Natural fractures are a prominent and dramatic feature of many outcrops and are commonly observed in core, where they govern subsurface fluid flow and rock strength. Examples from more than 20 fractured reservoirs show a wide range of fracture sizes and patterns of spatial organization. These patterns can be understood in terms of geochemical and mechanical processes across a range of scales. Fractures in core show pervasive evidence of geochemical reactions; more than is typical of fractures in many outcrops. Accounting for geochemistry and size and size-arrangement and their interactions leads to better predictions of fluid flow. Opening-mode fracture apertures commonly follow power-law size distributions with opening displacements ranging from approximately 1 µm to 1 m. A power law forms a straight line on a log cumulative frequency versus log aperture size plot. The slope of the line is the power-law exponent, reflecting the relative number of narrow and wide fractures in the set. The pre-exponential coefficient reflects the overall fracture intensity. We will examine the variation in power-law exponent and coefficient for fracture sets in carbonate and siliciclastic rocks and analyze why such variation occurs. Fractures may open in a single event or may repeatedly open and seal. During an opening event the rate of opening competes with the fastest rates of precipitation to determine if the fracture will seal before the next strain increment. Small fractures completely seal with cement precipitated synchronously with opening, whereas large fractures may retain some porosity. The aperture size at which porosity is preserved varies, and it is controlled by the temperature of the ambient fluid, the composition and texture of the host rock and precipitating minerals, and the length of time the fracture wall is exposed to mineral precipitation, which is dependent on burial history and fracture timing. If the widest fractures are not completely sealed before the next strain increment, they may act as planes of weakness, causing strain to progressively partition into fewer fractures, which will grow wider. The extent to which this process happens should partly govern the exponent in the power-law distribution. Cements deposited whilst fractures are growing may cause fracture size distributions to vary from those found in barren fracture arrays (including many of those in outcrop). Geochemical and fracture size interactions may also affect fracture spatial arrangements. Fractures may be evenly spaced, but more commonly fractures occur in complex and, in some cases, fractal arrays of clusters. We have developed a method, based on a two-point correlation integral, to rigorously identify different types of spatial arrangement, including periodic, random, and clustered. Our method provides a measure of preferred spacing relative to that expected from a random ordering of spacings. I will show examples from outcrop data sets and from fractures interpreted in image logs in shale gas wells.