Networked, embedded sensors allow for an instrumentation of the physical world at unprecedented granularities and from unimagined perspectives. The advent of a ubiquitous sensing era is evident. Yet, sensor network techniques are still far from entering mainstream adoption due to multiple unresolved research challenges, especially due to the high development cost of sensor network applications. Therefore, in this dissertation, we propose to design, implement, and evaluate an environmentally-immersive programming framework, called EnviroSuite, to reduce sensor network software development cost. The goal of our research is to create reusable sensor network development support for the community and reduce the adoption barriers for a broader category of users, ultimately leading to a transition of sensor networks from a research concept to a general-purpose technology available for use for a wide variety of research, government, industry, and everyday purposes. Current sensor network programming practice remains very cumbersome and inefficient for several reasons. First, most existing programming abstractions for sensor networks are either too low-level (thus too tedious and error-prone) or too high-level (unable to support the diversity of sensor network applications). Second, there is no clear separation between application-level programming and system-level programming. A significant concern is the lack of a general middleware library to isolate application developers from low-level details. Finally, testing sensor network systems is particularly challenging. Sensor systems interact heavily with a (non-repeatable) physical environment, making lab experiments not representative and on-site experiments very costly. This dissertation is targeted for a comprehensive solution that addresses all the above-mentioned problems. The EnviroSuite framework consists of (i) a new programming paradigm that exports environment-based abstractions,(ii) critical middleware services that support the abstractions and separate application programmers from tedious, low-level details, and (iii) testing tools geared for in-situ experimenting, debugging, and troubleshooting. First, we introduce a new programming paradigm, called environmentally-immersive programming (EIP), to capture the common characteristics of sensor network applications, the rich, distributed interactions with the physical environment. EIP refers to an object-based programming model in which individual objects represent physical elements in the external environment. It allows the programmer to think directly in terms of physical objects or events of interest. We provide language primitives for programmers to easily implement their environmental tracking and monitoring applications in EIP. A preprocessor translates such EIP code transparently into a library of support middleware services, central to which are object management algorithms, responsible for maintaining a unique mapping between physical and logical objects. The major outcome of sensor networks is observations of the instrumented environment, in other words, sensory data. Implementing an application mainly involves encoding how to generate, store, and collect such data. EIP object abstractions provide simple means for programmers to define how observations of the environment should be made via distributed coordination among multiple nodes, thus simplifying data generation. Yet, the next steps, namely, data storage and collection, remain complicated and fastidious. To isolate programmers from such concerns, we also include in the support library a set of data management services, comprising both network�…