Variability in the trophic level index in Lake Rotoehu from 1990 to 2021

Abstract

Setting an appropriate Trophic Level Index (TLI) target for lakes in the Rotorua Te Arawa region is critical for informing short- and long-term management decisions. Due to high variability in the observed TLI of Lake Rotoehu, coupled with contrasting understanding of historical water quality in the lake, the Bay of Plenty Regional Council requested a robust analysis of the drivers of the TLI over time to help inform the suitability of the current TLI target. To undertake this, we first quantified the uncertainty around the TLI in the 1990s, when the current TLI target was set based on the “best” observed water quality within the long-term monitoring program, to understand how a lower sampling frequency from 1990-2000 (~every 2-3 months) impacted TLI estimates. We found that while sampling fewer times in the year in the 1990s increased uncertainty around TLI estimates by 0.2-0.3 TLI units, a marked shift in water quality from a TLI of 3.6 in 1992 to 4.5 in 1993 remained evident. To understand if the main drivers affecting the TLI were different during the 1990s as opposed to later decades, we calculated the Pearson correlation coefficient between the TLI and several driver variables separately for each decade. Drivers in this analysis included meteorological variables, lake water level, and in-lake water quality not in the TLI calculation (e.g., bottom water nutrients, water temperature, stratification metrics), and the amount of aluminium sulphate dosed to the lake. While several variables were consistently important across all decades (e.g., bottom water nutrients, bottom water temperature, and mean monthly air temperature), average monthly water level and minimum windspeeds were only important in the 1990s, indicating that these variables had been related to major shifts in water quality seen during that time. Lastly, to better understand the relative importance of multiple drivers of the TLI and whether this importance changes over time, we conducted a moving window analysis. We fit autoregressive models, meaning the models included the previous month’s TLI estimate, as well as a single driver variable over a moving window of ~8 years, where each window moved forward one month at a time. We found that air temperature, bottom water temperature, and bottom water dissolved reactive phosphorus (DRP) were most often the top predictors of the TLI. However, the relative importance of average monthly water level increased sharply to the most important driver during a time period which also corresponded to very high water levels. Additionally, examination of model parameters over the simulation period demonstrated that the strength and magnitude of the relationship between the TLI and individual drivers changed over time, indicating that that relationships are not fixed through time. Overall, this work highlights the importance of re-evaluating the underlying relationships between the TLI and drivers over time, emphasising the dynamic nature of Lake Rotoehu. We demonstrate a clear shift in water quality between 1992 and 1993, likely driven by low water levels and windspeeds, which may have induced a pattern of increased external loading that still exists. Importantly, many of the variables which emerged as important for the TLI (water level, windspeed) will continue to vary with changing climate and are outside of the control of management. In light of this, while the current TLI target (3.9) may be feasible for Lake Rotoehu given historical observations and possible reference conditions, achieving this target consistently may be very difficult going forward due to catchment and climate pressures which have put Lake Rotoehu outside of an undisturbed reference state. This work can be built upon through future analyses which continue quantifying the changing relationships between the TLI and drivers over time, increasing the use of high-frequency buoy data to better inform water quality dynamics at shorter time scales, and testing our understanding of the drivers of water quality through predictions. The overall aim for future research would therefore be to establish water quality targets that not only align with community aspirations but are also technically feasible, especially in light of global change.