Abstract: Remote sensing and machine learning (ML) are transforming how we model environmental systems and enabling innovative solutions. This dissertation develops ML- and remote sensing-based frameworks for two environmental problems. In Part I, we build advanced ML models that (i) improve continental U.S. streamflow forecasts, (ii) scale to the global scope while reducing computational cost about 50% while preserving predictive skill for extreme events, and (iii) better represent inter-basin heterogeneity and improve flood-peak timing estimates. In Part II, we combine multisource remote sensing with ensemble ML to map fine-scale malaria risk in East Africa, supporting targeted surveillance and vector-control strategies.