in continuous soil nitrate concentration measurement, the technology is still too cumbersome and impractical for commercial agricultural applications since it is based on a wide-band UV spectrophotometer and a bulky deuterium lamp as a UV light source.
Reducing the cost, size, and energy consumption of a spectroscopy-based
apparatus for measuring soil porewater nitrate concentrations requires a
shift in applied technology from the commonly used UV deuterium lamp to a
less expensive LED-based UV light source. In addition, the commonly used
wide spectrum spectrophotometer can be replaced with a more affordable
photodetector. However, currently, most affordable, commercially available
UV LEDs do not possess the range of the high nitrate absorbance band
(200–240 nm), but only of the low nitrate absorbance band in the range of
300
In this paper, we present a novel approach, enabling nitrate estimation in
porewater samples containing DOC with a variety of concentrations and
chemical compositions. The analysis is based on UV absorbance spectroscopy
in the vicinity of 300 nm combined with simultaneous fluorescence
spectroscopy with excitation/emission at
Soil samples from five different agricultural fields, located on the coastal plain of Israel, were collected and analyzed for this study. The samples were collected from organic and conventional greenhouses that grow vegetable crops, an open field that is used for rotating mixed crops, and a citrus orchard (Table 1). These sites were chosen to represent a spectrum of typical agricultural practices in different soils. In addition, a commercial humus soil mixture (Dovrat LIMITED, Israel) was also examined to represent the potential impact of DOC originating from compost enrichment on a soil porewater spectral analysis. These sites have been intensively studied in the past, and additional information can be found in previous publications (Dahan et al., 2014; Turkeltaub et al., 2014, 2015, 2016) and in the Supplement (Sect. S1).
Sampled field sites location and soil type.
n/a stands for not applicable.
Soil water extract's samples were obtained by creating a mixture of the
different soils with double distilled water (DDW). The soil extracts were
left to stand for 24 h in order to achieve chemical equilibrium of the soil
DOC. The soil and liquid phases were separated by a standard laboratory
centrifuge, and the suspended solids were removed by a 0.22
The initial values of DOC and total nitrogen (TN) in the sample were
estimated by an Analytic Jena multi N/C 2100s TOC/TN analyzer, while nitrate
concentration was determined by a Dionex ICS 5000 ion chromatograph. The
absorption of the samples at 300 nm was determined using a TECAN Spark 10M
multimode microplate reader spectrophotometer. The light absorbance was
defined by the Lambert–Beer equation (Eq. 1):
A fluorescence spectroscopy technique was used to measure the DOC
concentration in the examined solution using a TECAN Infinite M200
spectrophotometer with excitation (EX)/emission (EM) at
The chemical and optical data of the soil water solution were analyzed by the
MATLAB 2019b curve fitting tool to obtain the polynomial equation for nitrate
estimation,
The application of UV absorption techniques to aqueous nitrate solutions commonly results in a linear correlation between the absorption rate and the nitrate concentration (Edwards et al., 2001; Ferree and Shannon, 2001; Michael et al., 2017; Moo et al., 2016; Tuly et al., 2009). However, this correlation is not straightforward in natural soil water containing DOC, as UV absorption by the organic matter, in conjunction with the nitrate absorption, leads to an increase in the overall absorption in the examined solution. For example, a series of solutions with variable nitrate concentrations and a relatively constant DOC concentration shows a linear correlation with the absorption values (curve a in Fig. 1). However, a similar series of nitrate concentrations, with an increased DOC level, shows a linear correlation as well, yet with a greater absorbance level (curve b in Fig. 1). Nevertheless, the similarities between the trend line associated with the low DOC levels and the trend line associated with the high DOC levels imply that the DOC contribution to the overall absorbance is consistent and quantifiable.
Nitrate concentration vs. absorption at 300 nm of soil water samples from the conventional greenhouse.
However, examining the relation between nitrate and DOC concentrations and
the UV absorption of a large solution matrix may not be as intuitive as the
2D model presented in Fig. 1. In order to assess the specific contribution of nitrate and DOC to the total absorption at 300 nm, a matrix of solution containing varying concentration of DOC and nitrate was made from the soil water extract for each study site. The solution matrix was obtained through the preparation of a series of replicates of different DOC concentrations for each soil water extract. Each replicate was spiked with a variable volume of 10 000 mg L
Database obtained by analyzing the water samples extracted from the conventional greenhouse soil.
The data from the three variables (DOC/nitrate concentration and absorption
values) were projected on a 3D domain (Fig. 2). The data distribution in space, which is geometrically defined as a plane, was mathematically quantified by applying a multivariate regression model. From this model, the nitrate concentration in the sample could be estimated as a function of the DOC concentration and absorbance at 300 nm (Eq. 2):
3D projection of the nitrate concentration as a function of the DOC concentration and absorbance at 300 nm for water samples obtained from the conventional greenhouse.
Yet, the contribution of the DOC to the overall absorbance can differ from site to site, due to variations in DOC chemical and optical characteristics in the different soils. The composition of the organic matter that constitutes the DOC is affected by site-specific characteristics such as type of agricultural crop, biological activity in the soil, and the type of compost and fertilizer used at the site (Kalbitz et al., 2000; Nelson et al., 1992; Tian et al., 2010). Therefore, the polynomial calibration equation for nitrate must be site-specific. Nonetheless, it was concluded that the impact of DOC on the absorption spectrum, due to its chemical composition, remained relatively constant through time and is a site-specific feature (Yeshno et al., 2019). This implies that once an initial calibration equation is obtained, it can be used for that particular site repeatedly in a long duration, making real-time continuous measurement of soil nitrate concentration feasible.
Further validation of the concept was performed by reproducing similar
multivariate calibration models and their calibration coefficients for
estimating nitrate in the soil water of five additional agricultural sites
(Table 3). The polynomial calibration equations were obtained for matrices of water samples containing variable DOC concentrations, ranging from 1.6 to about 200 mg L
Observed vs. predicted nitrate concentrations.
Calibration coefficients for the multivariate regression model of the study site's obtained water samples.
Nitrate predication statistics and DOC/nitrate concentration range for the examined soil water.
In order to perform real-time analyses of nitrate in the water samples using the presented concept, the DOC concentration in the examined solution needs to be continuously substituted in the polynomial calibration equation (Eq. 2), along with the absorption at 300 nm. Thus, an independent and simultaneous analysis of DOC, along with the absorbance measurement, needs to be performed to determine the DOC concentration in the examined solution.
One way of providing an independent, simultaneous measurement of DOC in the soil solution is by fluorescence spectroscopy. Fluorescence emission values of the soil water samples at 451 nm, which had known DOC concentrations, were used to develop a calibration curve for each site (Fig. 4). The relation between DOC concentration and fluorescence intensity in the water samples from the different fields shows that each field has its own unique calibration equation due to the variation in the DOC chemical characteristics. As previously mentioned, the composition of the organic matter that constitutes the DOC and its chemical characteristics are site-specific properties. Therefore, similarly to the polynomial of the nitrate calibration equation, the calibration curves for the DOC concentrations are also required to be site-specific.
DOC concentration vs. fluorescence intensity at 451 nm.
As previously mentioned in this paper, UV absorption by aqueous nitrate can
be divided into two distinct sections: (1) a higher absorbance band around
220 nm and (2) a lower absorbance band at 300 nm. While both absorbance
bands are known to have direct correlations with nitrate concentration, the
absorbance at 220 nm absorbs light at 2 orders of magnitude higher than
the absorbance at 300 nm. Consequently, the 300 nm absorbance, which
stretches at values from 0–1 (a.u.), corresponding to the relevant nitrate
and DOC concentrations in this research, is characterized by a low signal
(nitrate absorbance) to noise (DOC absorbance) ratio. Therefore, nitrate
measurement at 300 nm is naturally more vulnerable to measurement errors,
reducing the measurement accuracy, especially at the lower concentrations
(
Observed vs. predicted nitrate concentrations, for the humus soil mixture water extract, when the calibration equation was obtained for absorption at 235 nm using a standard laboratory spectrometer.
In the framework of this research, we have developed a new concept that
enables nitrate estimation in soil water using UV absorption spectroscopy.
The new concept allows the use of LEDs as a light source when performing
spectral analyses, which are more affordable, compact, and power-efficient
than the commonly used deuterium UV light source. Moreover, the newly
developed concept accounts for DOC measurement interference, which is a
known issue when applying UV absorption spectroscopy techniques to nitrate
analyses in soil water. Due to the currently limited availability of UV-LED
light sources in the deep UV range, the concept was developed and validated
at a wavelength of
The simultaneous analytical procedure for DOC and nitrate was tested on soil water extracted from five agricultural soils and from a soil mixture of commercial agricultural compost. The results indicate a reasonable correlation between the predicted and observed values of a series of samples obtained for each soil, with variable nitrate and DOC concentrations. Nevertheless, the non-ideal conditions when applying the new concept at 300 nm led to some measurement errors, which are not in line with a state-of-the-art standard laboratory apparatus. However, the concept is quite suitable for the practical demands of an on-site and real-time nitrate monitoring sensor in agricultural and environmental applications. We hope that this work will establish the groundwork for the development of an affordable LED-based, electro-optical nitrate monitoring sensor. Such a system, which does not rely on the standard cumbersome light sources, is expected to be efficient in power consumption, size, and production costs and thus suitable for the requirements of both farmers and researchers. We believe that this nitrate and DOC monitoring concept may lead to a higher level of precision farming and the sustainable use of associated water sources while optimizing yield and decreasing food production costs.
Data sets are available upon request by contacting the correspondence author: Elad Yeshno (eladyes@post.bgu.ac.il).
The supplement related to this article is available online at:
EY (PhD student) conducted the majority of the experiments, analyzed the data, and was the main writer of this article. SB (MSc student) conducted laboratory analysis and experiments and wrote parts of the article. OD (project PI) designed the experimental setups, analyzed the data, and wrote parts of the article. SA (project PI) formulated the analytical concept, designed the experimental setup, analyzed the data, and wrote parts of the article.
The authors declare that they have no conflict of interest.
The authors wish to express their great appreciation to Michael Kugel, who assisted with each and every technical aspect of the project while providing outstanding advice for the laboratory and field experiments.
This research has been supported by the KAMIN Framework (Israel Innovation Authority, grant no. 67854) and the Israel Ministry of Agriculture and Rural Development (Eugene Kandel Knowledge Centers) as part of the program The Root of the Matter: the root zone knowledge center for leveraging modern agriculture.
This paper was edited by Markus Weiler and reviewed by three anonymous referees.