Water quality monitoring of large freshwater bodies is usually slow and laborious, resulting in very sparse samples. This paper presents the design and preliminary field results from a network of low-cost mobile robots called NUSwan, a system aimed at addressing this issue. NUSwan robots can autonomously traverse a reservoir to provide real-time water quality data over the Internet, and allow users to modify the sampling missions interactively based on their interpretation of the data. The quality of measurements generated is comparable to that obtained from standard manual sampling. Moreover, the system incurs little logistics overhead while allowing higher monitoring frequency and spatial coverage. We present an example to show NUSwan's capability in detecting the emergence of water quality hotspots. Lastly, we demonstrate the potential of using data collected from a short mission carried out by NUSwan to rapidly assess the relationship between water quality indicators.