Linking riparian vegetation to precipitation using NDVI at Yuma Proving Ground, Arizona

Abstract

Measuring precipitation in semi-arid landscapes is important for understanding the processes related to rainfall and run-off. However, rain gauges are sparsely distributed. Linear regressions comparing rain gauge and RADAR precipitation estimates revealed that RADAR data is often misleading especially for monsoon type storms. This study investigates an alternative way to map the spatial and temporal variation of precipitation inputs along ephemeral stream channels using NDVI derived from Landsat TM imagery. NDVI was derived on 26 pre- and post-monsoon season Landsat images across Yuma Proving Ground (YPG) in southwestern Arizona. The mean NDVI values along ephemeral stream channels explained 73% of the variance in precipitation totals from a nearby rain gauge for 25 monsoon seasons. A 0.0006 increase in NDVI per day between pre- and post-monsoon season imagery was found to indicate high precipitation inputs and possibly indicate flow events. A second set of Landsat TM imagery were used to relate gains in NDVI during seven winter seasons to precipitation recorded from a nearby NEXRAD radar station. The NEXRAD Stage IV radar data were found to be more accurate during winter precipitation events when associated with rain gauge stations (adjusted R2: 0.81 & 0.84). High correlations were found between NDVI and precipitation at the 32, 48, 64 and 96d time intervals, though each season varied. The number of precipitation events with >5mm at the 96d interval showed significant correlation (0.63 & 0.77) while the number of events with >10mm had less correlation. Moreover, the combination and analysis of these two NDVI datasets revealed that wet winters may influence the vegetation for more than four years into the future.