1. Preparation of proficiency testing samples
In this proficiency testing program, all the sample preparations, quality control (homogeneity and stability), and statistical analysis procedures follow the requirements of ISO/IEC 17043:2010, ISO 17034:2016, ISO 13528:2015, ISO/IEC 17025:2005 and ISO Guide 35:2017. Two soil samples and two plant tissue samples including mango leaf and rice flour were prepared for this PT program. The procedures of sample preparation are as follows:
A. Soil samples (sample ID: S01-XXX and S02-XXX)
Two soil samples were collected from the surface soil of farmlands, which are located in two areas of Taiwan. The soils were air dried for more than three days, passed through 0.5 mm sieves, and stored in plastic under drying and room temperature condition. The air-dried soils were poured into a large and clean plastic container, mixed thoroughly with a plastic shovel, and divided into four portions. Each portion of soil was transferred into a clean plastic bag and mixed thoroughly again in the bag. Afterward, the soils in four bags were combined in a bag and well mixed again. The procedures from divided soil sample into four portions to combine in a bag and mixed well were repeated for three times. After mixing, the soils were divided into four portions and poured into four plastic containers. Thus the soil is ready for PT sample packing. Approximately 100 grams of soil was randomly collected from these containers, filled into plastic vials, and sealed to complete the preparation of each PT sample.
B. Plant samples (sample ID of mango leaf and rice flour are P01-XXX and R01-XXX, respectively)
The mango leaf and rice grain samples were collected in orchards and paddy fields, respectively. Two plant samples were rinsed with tap water and then with deionized water. The mango leaf and rice grains were oven dried at 70 °C for more than three days. Before grinding, the rice grains were divided into rice husk and brown rice. The brown rice was ground to fine powder and passed through 0.5 mm sieves. Dried mango leaves were ground to fine powder directly, and then passed through 0.5 mm sieves. The oven-dried plant powder was poured into a large and clean plastic container, mixed thoroughly with a plastic shovel, and divided into four portions. Each portion of soil was transferred into clean plastic bags and mixed thoroughly again in the bag. Afterward, the plant powder in four bags were combined in a bag and well mixed again. The procedures from divided soil sample into four portions to combine in a bag and mixed well were repeated for three times. After mixing, the plant powder was divided into four portions and poured into four separated plastic containers. Thus the plant is ready for PT sample packing. Approximately 25 grams of mango leaves powder were randomly collected from these containers, filled into plastic vials, and sealed to complete the preparation of each PT sample. For rice powder, approximately 20 g samples were randomly collected from these containers, filled into aluminum foil bags, and sealed (filled the nitrogen gas and vacuumized) to complete the preparation of each PT sample.
2. Homogeneity and stability testing
To avoid the effects of in homogeneity and instability of proficiency testing samples on the performance of participants, it needs to confirm the homogeneity and stability of testing samples before sending the samples to participants. The homogeneity and stability assessment were preformed after the testing samples preparation and packaging, and the principle and method of those assessments were refereed to ISO/IEC 17043:2015 and ISO Guide 35:2017. The batch size (number of subsamples) is 50 units in this proficiency testing program, based on the requirements of ISO Guide 35:2017, the size of homogeneity assessment should be larger than three units or 10 % of the batch size, and randomly selected from the batch. Therefore, it randomly selected six subsamples to assess the homogeneity of proficiency testing samples, and the repeatability standard deviation (sr) for the homogeneity procedure should be less than the desired standard uncertainty (utrg), which could be tested by following formulae (ISO Guide 35:2017):
nal is the number of observations on each of n aliquots taken from each unit for measurement.
If the measurements correspond to the above formulae, it presents that the homogeneity of proficiency testing samples achieve the precision requirement.
For stability assessment, it also selected six subsamples after those samples were stored in room temperature condition for one month, and the measured value was regarded as new measured value (xmon). The measured value calculated by homogeneity assessment was regarded as certified value. We used the single monitoring experiment to compare the new measured value with the certified value, and then evaluating the stability of proficiency testing samples. This assessment could be testing by following formulae (ISO Guide 35:2017):
the uCRM and umon are the standard uncertainties of certified measured value (xCRM) and new measured value (xmon), and k is the appropriate coverage factor at a level of confidence of approximately 95%. Once the measurements in above testing formulae are not met, it indicates signification change of the material (testing samples) and action should be taken.
3. The proficiency testing items and recommendation methods
The soil (S01-XXX and S02-XXX) testing items of this proficiency testing scheme contains soil pH, total organic carbon (TOC), total nitrogen (N), available phosphorus (P), exchangeable potassium (K), calcium (Ca), magnesium (Mg), and soil cadmium (Cd). The analysis items of mango leaf samples (P01-XXX) are total-N, P, K, Ca, Mg, Fe, Mn, Cu, Zn, and the analysis item of rice grain samples (R01-XXX) are Cd. To compare the differences in the analytical results of soil and plant tissue among different laboratories, participants of this proficiency testing program are suggested to analyze the samples using the recommendation methods as shown in Table 2. If participants have already adopted the suggested methods in their laboratories, they just follow their existing testing procedures. If participants have not yet established the methods suggested, they can refer to the methods cited in the references column.
Table 2. The soil testing/ plant analysis items and recommendation methods of FFTC 2018 proficiency testing program
# PT: proficiency testing
4. Data statistical analysis for proficiency testing
The statistical methods of this proficiency testing scheme were referred to ISO/IEC 17043:2015 and ISO 13528:2015, and the statistical analysis for the interlaboratory comparison has three steps, including (1) determination of the assigned value, (2) calculation of performance statistics and (3) evaluation of performance, the methods of these steps were described as follows:
A. Assigned value determination: the assigned value (X) of this proficiency testing is consensus value from participants, which is calculated by the average of participant's results (x), and the consideration of the effects of outliers. The outliers among reported values are detected by boxplot, and the outliers are not used for the determination of assigned value.
B. Performance statistics calculation: To compare the difference in analytical results among the various laboratories, it needs to transform the proficiency testing results into a performance statistic. The aim is to measure the deviation from the assigned value in a manner that allows comparison with performance criteria (ISO/IEC 17043:2015). There are many methods to calculate the performance statistics, including difference (D), percent difference (D%), z scores, zeta score and En number. In this proficiency testing scheme, we used the z score to expressed as performance statistics, the equation of z score is shown below:
σˆ is the standard deviation for proficiency assessment, x and X are the participant's result and assigned value, respectively. The z score was calculated by above equation based on the results submitted by participants.
C. Evaluation of performance: In general, the z score within the range of -2 to 2 indicates a satisfactory result, while the z score outside the range of -2 to 2 indicates a questionable result. Once the analytical result belongs to outlier (determined by box plot), the z score of this analytical result will not be shown in statistical comparison results.
∣z∣ ≤ 2.0 “satisfactory” performance
∣z∣ > 2.0 “questionable” performance
RESULTS AND DISCUSSION
The results of homogeneity and stability testing
For homogeneity testing, we followed the ISO Guide 35 (2017) and used the equation (1) to check the homogeneity of proficiency testing (PT) samples including two soils and two plants, which was calculated by repeatability standard deviation (sr /√na1) and the standard uncertainty (utrg) of PT items, and the testing results was presented in Table 3. This result shows that the standard uncertainty (utrg) of all PT samples and measured items are higher than repeatability standard deviation (sr /√na1), which indicates the homogeneity testing of four PT samples corresponded to the precision requirement. Therefore, based on the homogeneity assessment, it reveals the all testing samples of this PT program are sufficiently homogeneous.
Table 3. The repeatability standard deviation (sr /√na1) and the standard uncertainty (utrg) of PT (proficiency testing) items for homogeneity testing
According to the ISO Guide 35 (2017), the assessment of stability is to estimate and confirm the stability of the PT samples after processing, storage and transportation, which also needs to consider the physical, chemical and biological properties of the samples. We used the equation (2) to check the stability of PT samples including two soils and two plants, which calculated by the difference between the certified measured value (xCRM) and new measured value (xmon), and the standard uncertainties of these two values. The results of stability testing were presented in Table 4, it shows the k√u2CRM-u2mon values of all PT samples and measured items are less than k√u2CRM-u2mon values, which indicates the stability of four PT samples, are corresponded to the requirement of single stability monitoring. In this testing, we only tested the stability of PT sample storage for one month, but it had no evaluation for the stability of PT samples during transportation, the reason is that the chemical properties of soil and plant tissues such as this PT items could not significantly change under the room temperature and low humidity condition for a long-term storage (sealed in bottles and bags) based on the previous study and experience. Therefore, it predicted that the all testing samples of this PT program are sufficiently stable.
Table 4. The ∣xCRM-xmon∣and k√u2CRM-u2mon values of PT (proficiency testing) items for stability testings
1. The assigned value of PT items of soil and plant tissue samples
Generally, in the PT program, the evaluation of measured result of participants is based on the comparison with the assigned values, and the assigned value can be determined by known values, certified reference values, reference values, consensus values from expert participants and consensus values from participants, which based on the purpose of PT program. Due to the soil and plant testing samples of this PT program were collected from natural farm land, orchard and paddy rice field, it has no known values, certified reference values and reference values. Therefore, the assigned values of this PT items were the average of participant's measured values (consensus value), but the outliers were not used for the calculation. The assigned values of different items of two soil and two plant testing samples were presented in Table 5.
Table 5. The assigned value of soil and plant PT (proficiency testing) items
2. Evaluation of performance
In this PT scheme, the z scores were used to evaluate the performance of participants, which is calculated by the equation (3), and the assigned values used in this calculation are listed in Table 5. Based on the results of homogeneity and stability assessment, it revealed that the differences between the subsamples were small, and these testing samples have high stability while stored at room temperature and low humidity condition. Therefore, it predicted that there were two reasons caused the differences in analytical results among the participating laboratories. The first reason is the analytical error within the participating laboratories, which could be confirmed by repeated testing using the identical method. The second reason is the differences in analytical method, procedure and instrument among the participating laboratories, which could be clarified and evaluated through the interlaboratory comparison such as PT program. Figures 1 to 4 show the z score distribution of different PT items and testing samples (two soils and two plants), it indicated that the differences in degree of the measured values deviated from the assigned values among the participating laboratories, and the degree of deviation was increased with the absolute z scores. In general, the uses of 2.0 as an action signal for z scores, reveals about 5 % of z scores would be expected to fall outside the range -2.0 ≤ z ≤ 2.0. Due to the probability of z falling outside -3.0 ≤ z ≤ 3.0 is very low (about 0.3 %), thus it is not suitable as an action singnal. Therefore, in this PT program once the absolute z scores are higher than 2, indicates the results are questionable (warning signal). On the contrary, the absolute z scores less than 2, indicates the results are satisfactory.
For the z score results of two soil testing samples (Fig. 1, 2), it shows the absolute z scores of the most of participants were less than 2.0 for all PT items, only 1 or 2 participants out of the range, the proportion of those participants fall inside this range (-2.0 ≤ z ≤ 2.0) is corresponding to the probability of two standard deviation (95 %). This result indicates the most of participating laboratories in this PT program have the good capability of soil testing, and the most of measured results is recognized as satisfactory performance, only a few questionable measured values and outliers were found, which mainly resulted from the participants using similar methods. However, for a few PT items of two soil samples, it was found that there was high deviation degree of exchangeable Ca measured values among the participating laboratories, and the coefficient of variation (CV) was also higher than 30%, which might be caused by the different extraction methods (ammonium acetate method and Mehlich's method) and analytical instruments (EDTA titration, atomic absorption spectrometry (AAS), flame photometer (FA) and inductively coupled plasma-optical emission spectrometry (ICP-OES)). In addition, it was also found that the high deviation degree of Cd concentration measured values of S01 among the participating laboratories, and the CV of S01 was about 60%, which mainly resulted from the Cd concentration (assigned value: 0.339 mg kg-1) of S01 sample was too low. On the contrary, the assigned value of Cd in S02 is 8.75 mg kg-1, and the deviation degree and CV of measured values of S02 among the participating laboratories were low. It revealed that the performance of participants also affected by the concentrations (assigned value) of PT sample.
Similarly, the z score results of two plant analysis samples (Fig. 3, 4), also shows the absolute z scores of the most of participants were less than 2.0 for all PT items, only 1 or 2 participants out of the range. According to the distribution of z scores, it also indicates the most of participating laboratories in this PT program have the good capability of plant analysis, and the most of measured results is recognized as satisfactory performance. However, due to the differences in digestion methods (conc. HNO3/HClO4/HCl, conc. H2SO4/HNO3/HClO4, conc. HNO3/HClO4, 1N HCl, conc. H2SO4) were used among the participating laboratories, it found that the measured values of Cu, Zn and Fe had high deviation degree and CV (＞30%), but the deviation degree and CV of P, K, Ca and Mn are were low. This result reveals that different digestion reagents have different digestion capabilities for the same elements. For most of the laboratories, the same digestion reagent usually used to determine the all elements (macronutrients and micronutrients) in plant tissues, it is easy to find the underestimating problem for some elements if the strength of digestion reagent is not enough. For the results of rice samples, it observed that the deviation degree and CV (＞50%) of measured values of participating laboratories were high, which might be caused by the low assigned value (0.157 mg kg-1) and the small number of participating laboratories, and leading to it is not easy to distinguish the performance of participants. For this case, it can be improved by narrowing the range of z score for satisfactory performance, such as reducing the absolute z score from 2 to 1.5.
Fig. 1. The z score distribution of different proficiency testing items (soil samples, S01) calculated by participant's measured values.
Fig. 2. The z score distribution of different proficiency testing items (soil samples, S02) calculated by participant's measured values.
Fig. 3. The z score distribution of different proficiency testing items (mango leaf samples, P01) calculated by participant's measured values.
Fig. 4. The z score distribution of Cd concentration in rice samples (R01) calculated by participant's measured values.
The results of homogeneity and stability testing indicated that the differences between the subsamples were small, and these testing samples have high stability while stored at room temperature and low humidity condition. The results of performance evaluation indicated that the z scores of most of participants fell inside the range of -2 to 2 (satisfactory results) for most of PT items of two soil samples and two plant samples, only 1 or 2 participants fell outside this range. It represented that the capability of soil and plant testing for most PT participants were satisfactory. However, it also found that the high deviation degree and coefficient of variation (CV) of measured values for a few PT items such as the exchangeable Ca for soil sample and the Cu, Zn and Fe for plant sample (mango leaf), which mainly resulted from the use of different analytical methods and instruments. In addition, it also found that the performance of participants also affected by the low concentrations (assigned value) of PT samples and the less PT participants (e.g. Cd for S01 and R01 samples). In summary, it suggests the participation in the PT program regularly will help to maintain the soil/plant testing quality of laboratory, and further to elevate the accuracy of the fertilizer recommendation and plant nutritional diagnosis.
ISO/IEC 17025. 2005. General requirements for the competence of testing and calibration laboratories.
ISO/IEC 17043. 2010. Conformity assessment－General requirements for proficiency testing.
ISO 13528. 2015. Statistical methods for use in proficiency testing by interlaboratory comparison.
ISO 17034. 2016. General requirements for the competence of reference material products.
ISO Guide 35. 2017. Reference materials－Guidance for characterization and assessment of homogeneity and stability.