he rapid evolution of fifth-generation (5G) wireless communication systems has placed unprecedented demands on radiofrequency (RF) front-end components, with requirement of compact, high-frequency and low-loss resonators capable of reliable operation under severe thermal constraints. Micro-electro-mechanical system (MEMS) piezoelectric resonators have emerged as key building blocks for both RF signal processing and advanced sensing technologies. III-nitride piezoelectric materials, including aluminum nitride (AlN), gallium nitride (GaN) and emerging aluminum scandium nitride (AlScN) alloys, have attracted significant attention due to their high phase velocity, strong electromechanical coupling, high quality factor and compatibility with semiconductor manufacturing. These materials further enable multifunctional resonant architectures in which frequency selectivity, nonlinear dynamics and environmental sensitivity can coexist in a single device. However, thermal properties such as small thermal conductivity and large temperature coefficient of frequency (TCF) introduce critical challenges that limit quality (Q) factors and power handling capabilities in 5G systems. This dissertation investigates the design, fabrication, and characterization of III-nitride piezoelectric resonant devices for emerging 5G filter technologies and high-performance resonant sensing systems, with an emphasis on the thermal behavior of AlScN and the exploitation for high-sensitivity resonant sensing applications. The thesis begins with a demonstration of a AlScN/GaN/sapphire SAW resonator with ferroelectric switching capability suitable for super high frequency (SHF) operation. We present the characterization of the thermal conductivity of AlScN and use heat-transfer FEA model and ab initio calculations as approaches to study self-heating and quality factor degradation phenomena of AlScN based resonator. We explore the potential of using AlScN as a piezo material for resonant temperature sensors by harnessing device/material nonlinearities that give rise to parametric frequency conversions and phononic frequency combs. Our research opens a new chapter for engineering high-sensitivity tunable physical sensors, extending toward deeper understanding of noise limits, stability and closed-loop operation.