Photogrammetric and structure from motion (SfM) techniques are increasingly being used to monitor active lava domes (e.g. JAMES & VARLEY, 2012, DIEFENBACH et al., 2013). This study applies SfM techniques to digital single lens reflex (DSLR) and thermal images acquired during observation overflights of Volcán de Colima prior to an eruption and associated dome collapse in July 2015. The collapse triggered a pyroclastic flow which travelled ~10.7km's, threatening several ranches and the town of Quesaría. Models of the dome were constructed from DSLR and thermal images, and georeferenced by comparison with Google Earth imagery. Models built using DSLR images were found to be substantially more sensitive to degassing and poor lighting, but were of superior quality during favourable conditions. Conversely, models produced from thermal images were less detailed but more robust in non-optimal circumstances. Thermal models were constructed from most flights, while DSLR models could only be built for about 60% of the datasets. Georeferenced models were exported as triangular meshes and aligned with a pre-dome model to improve relative georeferencing, using MeshLab's iterative closest point algorithm (CIGNONI et al., 2008). Volume differences were then calculated using an implementation of the signed tetrahedron method (ZHANG & CHEN, 2001). In our application of this method, 'regions of interest' are interactively selected and the volume between a reference surface (pre-dome model) and test surface (each dome model) calculated. Hence, the volume of the dome, top portion of the main lava flow, and two reference areas (zero volume change assumed) were estimated (Fig. 1). Estimations derived from the DSLR and thermal models generally correspond, suggesting (along with low reference area volumes) that they are reasonable, though this method assumes constant underlying topography and hence likely produces underestimates. The data show that dome growth occurred in three distinct episodes. Between January and April 2013 the dome grew at a rate of ~0.05 – 0.12 m 3 /sec, slowly filling the pre-dome crater (Figure 2a). By late April the crater was overtopped and a lava flow formed on the volcano's west flank, creating a stable configuration where inflow ≈ outflow and dome growth dropped to <0.01 m 3 /sec. The second period of dome growth occurred between July and November 2014, growing at ~0.06 m 3 /sec (Fig. 2c) and forming several new flows (which accommodated most new lava). The dome then underwent a period of substantial subsidence, deflating at ~0.03 m 3 /sec, accompanied by endogenous growth from a second (easterly) vent (Figure 2c & d). Finally, the dome inflated again (at ~0.05 m 3 /sec) from May 2015, before collapsing to the south in early July. Photographic evidence suggests effusion rate may have increased dramatically in the hours preceding collapse. The geometric and thermal evolution of the lava dome was also examined using the photogrammetric dataset. Most significantly, the models indicate that the July eruption was preceded by effusion from two separate vents (Fig. 2d) and by substantial south directed bulging of the dome. These results suggest that photogrammetric monitoring provides both important insight into volcanic processes and a useful dataset for risk forecasting.