Landscapes encode the history of the climate. For example, saprolite, the intermediate material  between rock and soil, plays a critical role in the evolution of topography, nutrient supply,  landslide hazards, and the global carbon cycle. The subsurface also bears resources used for the  production of energy and construction materials. It is thus important to assess the response of soils,  rocks and other geomaterials to a varying climate and to increasing societal demands. Here, we  present analytical and numerical models to predict the propagation of microscopic cracks and  metric fractures in porous media subjected to variable mechanical, hydraulic and chemical  conditions. We first explain a micro-mechanical model of anisotropic damage induced by the  weathering of biotite minerals. The bedrock is modeled as a matrix that contains and anisotropic  distribution of microscopic cracks and biotite inclusions endowed with a weathering stress field,  called eigenstress. The mechanical properties of the bulk rock are obtained by homogenization. With this model, we conduct a series of finite element simulations in bedrock under gently rolling  topography and a range of biotite orientations. In our simulations, crack growth and stress  redistribution are far more sensitive to biotite weathering than to the topographic or regional  stresses, which suggests that biotite weathering is capable of dominating the development of  bedrock damage. We deploy cohesive zone (CZ) elements at the boundary of the finite elements  and show that biotite weathering can affect the trajectory of metric fractures. Furthermore, we propose a strategy to simulate multi-scale fracture propagation coupled with fluid flow with the  extended finite element method (XFEM) and simulate a series of injection tests in an inclined  fracture embedded in a transversely anisotropic porous medium. The fracture propagates along the  bedding direction in the absence of in situ stress and along the direction of maximum compressive  stress when anisotropic stress boundary conditions are applied. The proposed XFEM model allows 

simulating multiscale hydraulic fracturing. To upscale the transport properties of the bedrock from  the microscopic to the metric scale, we then extend the theory of homogenization to diffusion  problems by using imperfect interfaces. We develop a self-consistent model for halite that explains  why rock made up of larger crystals exhibit lower creep deformation, while specimens with smaller  grains have a lower diffusivity. Simulations show that grain boundary healing accelerates  specimen compaction, while precipitation in the pores controls the evolution of effective  diffusivity. These results are impactful for geological storage design, but it remains challenging to  formally link the microstructure changes induced by chemical reactions and pore deformation to  macroscopic physical and mechanical properties, because the key microstructural features that govern macroscopic fluid flow differ from those that dominate elastic, plastic and brittle behaviors.  We thus conclude this talk by discussing Artificial Intelligence strategies to extract microstructure  features from various image data.