We present a data-driven investigation on the inaccuracies that are inherent with the paradigm of monitoring treatment-induced microseismicity from a limited set of observation wells. For this, we select a set of microseismic events detected on three vertical arrays with sufficient quality to be locatable on each individual array. This fact allows us to examine how the azimuthal coverage of the treatment zone affects the locations of the microseismicity by systematically excluding arrays from the location analysis. From there, we show how interpretations may be biased from “ground truth”, here taken to be the three-array locations. We observe a general increase in the degree of scatter of the data in going from three arrays to two and finally one. The two array solutions do show some small geometrical biases in terms of azimuth and asymmetry of the cluster of events. The distortion of fracture geometry is exacerbated for the one-array locations, where the location algorithms are more reliant on the particle motions of the events to constrain the azimuth from the hypocenters to the arrays. We relate the decimated arrays location artifacts to the geometry of the geophone arrays, inferring how considerations like the viewing angle of the sensors and the distance to the events impact our observations.
Industry frac monitoring has standardized a single observation well configuration due to many factors, mainly availability of wells in proximity to the stimulation, costs associated with closing of production wells, etc. This pattern is replicated in many longer term or permanent installations. One question that arises is how robust are the distributions observed? Are there artifacts introduced into the interpretation of fracture trends such as length, width, height, azimuth induced by monitoring the microseismicity from a single-well? How important is the orientation and distance of the observation well with respect to the treatment zone and the expected fracture azimuth? Modeling expected errors and event detectability though algorithms like Monte Carlo simulations can be one line of inquiry to answer these questions, by perturbing event locations and seeing theoretically what effect that will have on the recovered waveforms. Another approach is to use data analysis to attempt to quantify how single-well observation can bias event distributions. The latter approach is how we choose to investigate this problem. In this paper, we discuss data-driven approach to describing single-well recording biases. We use as ground truth a dataset of 52 events with moment magnitudes ranging from -3.0 to -2.3 recorded on three monitoring wells. We progressively decimate the array coverage, first by considering all possible two-array locations and then moving to all one-array scenarios. The resulting event distributions from the dual and single arrays are then evaluated with respect to the full array solution. The overall trends of the fracture growth, the scatter of the distributions, and the overall spatial deviations from ground truth will reveal how incomplete array coverage influences the observed microseismic event clouds, and interpretations of fracture effectiveness.