In the Internet-of-Things (IoT) things of the physical world communicate with humans and other objects by means of the Internet. The concept extends the traditional Internet of Computers (IoC) that uses computers such as desktop PC, notebooks or smartphones as access points to the Internet by a multitude of end points represented by physical objects. Since things, human beings but also events in the physical world possess various spatial properties, their actions, observations or happenings inhere geospatial information. Hence, the exchange of digital geospatial information in the IoT leads to the Geospatial IoT. The Geospatial IoT is fundamentally different from established geoinformation technologies. Things in the Geospatial IoT provide geodata with a much higher frequency and quantity. Its integration is foreseen as the most revolutionary change in the history of geoinformation technologies. The ubiquitous geodata collection and provisioning through the Geospatial IoT as an all-encompassing infrastructure holds huge potential for better spatially understanding, modeling and visualizing of our natural and artificial ecosystems. However, a lot of challenges occur in integrating spatial and spatiotemporal data from IoT devices into new or already established systems. Novel requirements concerning architectures, messaging mechanisms and technologies such as scalability or efficiency appear, which must be analyzed and considered when implementing an appropriate infrastructure. This thesis addresses different research questions regarding an infrastructure for the Geospatial IoT and its integration into established geoinformation technologies. Approaching these issues, typical concepts, architectures and buildings blocks of the IoT are investigated first. This holds especially for recent developments in the field of IoT communications on different layers. Based on these typical structures, a concept for a Geospatial IoT architecture is designed. The conceptual architecture uses GeoEvents as a base data type for exchanging spatiotemporal messages between actors in the Geospatial IoT. The data type is derived from real-world geospatial events and states, which represent occurrences and states of continuants. Basically, a GeoEvent is a four tuple with a name, a spatial component, a temporal component and a message payload. With that event type defined, a GeoEvent-driven architecture for the Geospatial IoT can be specified. It follows an event-driven pattern, so that GeoEvents are send to interested consumers when they occur. Thereby, consumers may specify their interest in GeoEvents by GeoSubscriptions. In the architecture, GeoEvents are distributed by a GeoEvent processing engine, which evaluates the meta information of the GeoEvents against filters of the GeoSubscriptions. The IoT communication protocol Message Queuing Telemetry Transport (MQTT) is chosen to implement the GeoEvent-driven architecture for the Geospatial IoT. It meets requirements for resource-constraint devices such as small message size and efficiency, but has also beneficial architectural properties such as the messaging paradigm or the scalability of the server. In the thesis, the topic-based publish/subscribe protocol MQTT is extended by a GeoEvent and a GeoSubscription data type. In the developed extension called GeoMQTT, new messages are introduced to encode these data types. A GeoMQTT broker distributes the GeoEvents between the clients based on the GeoSubscriptions. To achieve this, filtering capabilities on temporal and spatial components are integrated, so that GeoMQTT can be used as a content-based publish/subscribe protocol. Several clients in different programming languages are implemented, as well as a GeoMQTT-SN extension for unreliable networks. The latter one focuses especially on small message sizes. GeoMQTT is evaluated with respect to multiple requirements for IoT environments. The expressiveness of the message types as well as the efficiency of the subscribe mechanisms meet the requirements for the Geospatial IoT. Further, scalability is ensured by the GeoMQTT broker and, thus, prepared for a dynamically increasing number of things in the Geospatial IoT. Finally, the spatiotemporal messaging protocol GeoMQTT is integrated in established geoinformation technologies. For instance, a GeoMQTT plug-in for QGIS is implemented to receive GeoEvents in a desktop GIS in real-time. The WPS for executing geoprocesses is extended by new input and output data types, namely GeoPipes, so that processes on geospatial data streams can be invoked. Additionally, bridges to Sensor Web services and to REST servers provide end users or software agents access to historical or new GeoEvents by means of the WWW.The thesis approaches and solves some of the challenges and tasks occurring in building an infrastructure for the Geospatial IoT. The prototypical implementation of the GeoEvent-based architecture and its related services show that things in the Geospatial IoT can be interconnected efficiently by a spatiotemporal messaging mechanism. Real-time access is realizable by services and established geoinformation technologies may also participate. Nevertheless, there are still many ongoing issues to solve, which are addressed in the end of the thesis.