We integrated several independent geophysical and geologic methods to examine the effects of stratigraphic and structural heterogeneities on the growth of hydraulic fracture networks from two horizontal wells in the Barnett Shale, Fort Worth Basin, Texas. Our data set included time-lapse 3D seismic surveys, microseismic data, wireline logs, and distributed temperature sensing (DTS) data. We first created a local stratigraphic framework using wireline logs. In our area, the lower Barnett Shale consists of siliceous mudstones (the primary reservoir) intercalated with carbonate submarine fan deposits. The latter are low porosity (i.e., nonreservoir) and, if thick enough are potential baffles to the growth of hydraulic fractures. We used stochastic inversion to define the 3D distribution of fan lobes with much better resolution than could be obtained using deterministic inversion and obtained a geologically reasonable lithology prediction. The lowest of the fan lobes partially overlies the two horizontal wells, and its limits could be defined using wireline logs, the stochastic inversion, and seismic attributes (e.g., coherence, seismic facies classification). As suggested by the distribution of microseismic events, the extent of this lobe (locally up to 80 ft/24-m-thick) had a significant impact on the growth of the hydraulic fracture networks. The DTS data showed that high production correlates to dense microseismic activity in this area. Our time-lapse seismic analyses suggested that velocity changes induced by the hydraulic injections are detectable, although (largely because of logistical problems) the data were inadequately sampled to quantitatively define these changes. Alone, none of the analyses described herein provided an adequate understanding of the subsurface. However, once integrated, our multidisciplinary work provided a coherent, if still largely qualitative, understanding of the relationships between the geology and the growth of hydraulic fracture networks and some of the geophysical and engineering methods that can be used to define those links.