Performance of commercial soil laboratories in a proficiency test program in Brazil

Abstract

A soil proficiency test (PT) was administered to 50 participant laboratories in which two sets of samples, consisting of 20 yearly PT samples and 5 ‘blind’ samples in clients’ names were analyzed for pH, organic matter, total acidity, extractable calcium, magnesium, potassium and phosphorus by the laboratories. Our objective was to determine whether laboratories take extra care to analyze clients’ samples as they do with regular PT samples. The analytical data were evaluated essentially by the procedure described in the international harmonized protocol for proficiency testing of analytical chemistry laboratories. Performance of participant laboratories was assessed by z-scores and summary z-scores statistics involving sum of squared z-scores interpreted as chi-square $\left( {\chi_{n}^{2} } \right)$ distribution for zero-centered z-scores with unit variance. From 8 750 determinations, outliers and stragglers accounted for less than 2% of the entire data. Over 93% of the data were satisfactory, whereas between 2 and 4% were either unsatisfactory or questionable in both the PT and ‘blind’ tests. On the basis of sum of squared z-scores interpreted from $\chi_{n}^{2}$ distribution table, between 30 and 40% of the laboratories had more than 90% probability of having their measurement data within the robust mean and standard deviation for each soil parameter, while another 30–42% of the laboratories had less than 50% probability of having measurement data within the robust mean and standard deviation. Overall, 21 laboratories (42%) were ranked in Class A either in the PT or ‘blind’ tests out of which 12 of them (57%) retained this ranking in both tests. Fourteen laboratories (28%) were ranked in Class C in either the PT or ‘blind’ tests with only 5 of them (36%) consistently ranking in this class in both tests.