Nitrogen-vacancy (NV) centers in diamond excel as room-temperature quantum sensors by virtue of their long-lived spin coherence and experimental addressability at the single-spin level. When isolated deep within bulk diamond, NVs' spin coherence times and relaxation times are limited to several milliseconds by internal nuclear and electronic spin baths and vibrations in the crystal structure. However, when NVs are placed just nanometers from the diamond surface, which is necessary for nanoscale imaging of external fields, NV spin properties are impacted by a host of new decoherence sources that must be understood and mitigated to optimize the utility of the NV as a magnetometer.

This dissertation addresses the questions: 1) What is the length scale over which near-surface NV spins experience decoherence due to the diamond surface? 2) What are the physical noise sources, and their frequency spectra, that cause surface-induced decoherence in NV centers? In addressing these questions, we also develop a NV on a scanning probe tip platform and use it to perform nanoscale imaging based on the NV spin-relaxation rate in the presence of magnetic and electric field fluctuations.

First, we develop a method of nitrogen delta-doping during single-crystal diamond growth to create near-surface NV centers localized at multiple few-nanometer layers. Through a technique of scanning probe magnetic resonance imaging, we measure the depths of these shallow NVs with nanoscale precision. We correlate these depths to spin coherence times measured with dynamical decoupling and model this depth dependence with a combined model of surface-related and bulk magnetic noise.

We find that significant discrepancies between the maximum measured coherence time and its maximum theoretical limit -- twice the spin relaxation time -- necessitate further study of the relaxation rates of near-surface NV centers. We develop a method to measure relaxation rates between all three NV spin-triplet ground state levels and find that a double-quantum (DQ) spin relaxation channel is a major, and under some conditions dominant, contributor to the total spin decoherence rate. We demonstrate a surface-noise spectroscopy technique combining dephasing and DQ relaxation data to identify which parts of the derived noise spectral density are due to magnetic fields or electric fields. The susceptibility of the NV to electric field noise through DQ relaxation, as with other decoherence channels, is simultaneously a problem and a potential resource for sensing and imaging.

Finally we employ the near-surface NV as a scanning probe to perform nanoscale decoherence-based imaging of electromagnetic noise from various target samples. Using NV spin relaxation as a signal, we demonstrate two-dimensional imaging of magnetic noise from a few-thousand Gd$^{3+}$ spin labels with 20-nm spatial resolution, limited by setup drift in ambient conditions over long time scales. We then make a number of magnetometer enhancements towards imaging at the level of single-spin sensitivity, including engineering diamond nanopillars, forming shallower NV centers, and improving scanning probe microscope stability. We apply the three-level relaxometry techniques to study both magnetic and electric field noise at the NV as a function of nanometer-scale distance from metallic surfaces. Surface-induced decoherence is a major challenge in a variety of qubit systems and hybrid qubit interfaces, and the shallow NV center in diamond is positioned to be a valuable scientific tool for studying these noise sources at nanometer, and even sub-nanometer, length scales with high precision.