|
Water column transparency is an important feature of water quality and has ecological implications. Traditionally transparency has been measured with a Secchi disc. A Secchi disc is a small black and white disk that is lowered below the water surface until it is just visible. This depth is referred to as the Secchi disc depth. Secchi disc measurements are made manually. However, it is often possible to create a surrogate measure of Secchi disc transparency using turbidity. Here we apply this approach to estimate transparency from data collect remotely.
Background
Turbidity (Tn, NTU), as measured with a nephelometer or turbidimeter, measures light scattered by particles from a beam within a rather wide angle centered on 90º. Turbidity is a surrogate metric of the spectral average scattering coefficient (b, m-1), an inherent optical property of water that quantifies the intensity of the light scattering process. Various authors have reported Tn to be approximately equal to b. As a surrogate of b, Tn has been found in some cases to be a reasonably good predictor of common limnological optical measures of water quality, such as the diffuse spectral average light attenuation coefficient for downwelling irradiance (kd) and Secchi disc transparency (SD).
Estimation of Secchi Disc from Robotic Turbidity Measurements
Secchi disc transparency is highly sensitive to changes in b. Accordingly, some success in estimating changes in SD for individual systems, and differences in this metric in cross-section studies, from Tn measurements has been previously demonstrated. The relationship between SD and Tn has been described by
SD = N”/Tn
However, the coefficient N" can differ widely between systems and with time in an individual system, as it depends on the ratio of scattering to absorption, the relationship between Tn and b, and conditions that influence the SD measurement.
|
|
|
Figure 1. Regression of Secchi disc versus the inverse of robot turbidity measurements. From Onondaga Lake South Deep robot over the years of 2000, 2001 and 2002
|
|
The relationship between SD and the robotic Tn measurements for the upper waters of Onondaga Lake was evaluated from approximately paired (within two hours) observations in 2000, 2001, and 2002, years that included the substantially higher SD observations during the clear water phase. This population is far from normally distributed, but it does include a number of clear water phase (e.g., SD>3 m) observations. The best fit linear least squares regression expression for the SD versus Tn (for 1/Tn < 0.9) relationship was strong (r2 = 0.85) and highly significant (p < 0.0001), though the intercept deviated from zero.
The scatter around the best fit line reflects variations in N”, as well as inaccuracies in the SD and Tn measurements. The scatter for conditions outside of clear water phase intervals compromises the utility of Tn measurements to track SD dynamics when values less than 3 m prevail. Thus the robotic Tn measurements do not represent an adequate basis to establish future compliance with New York's swimming safety SD standard of 1.2 m. The primary utility of robotic Tn measurements as an estimator of SD is to aid resolution of the dynamics of water clarity experienced as part of the clear water phase, when relatively large changes in SD occur. This is illustrated for the May-June intervals of 2001and 2002 ((See Figure 2). The predictions of SD tracked the observations well, with the exception of the false high prediction for June 10, 2001. The daily robotic measurements of Tn support finer temporal resolution of SD dynamics than the weekly measurements of the long-term monitoring program. For example, from figure 2 we see that SD apparently decreased rapidly following the observation on June 18, 2001.
|
|
|
Figure 2. Secchi disc prediction utilizing regression from Figure 1 for the years 2001 and 2002.2
|
|
Two SD maxima, separated by a decrease, are indicated over the late May to early June interval of 2002, a temporal pattern that has been documented by more frequent SD measurements in an earlier year (1999). SD is predicted to have decreased rapidly after the peak of early June of that year.
|
