Abstract

A magnetic resonance sounding (MRS) campaign was carried out at Nam Loỏng, northwest of Lai Châu Town to investigate the fractured-karst aquifer of the Đồng Giao Formation (T₂a đg). The borehole LK.1 was drilled to evaluate the results of sounding d5 which uses 37,5 m to turn square 8 shaped loop. The low quality data of d5 is caused by the noise, generated firstly by power line and lightning, particularly at Larmor frequency - 1950,1 Hz, secondly by systematic error of the software ProDiviner in Larmor frequency calculation and at last by geomagnetic diurnal variation. As a result, the signal of noise ratio is low (S/N = 2.3) and the frequency shifts of 11 in 16 moment pulses are large in the 1.2 to 2.5 Hz range. Although the solution of MRS inverse problem is not unique, under high S/N and shallow aquifer top conditions, the estimated aquifer top is quite stable according to the numerical modelling [11] and well agrees with the static water level measured in boreholes [12, 32]. In this article, the effects of inversion parameters used in SAMOVAR Software on MRS inversion results (especially, the elimination of bad pulse moments, regularization parameter R and number N_L of 1D model layers) using bad data of d5 sounding to detect deep aquifer are presented. Until now, there is no general rule about marking the position of the groundwater table in smooth inversion [16], thus in this case the position was determined by inversion result of water content with the principle that the water content at this position must be larger or at least equal to the error of the maximal water content of the aquifer. The inversion results (deep, water content, etc...) are significantly affected by R and N_L but they take only one to four values, so 3 values of R (R = Ra in automatic procedure, R < Ra, R > Ra) and 2 values of N_L (N_L = 16, optimal value and N_L in automatic procedure) can be chosen to implement the inversion. Using inversion technique with all or the rest after removing one or even 2 to 4 bad pulse moments together from 16 pulse moments, 68 inverse data have been generated. The statistic processing results of this complex data reveal that there are two aquifer zones, the shallow one with depth L₀ = (9.1±0.9) m and the water content w₀ = (0.8±0.8) % and the deep one with L₁ = (24.0±3) m and the maximal water content is (20±7)%. In conclusion, even though signal noise ratio was low about 2,3 and depth of L₁ was quite large in comparison with the size of the loop (a = 37.5 m, L₁ /a = 0.64), depths of aquifers received by MRS inversion fairly agree with borehole data.

Furthermore, during the data processing, the systematic error in Larmor frequency calculated by ProDiviner software has been discovered. The error is caused by excessive rounding of proton gyromagnetic ratio over 2 pi value to 0.0426 Hz/nT.