Illinois
Fertilizer Conference Proceedings
January 24-26, 2000
| Home | 2000 Index | Search |
| Home | 2000 Index | Search |
J.W. Hummel, R.R. Price, R.G. Hoeft, T.R. Peck, and S.J. Birrell1
The objective of this work was to develop and test a real-time soil nutrient analysis system, based on ion-selective field-effect transistors (ISFETs). The development of a real-time soil nutrient sensor could allow the automated collection of soil nutrient data on a fine resolution to accurately characterize within-field variability for site-specific fertilizer application.
The application of the sensor in Illinois agriculture will allow producers to identify fields and soils where variability can be profitably addressed by site-specific management technology. In addition, the sensor will provide accurate maps of soil nutrients, so that georeferenced nutrient applications can be precisely made. Improved correlation between measured soil nutrient levels, nutrient removal due to harvested crops, and spatially applied nutrient additions will foster increased confidence in soil nutrient analysis. Agricultural producers will benefit from more efficient utilization of purchased inputs, and Illinois consumers will benefit from reduced adverse impact of agricultural practices on the environment.
Previously completed research (Birrell 1995) resulted in a multi-ISFET nitrate sensor that was integrated into a flow injection analysis (FIA) system to measure soil nitrates. A prototype automated soil extraction system was also developed and tested in conjunction with the ISFET/FIA system. The prototype automated extraction system did not consistently provide soil extracts that could be analyzed by the ISFET/FIA system, and the predicted concentrations were much lower than the actual soil nitrate levels, although the predicted levels did show the same trends between different soils. The major sources of error resulted from inconsistent sample flowrate to the FIA system due to blockages in the filtration process and variation in the volume of soil metered for individual cycles. Although the automated extraction procedure was not very successful, the results were encouraging and warranted further development.
We previously reported (Birrell et al., 1999) on the development of a system to test nutrient extraction performance (Figure 1). The system was designed to evaluate the effect of soil type and texture, soil moisture content, soil sample volume, sample preparation, and rate of extraction solution injection into the sample on extraction efficiency and extraction time. Extracting solution is forced through small circular soil cores of different sizes and then drawn through a filter into a flow cell containing an ISFET sensor.
For this experiment, a 4x2x2x2x2x2x2 factorial was used (four soil types [Table 1] and two levels of each of the remaining variables) to investigate the effects of all of the variables on real-time soil nitrate extraction and analysis. The soils were selected to provide a broad range of both textural properties and organic matter content. All soils had been screened using a 2-mm sieve, air-dried, and stored in sealed containers at room temperature. The soils were placed in columns and immersed in de-ionized water to allow the nitrates to either denitrify or leach out in the water. The soils were dried at 104°C for 48 hours, crushed, sieved through a 2-mm sieve, and stored in plastic bags at 5°C. Laboratory analysis indicated that the nitrate-N concentration had been reduced to a range of 6 ppm – 12 ppm. Each of four lots of each soil was adjusted to a different combination of the two levels of nitrate-N and soil moisture, resulting in a total of 16 test soils.
A multi-ISFET chip was prepared for the experiment by applying membranes composed of a ligand, tetradodecylammonium nitrate (TDDA), and a plasticizer, 2-nitrophenyl octyl ether (NPOE), and high molecular weight polyvinyl chloride (PVC) to the gate areas of all four ISFETs using the same procedures as Birrell (1995).
The ion-selective field effect/flow injection analysis (ISFET/FIA) system was calibrated at the beginning of the experiment, and again after each set of 16 soil samples, using injections of standard calibration solutions (5.0[10-3] M, 2.5[10-3] M, 1.25[10-3] M, 6.25[10-4] M, and 3.125[10-4] M NaN03 in 0.01 M CuSO4) and base solution (0.01 M CuSO4). The calibration data were saved in a separate spreadsheet, which accelerated the calibration process by allowing rapid comparisons of the ISFET responses to the range of standard solutions.
During testing, soil was packed into the soil core chamber and loaded onto the test stand base. A filter disc and holder, a sampling needle, and a sampling needle holder were inserted and clamped in place. Extracting solution was forced through the soil core, and the needle collected a sample of the extraction solution exiting the soil core, which was injected into the ISFET/FIA system for analysis. The ISFET output voltage was sampled at a rate of 100 Hz, for a total of 5000 data points for each soil core. A sample of soil was drawn from each of the 16 test soils during each test day. These samples were split and used for laboratory nitrate-N analysis and gravimetric moisture analysis.
The results of typical nitrate extraction tests (Figure 2 and Figure 3) illustrate the response of the ISFET/FIA system, as well as extraction pump position and pressure. A noise-reduced ISFET response curve, produced by conducting a 20-point running median filter of the data, is superimposed on the raw data. A comparison of the results shows that the shape of the ISFET response curve was affected by the soil type being tested. A maximum response occurred for the lighter texture soil, Ade loamy sand (Figure 2), and then decreased quickly, even as extracting solution was continuing to flow through the soil core. In contrast, the ISFET response for the heavier-textured Proctor clay loam soil (Figure 3) did not exhibit a maximum response as quickly, nor did it decrease in response level as quickly as did that of the lighter texture soil.
The differences in ISFET response among soil types led to the investigation of a suite of response data descriptors that might be useful for nitrate level prediction. Obvious data descriptors to investigate included peak, slope, and cumulative area-to-peak (Figure 4). In addition, three values of slope were calculated as the difference between the baseline value prior to injection and the response value at 0.5 s, 1.0 s, and 2.0 s, respectively, after injection was initiated, divided by the elapsed time. By fitting a regression curve to the initial part of the ISFET response curve, we were also able to add a slope value calculated at 0.25 s after injection initiation.
Statistical analysis was used to identify the design variables that significantly affected the values of the data descriptors. All of the data descriptors were analyzed, with the goal of identifying all variables and all second and third order interactions between and among variables, that significantly affected each data descriptor.
A typical statistical analysis for one of the data descriptors, the regression peak response (Table 2), shows that nitrate-N level, soil type, core length, core diameter, and extraction pump flowrate were variables whose effects on regression peak response were highly significant. In addition, second-order interactions between soil type and nitrate level, and between nitrate level and flowrate, were also highly significant. Nitrate-N and soil type were the variables with the greatest effects on the ISFET data descriptors, suggesting that a priori knowledge of soil type might be necessary for the ISFET technology to make accurate real-time measurements of soil nitrate-N.
The time required to obtain a nitrate-N value is important in the development of a real-time sensor. The results of these calculations (Table 3 and Table 4) indicate that, on average, from 2 s to 5 s are required to produce a value, but the elapsed time could be as long as 27 s. Obviously, one of the slope data descriptors, with elapsed times of 0.25 s to 2.0 s, would be a better choice if response time is critical, even if the nitrate-N response is less than optimum.
The nitrate extract collected from an intact soil core was indicative of the nitrate concentration in the soil core. The method of extraction and sample injection into the ISFET/FIA system appeared to be satisfactory, although conclusions are limited somewhat by large variations in laboratory nitrate-N analyses.
Nitrate-N level, soil type, core length, core diameter, and extraction pump flowrate were variables whose effects were highly significant. These results suggest that a priori knowledge of soil type might be necessary for the ISFET technology to make accurate real-time measurements of soil nitrate-N.
The peak, 90% peak, and cumulative area-to-peak data values could estimate nitrate-N concentration, on average, in 2 s to 5 s, but the elapsed time could be as long as 27 s. Slope of the injection curve, with elapsed times of 0.25 s to 2.0 s, may be a better technique for a real-time sensor.
Table 1. Textural classification, particle size, and organic matter content of soils (Womer, 1989).
Table 3. Time to attain a data descriptor value for ISFET 2 data.
Table 4. Time to attain a data descriptor value for ISFET 3 data.
Figure 1. Schematic of soil extraction device
Figure 2. Nitrate extraction curve for an Ade loamy sand - ISFET #3
Figure 3. Nitrate extraction curve for a Proctor clay loam soil - ISFET #2
Figure 4. Typical ISFET response showing peak, slope, and cumulative area-to-peak data descriptor
1 J.W. Hummel is Agricultural Engineer, USDA Agri. Res. Ser. and Professor, Dept. of Agri. Engr., Univ. of Illinois; R.R. Price is Graduate Research Assistant, Dept. of Agri, Engr., Univ. of Illinois; R.G. Hoeft is Professor, Dept, of Crop Sciences, Univ. of Illinois; T.R. Peck is Professor, Dept. of Nat. Res. and Environ. Sciences, Univ. of Illinois, Urbana IL; and S.J. Buren is Assistant Professor, Dept. of Agri. and Bio. Engr., Iowa State Univ. , Ames IA.
Birrell, S.J. 1995. Multi-ISFET sensor system for soil analysis. Unpublished Ph.D. Dissertation. Library, University of lllinois at Urbana-Champaign, Urbana IL.
Birrell, S.J., J.W. Hummel, R.G. Hoeft, and T.R. Peck. 1999. Development and field testing of real-time soil nutrient sensors for precision fertilizer application. In R.G. Hoeft (ed.) Proc. Ill. Fert. Conf., pp 65?70, Univ. of BI. Coop. Ext. Set., Urbana IL.
Price, R.R. 2000. A real-time core extraction system for soil nitrates. Unpublished Ph.D. Dissertation. Library, University of Illinois at Urbana-Champaign, Urbana IL.
Womer, C.R. 1989. Design and construction of a portable spectrophotometer for real-time analysis of soil reflectance properties. Unpublished M.S. Thesis, Library, University of Illinois at Urbana-Champaign, Urbana IL.
