THE ROLE OF DROUGHT AND STORMS IN HARMFUL ALGAL BLOOM EVENTS IN TWO GEORGIA PIEDMONT RESERVOIRS

Cyanobacterial harmful algal blooms (cyanoHABs) are an increasing concern in the Piedmont reservoirs of Allatoona and Lanier, Georgia. Effective prediction and prevention requires an understanding of the complex conditions which may trigger cyanoHABs in these reservoirs. This study seeks to determine the relative impacts of precipitation intensity, frequency, and duration as well as drought severity, frequency, and duration on cyanoHAB event onset, severity, and duration in these lakes, and put those relationships into the context of other hydrological and climatological factors within their watersheds.

Data sets for hydrologic and meteorological conditions were gathered and compared to a seven-year dataset of in-situ chl-a and water quality for sites across both reservoirs. Drought and storm events were identified, classified, and ranked. The in-situ chl-a data were used as a proxy for the presence of cyanobacterial harmful algal blooms during the bloom season of May through October. Satellite data were incorporated to increase the spatiotemporal resolution of the study and to model phycocyanin for verification of cyanoHAB events. Landsat 5, 7, and 8, and Sentinel-2 data from 2008 to 2014 were processed using several algorithms such as the Floating Algal Index, and were ground-truthed as well as compared against historical in-situ data where passover occurred within a one day margin (r2=0.64).

Regression analysis revealed relationships between cyanoHABs in the two reservoirs and extreme hydrologic events. Both in-situ chl-a data and algal blooms in Landsat scenes show a positive correlation with hydrologic influence from major tributaries. In-situ chl-a data was positively correlated with heavy flow in Allatoona (r2 = 0.65, 0.51) and Lanier (r2 = 0.81, 0.79). Remotely sensed algae showed relationships in zones of Allatoona nearest the Etowah River with weekly mean stormflow (r2 = 0.77, 0.71), but the relationship with PDSI varied between algorithms (r2 = 0.52, 0.04). Both reservoirs exhibited trends between bloom severity and streamflow conditions, but only after thresholds had been met, such as tributary discharge above a 3-year storm. Relationships between drought and precipitation indices are more complex, likely due to the temporal resolution of the data as well as other influencing variables.