proportional to the 241 Am value. Therefore, the ratio and error estimators were changed to reflect this approach (Doctor and Gilbert, 1978). TRU Activity Estimation. Before the TRU activity calculation from 241A4m data could be performed, several corrections had to be made to the raw 24lam data. The first correction was for detector effective area (detector efficiency), which was required because the program which computed 24lam activity from the gamma spectrum peak area assumed all the detector crystals were 19cm? in area. However, some of the crystals were actually smaller in area, and the effective area of the crystals tended to change while the detectors were in the field. The crystal effective areas were checked routinely by the EG&G scientist and any changes were reported to DRI so that the data could be corrected appropriately. For results of these calibration procedures, see Appendix C. No estimate of the variance of this correction factor was available. Another correction was for signal attenuation due to the presence of vegetation in the detector field of view. The eorrection factor, called the Brush Correction Factor (BCF), was estimated using the data from an experiment on Pearl and corroborated by later experiments. The experiments and results are described in Tech Notes | and 1.1. (All Tech Notes ean be found in Appendix B.) The standard deviation of the BCF estimate was included in the error propagation. The proportion of the detector field of view that was covered by brush was estimated by the IMP technician in the field. In some cases, corrections were made for efficiency losses caused by operating the detector at an incorrect bias voltage. The necessary correction factors and corresponding standard deviations were estimated from remeasurements using the correct voltage, as described in Tech Notes 5.0 and 5.1. These standard deviations were included in the error propagation. Finally, there was one instance when a detector suffered a step-function loss in efficiency as a result of mechanical damage, but the loss was not noted until some time later. A detector efficiency check was performed to estimate the correction but no variance estimate was made (see Tech Note 5.2). The correction was applied to all data taken with this detector after the date of mechanical damage. After all the necessary corrections to the 241 am data had been made, these values were multiplied by the estimated TRU to 241 4m ratio to arrive at the estimated TRU activity. The estimated variance of the ratio was propagated into the estimate of the variance of the TRU activity. Details on the corrections, TRU computations, and propagation of error are given in Tech Note 20. The computed TRU activity and propagated error values were used as input to the kriging programs for initial surface characterization. The kriging routines on-island could be used to estimate the average over a square area of side d, where d is the grid spacing, using a 3 x 3 array of data points. It was also possible to use a 4x 4 array of data points to estimate the average over a square area of side 2dorside 2d centered on the center four data points. For example, with data taken at the usual 50 m grid spacing, average TRU activity could be estimated over 0.25 ha, 1.0 ha or 0.5 ha. The programs were set up to estimate the average activity over the square area even when some data were missing, such as when a sampling location coincided with a large bunker and no data could be taken. On the island edges, the programs would check which points in the standard 3 x 3 or 4 x 4 array were missing, to determine how much of the square area actually lay on the island rather than over water. Then the average activity would be estimated only on the region of the square actually on the island. The results of the area estimates were output in several forms. The computer printed a data map with the averages centered in the square they represented. A similar printout showed the 0.5 s upper bounds, where s is the standard deviation of the kriging error, on the area-average estimates. Another set of printouts consisted of maps with the sections of the island having estimates or upper bounds less than a criterion shaded one intensity and the sections above the criterion shaded a different intensity. These printouts could be done several times using different criteria or different multipliers on s, thus making comparisons of various alternatives easier for the project managers. 5.2.2 Surface Cleanup Once it was established that an area of an island would require cleanup, additional data would be collected to try to get complete coverage of an area. Prior to cleanup, the entire boundary of the area (as determined from the kriging estimates), plus a row of points on either side of the 142