The recent growth of the Internet of Things (IoT) brings a wealth of devices that collect environmental data, enabling data-informed decision-making. Low Power Wide Area Networks (LPWANs) are emerging IoT protocols. This work designs LPWAN networks at three scales to drive community impact and guide future adoption. The first scale consists of a 1-gateway, 100-device IoT network in a residential building. A combined quantitative and qualitative approach is taken to understand changes in the indoor thermal environment and energy demand after two distinct building energy retrofits. The second scale comprises LPWAN networks for individual cities. Novel LPWAN coverage predictive models were created using transmission datasets from Geneva, Ithaca, and Brooklyn, NY. Model evaluation demonstrates the effectiveness of predicting Packet Reception Rate (PRR) as an alternative to existing Received Signal Strength Indicator (RSSI)-based techniques. The third scale consists of LPWAN networks at state and county levels. I explore the generalizability of my PRR models, comparing predictive performances of Logistic Regression, XGBoost, and Convolutional Neural Network (CNN) models on out-of-sample sets from different regions. The suite of tools I developed will equip municipalities with capabilities for optimal network planning, helping achieve the vision for a statewide public IoT network.