This appendix describes a series of correlation criteria that EPA considers to be reasonable for the purpose of demonstrating equivalency between two test systems designed to measure the same emissions during FTP locomotive testing. These criteria are presented here only as guidelines. When requested to make a finding of equivalency, EPA could base its decision on criteria other than those listed here, where EPA has reason to believe that these criteria are not appropriate.

(a) General approach.

(1) Multiple tests should be conducted in pairs on the same locomotive or engine using each of the measurement systems.

(2) Variations for other parameters, such as test fuel, should be minimized to the maximum extent possible.

(3) Locomotive and/or locomotive engine tests conducted in accordance with the provisions of Subpart B of this part are preferred. Where appropriate, engine tests conducted in accordance with 40 CFR part 89 may also be used.

(4) Equivalency of the systems should be determined by comparing individual modal data, individual cycle-weighted data, and the average cycle-weighted results from each system.

(b) Correlation criteria for particulate measurements.

(1) The correlation coefficient (R2) for individual modal data should be 0.90, or higher.

(2) The maximum deviation between any pair of cycle-weighted data should be 15 percent, or less.

(3) The ratio of average cycle-weighted results using the alternate system to the average cycle-weighted results using the specified Part 92 system (i.e., avgalt/avgspc) should be between 0.97 and 1.05.

(c) Correlation criteria for other measurements. Correlation parameters for gaseous pollutants should be better than those specified in paragraph (b) of this appendix for particulate measurements.

(d) Minimum number of tests. The recommended minimum number of tests with each system necessary to determine equivalency is:

(1) Four locomotive or locomotive engine tests, conducted in accordance with the provisions of subpart B of this part; or

(2) Seven 8-mode nonroad engine tests, conducted in accordance with the provisions of 40 CFR part 89.

(e) Statistical outliers. Statistical outliers may be excluded consistent with good engineering judgement. Outliers should be replaced by rerunning each excluded test point. Where more than one outlier is excluded, is recommended to perform one additional pair of tests (in addition to the minimum number specified in paragraph (d) of this appendix) for each two outliers excluded.

[63 FR 18998, Apr. 16, 1998, as amended at 70 FR 40457, July 13, 2005]