The growing contamination of surface water by stormwater runoff parallels increasing urban development. Heavy metals, polycyclic aromatic hydrocarbons (PAHs), persistent organic pollutants (POPs), and contaminants of emerging concern (CECs) are discharged from point sources and washed from impervious surfaces into surface waters, impacting the ecology of these systems, food supplies, and the source waters for drinking water. Green stormwater infrastructure reduces peak runoff and removes contaminants, while providing the structure to support microbial communities and stabilize the soil. Vegetation health plays a large role in the effectiveness of green infrastructure installations, with unhealthy vegetation reducing uptake and transpiration rates, as well as filtration effectiveness. Monitoring the health of vegetation in stormwater green infrastructure can indicate signs of water stress, disease, as well as pollutant induced toxicity. Early detection of stress in vegetation can inform management and maintenance decisions. This study examines one remote and one ground-based method for monitoring biomass and primary production in two cells of a bioretention facility, one cell (Cell 2) containing sedges, rushes and grasses and the other (Cell 3) containing a mix of sedges, rushes, grasses and broadleaf vegetation. Ground-based measurements of the fraction of intercepted photosynthetically active radiation (fIPAR) produced unreliable results with challenges leading to over and under-estimates of intercepted PAR. fIPAR and the normalized difference vegetation index (NDVI) were poorly correlated for both bioretention cells, with coefficients of determination for Cell 2 (0.13) being lower than Cell 3 (0.49). However, radiometric calibration of UAS data from inexpensive sensors using the empirical line method (ELM) and three inexpensive ethylene-vinyl acetate (EVA) foam panels produced reasonable results. Linear regression equations were derived for the red and NIR bands of imagery with coefficients of determination from the three image sets ranging from 0.636 to 0.999. To increase the accuracy of the method, additional calibration targets should be used. The resulting NDVI data was tracked over a two-month period during the transition from spring to summer. The NDVI data was useful in exploring the spatial distribution of NDVI and how NDVI value coverage areas change over time.