BACKGROUND Global harmonious high-quality geochemical data and accompanying maps are reference baselines for quantifying future human-induced or natural environmental changes.

OBJECTIVESTo ensure the accuracy of geochemical reference values and analysis data, advices were made for laboratory analysis.

METHODSThe analytical data of Cd, Hg and W from China, USA, Europe and Australia were compared and two analysis data collected 15 years apart from one laboratory in China were also compared.

RESULTSAll of the cadmium analysis data were consistent with a correlation coefficient of 0.96. Mercury was poorly consistent with a correlation coefficient of 0.74, and tungsten was not comparable with a correlation coefficient of 0.56. The analysis results of cadmium were highly consistent because the analysis method was the same and the detection limit was comparable. The consistency of mercury was poor, especially the low-content mercury, which was significantly different due to different analysis methods and detection limits. Tungsten was not comparable due to different laboratory analysis methods. The prerequisite for recognition of environmental changes (RC_env > RE_smpl+RD_lab) was that the change value must be larger than the value of field sampling error (RE_smpl) and laboratory analysis error (RD_lab). Therefore, sampling error and laboratory analysis error must be minimized.

CONCLUSIONSSix general guidelines are proposed. (1) The original sample is separated through a 10-mesh sieve and processed to a particle size of less than 200 mesh using a pollution-free method. (2) A total of 71 elements plus other parameters should be determined by well-established multiple analysis methods, e.g., fused glass bead-XRF for major elements and 4-acids ICP-MS and ICP-OES for minor elements in combination with other methods. (3) The method detection limits must be lower than crustal abundance of the chemical elements and reportable data percentage must be more than 90%. (4) The geochemical reference materials used for quality control should contain the reported certified CRM values for all elements. (5) Analytical relative errors for the triplicate samples should be less than 40% (RD ≤ 40%) if concentration of the element is less than 3 times detection limits, and less than 20% (RD ≤ 20%) for the elements with concentration more than 3 times detection limits as well as major elements, Fe-group elements and toxic metals. (6) The total contents of major elements SiO₂, Al₂O₃, Fe₂O₃, FeO, MnO, MgO, CaO, Na₂O, K₂O, TiO₂, P₂O₅, H₂O⁺, CO₂, SO₂ and organic matters or those of SiO₂, Al₂O₃, Fe₂O₃, FeO, MnO, MgO, CaO, Na₂O, K₂O, TiO₂, P₂O₅ and LOI should be at 99.3%-100.7%.