RESTRICTIONS IN THE USE OF CORRELATION-COEFFICIENTS IN COMPARING METHODS FOR THE DETERMINATION OF THE MICRONUTRIENTS IN SOILS

Fifty nine soil samples from the State of Sao Paulo were collected fro m the plow layer, with about half of these samples presenting low or m edium concentrations of micronutrients. Micronutrients were extracted by one of the following solutions: DTPA at pH 7.3, DTPA at pH 5.3, Meh lich-1 BR, Mehlich-1 USA, Mehlich-3, and 0.01 mol L-1 calcium chloride (CaCl2), and then determined by inductively coupled plasma emission s pectrometry (ICP AES). Simple linear correlations were calculated for the results of all methods against the DTPA method at pH 7.3, Two sets of calculations were always considered: one including all 59 soil sam ples and another set including only those samples containing low and m edium amounts for each element. Higher correlation coefficients were o bserved when samples with high contents were included. For example, th e range of correlations (R-2) obtained for zinc (Zn) varied from 0.41 to 0.98, if all samples were considered in the calculations, but the r ange decreased to 0.41 - 0.55 if samples with high contents of Zn were not considered. The correlation coefficients for CaCl2-extractable Zn were not significant. The correlation coefficient(R-2) range for copp er (Cu) from 0.83 to 0.94 for all samples, decreased to 0.66 - 0.72 wi th the exclusion of the samples with higher Cu contents. For manganese (Mn), the range of correlation coefficients (R-2) varied from 0.74 to 0.92 for all samples but was reduced to 0.30 - 0.50 when the samples with high contents of Mn were excluded. In the case of iron (Fe), the DTPA method at pH 5.3 presented results dose to DTPA at pH 7.3 (R-2 = 0.92), while the CaCl2 procedure had the poorest correlation (R-2 = 0, 42), In conclusion, the substitution of an extraction procedure based only in the correlation coefficients between methods may be misleading if the interest is diagnoses in the deficiency range.