In my experience teaching geology, many people are fundamentally uninterested in rocks; they're mundane, pretty static, boring, etc. Many people are also not particularly spatially-inclined so maps aren't interesting. I think the spatial and temporal scale issues you bring up are also factors.

People do seem to be interested in landscape development, minerals (pretty crystals are cooler than pieces of shale), earthquakes, etc. Paleontology is typically housed in geoscience departments too.

From a research perspective, geoscience is extremely fun. Even to those infatuated with the Earth such as myself, the individual topics of study can get a little dry. However, geologic problems can be approached from a wide variety of methods, and to solve those problems that are still standing, a research group will come at it from many angles. This includes, but is not limited to, physics-based computational modeling, hiking, isotope geochemistry, fossil collection, statistics, satellite imagery analysis, acoustic physics, technical drawing, computer simulation, spherical geometry, and smashing boulders with sledge hammers.

For example, one chapter[0] of my PhD thesis (studying the development of fault-bounded mountain ranges in Tibet) I spent several months camping in beautiful Tibetan mountains doing geologic mapping and sampling (which is hiking, making qualitative and quantitative field observations and measurements, as well as digging holes and picking up rocks), then I came home and did a lot of radioactive isotope geochemistry of the minerals apatite and zircon within the rock samples: These minerals contain uranium and thorium; as the U and Th decay, they produce 4He (an alpha particle). At high temperatures, those particles diffuse out of the apatite and zircon crystals, but at low temperatures the 4He is retained in the crystals. So you measure the ratio of 4He to U and Th, and get an idea of when the crystals cooled below the diffusion 'closure temperature' (which is dependent on cooling rate). By taking these measurements over almost a vertical kilometer along the mountain range, I could estimate how fast the mountain range was rising, because it cools as it rises out of the surrounding crust. But it's a bit complicated to do the calculations right because there are lots of small issues (how fast the rock in the mountain range cools is dependent on the ambient temperature of the crust, the thermal properties of the rock, how much radiogenic heat production there is from K, Th and U decay, the temperature of the surface of the earth, the irregularities of the topography, ...) so to do it right you have to make a finite element model of the crust and simulate the thermal evolution as the crust deforms due to faulting and mountain range development. This is only a forward model, so to figure out the whole history of deformation including faulting and mountain range uplift rates through time, and deal with uncertainty in the thermal parameters of the crust, I had to run thousands of forward models. Lacking a cluster, I had to teach myself Python and how to use AWS EC2 instances to do embarrassingly parallel finite element modeling (I used the PiCloud library which was incredible but died in an aquihire event).

[0]: http://onlinelibrary.wiley.com/doi/10.1002/tect.20053/abstra...