Long-term trends in water quality, land cover, and pesticide use in watersheds of the Southern Great Plains and their association with Prymnesium parvum

Prymnesium parvum is a harmful alga responsible for numerous fish-kill events worldwide and in the United States. Different biotic and abiotic factors may influence P. parvum distribution, growth, and toxicity. Several studies have been conducted on the association of water quality, lake inflow, and some biotic variables with P. parvum presence and bloom formation; however, the combination of water quality, weather and climate, hydrologic, land cover, and pesticide variables has not been applied before. In Texas, two of the most impacted river basins since 2001 are the Colorado River and Brazos River basins. This study examined one reservoir impacted by P. parvum and one unimpacted (reference) reservoir from each of these basins to address the following objectives at the watershed-scale: 1) compare and contrast environmental conditions in impacted reservoirs in the periods before and after the onset of toxic blooms in 2001, 2) determine environmental conditions that are unique to impacted reservoirs in the two study basins, and 3) identify potential environmental drivers of toxic blooms. The temporal scale of the variables was annual (average or cumulative as appropriate). Principal component analysis (PCA) was performed to address objectives 1 and 2. The period of record (POR) for objective 1 was 1992-2017 and only data from P. parvum-impacted reservoirs were used. Results showed that the use of most pesticides generally declined since 2001, coincidentally with the onset of toxic blooms, while the use of the herbicide Glyphosate and atmospheric CO2 concentration increased. The POR for objective 2 was limited to 2001-2017 and data from all four reservoirs were included. Results showed that P. parvum-impacted reservoirs have higher specific conductance, lower percent wetland area, and lower use of certain pesticides. A classification and regression tree (CART) analysis was used to address objective 3. All reservoirs were included in this analysis, the POR was 1992-2017, the dependent variable was toxic bloom occurrence (YES or NO at annual scale), and independent variables included non-redundant variables selected from a correlation matrix. Salinity (specific conductance) was the primary split, where the risk of toxic bloom increased greatly at levels ≥ 1613 µS/cm. The second split was the insecticide Terbufos, where the risk of toxic bloom increased at relatively low use levels when combined with high salinity. To avoid collinearity, percent wetland area was not included in the original CART model because it was highly correlated with salinity, but a new model was rebuilt replacing salinity with percent wetland area. The first split was percent wetland area, where the risk of toxic bloom increased with values below ≤ 0.43%. Overall, these results suggest that 1) the reduced use of potentially growth inhibitory pesticides and the increased use of Glyphosate – a known stimulant of P. parvum growth – coupled with the rising levels of air CO2 – also a known stimulant of growth – may have contributed to the onset of toxic blooms in the early 2000s; and 2) high levels of salinity, lower percentages of wetland area, and lower use of certain pesticides are unique traits of P. parvum-impacted reservoirs and are also associated with increased risk of toxic blooms.