In earlier use of these algorithms within the Swedish network (Slunga et al. 1984) a combined interpretation of fault plane solutions and location was tested. This interpretation algorithm has now been modified to fit the data format of the Icelandic network. The multievent interpretor makes a systematic search for groups of microearthquakes with hypocenters on a plane and having acceptable fault plane solutions close enough to this plane. The output of the algorithm is therefore a number of planes, each defined by a group of earthquakes. A quality measure is computed for all the groups. Its value is determined by the number of earthquakes, the degree of fit between the fault plane solutions and event distribution, and on the events' closeness to the plane. The basic idea behind this interpretation is that several densely spaced earthquakes often lie on a common fault plane and their locations can then be used to constrain the strike and dip of that plane. In addition, the events should also have similar fault plane solutions, which should be close to the best fitting plane through the group. We expect the chosen fault plane to represent the true fault, as we do not expect the spatial distribution to favour the auxiliary plane.
We selected data from two subareas, one near the Nesjavellir geothermal field north of Mt. Hengill (lower left corner at -21.3E, 64.085N and upper right corner at -21.2E and 64.15N), the other near Ölkelduháls (corners at -21.25E, 64.045N and -21.15E, 64.085N). Within the Nesjavellir area, accurate relative locations were obtained for 427 earthquakes occuring from January 1, 1998 to January 31, 1999. After relocation, four clusters of earthquakes become apparent in the dataset, located beneath Kýrdalsbrúnir, Nesjavellir, Stangarháls and Hvanngil (Figure 3b). The strike of the best fitting plane through each of these groups is 25E. All, except the plane through the Stangarháls sequence, are nearly vertical and faulting is mostly strike-slip. The group beneath Stangarháls delineates a plane dipping 65 to the west and movement on the plane is normal faulting. In general, the fault plane solutions of individual earthquakes agree with the fault aspects estimated from the distribution of events with median deviation from the best fitting plane through each group around 10.
For the Ölkelduháls area, accurate relocations were obtained for 505 earthquakes which occurred from January 1 to May 20, 1998. After relocation, planes were fit through each of the seven subgroups of the data (Figure 4). Their strike is 10-20 and the dip is close to vertical. Average distance of earthquakes used to constrain each plane, from the best plane, is 5-20 m for all seven groups. Fault plane solutions show a combination of right-lateral strike-slip and normal faulting on northerly striking faults generally agreeing with the event distribution. The median of smallest deviation of fault plane solutions from the best plane through each group is usually less than 6.
An automatic search for planes that fit both the locations of earthquakes and their fault plane solutions suggests the presence of possible E-W faults on top of the three N-S elongated groups near Katlatjörn efri (Ktl, Figure 5), as well as on top of the N-S groups under Hrómundartindur (Hr) and south of Djáknapollur (Dkn). Events on these faults are distributed in time and the lineaments only become apparent when long time periods (several months) are examined.
The methods used in this study to map faults with microearthquakes are still under development.