Development of Real-Time Soil Carbon Ecoinformatics Infrastructure Using Observational Network Data

To understand and model the temporal variability of soil respiration, we need high frequency, long-term data sets for model development and validation. Three observational stations equipped with Continuous Timeseries-Forced Diffusion (CTFD) probes were deployed in the summer of 2010 across a 1000 km transect in Atlantic Canada. At half hourly resolution, each observational station records soil CO2 efflux from two (2) probes and from a suite of meteorological sensors and peripherals. Each station is equipped with telemetry and data is continuously downloaded, quality controlled, processed, and made available for online display via several CGI, Java, and Perl scripts (http://fluxlab.stfx.ca/fieldsites/). This small network is intended to be the beginning developments of a larger Ecoinformatics Network. This presentation will display early data from this network and summarize real-time modeling efforts. The high-frequency observations show extremely dynamic systems which demonstrate CO2 efflux dependency to temperature and other important environmental drivers; pronounced increases in CO2 efflux after rain; differences across spatial scales; and short-term lags in data owing to gas or thermal transport. Other measurement methods (i.e. chambers) may miss many of these short-term flux variations in the absence of continuous data collection. Intra-site temporal observations (at sub-meter scale) show that spatially variable fluxes have similar scales of amplitude variation. All sites seem to show similar scales of temporal variability but CO2 fluxes can lag between probes across various time scales. These results suggest that site variably may be captured by measurements at only a few representative locations with high temporal frequency. Observation efforts will continue to monitor over winter and will provide unique data measuring fluxes under the snow pack at the soil interface. A key goal of this Ecoinformatics Network system is to develop improved soil models, capable of simulating soil process on multiple timescales. Most existing soil modeling efforts are focused on long-term processes and are useful for predicting long-term (decadal or greater) changes in soil carbon stocks but less useful for applications like Carbon Capture and Storage Monitoring, and Cap and Trade, both of which require modeling of CO2 emissions at sub-annual (daily/weekly) timescales. Taking advantage of our real-time soil CO2 data, we are pre-programming models and are testing them in an automated fashion. Model types being developed include artificial neural networks and other customized pattern recognition algorithms and real-time empirically based models which consider both historic data and the current state to calculate the future trajectory. This poster will discuss early results of this new modeling approach.