Having accurate environmental data with a good spatial and temporal resolution is not possible using the traditional techniques. Wireless sensors provide a new way to attack these problems. Two years ago we started to build a wireless sensor network for soil monitoring, and 18 months ago we deployed the first system. Soon a second system was deployed in one of the Baltimore Ecosystem Study urban forest plots. Here we report the data we collected by the two deployments. Over 18M data points were collected on air and soil temperature, and soil moisture. Building the system was a very useful experience. First, it became obvious that the custom programming of the motes was a non-trivial problem, and it required more computer science expertise than most biologists have. Also the road from raw data to a calibrated spatio-temporal data set suitable for the scientific analysis proved to be long and cumbersome.  Third, the traditional data analysis tools and processes were not adequate to these data volumes.  Finally, we realized that we needed to redesign the hardware which also meant new software development. Our second generation system is based on the Telos SkyMote platform, which has lower power consumption and better sleep mode properties. We have designed our own 4-channel analog interface board with its own stabilized reference voltage. The embedded software on the motes can now be upgraded remotely, without opening the box.  The USB port of the mote is accessible from the outside, should a total code upgrade be necessary.  In order to connect the dense patches of sensor deployments to one another across larger separations, we have started experiments with long range radio transmitters that use the 900MHz band, and have a range of several miles. The first batch of this new system has been deployed in an urban forest