As a component of a multisensor approach to monitoring carbon sequestration sites for possible leaks of the CO₂ gas from underground reservoirs, a low-cost multispectral imaging system has been developed for indirect detection of gas leaks through observations of the resulting stress in overlying vegetation. The imager employs front-end optics designed to provide a full 50° field of view with a small, low-cost CMOS detector, while still maintaining quasi-collimated light through the angle-dependent interference filters used to define the spectral bands. Red and near-infrared vegetation reflectances are used to compute the normalized difference vegetation index (NDVI) and spatial and temporal patterns are analyzed statistically to identify regions of anomalous stress, which are then flagged for closer inspection with in-situ CO₂ sensors. The system is entirely self-contained with an onboard compact computer and is housed in a weather-proof housing to enable extended outdoor deployment.