Clarifying water clarity: A call to use metrics best suited to corresponding research and management goals in aquatic ecosystems

U.S

its fluorescence (Stedmon et al. 2003), chlorophyll a pigment concentration (Chl a; mg m À3 ) commonly measured by its fluorescence (Holm-Hansen et al. 1965;Welschmeyer 1994), and total suspended solids concentration (TSS; mg L À1 ) also known as suspended particulate matter, total suspended matter, or suspended sediment concentration (Ball 1964). For CDOM and Chl a, conversions from fluorescence to aCDOM and Chl a pigment concentration may need to account for confounding factors, such as non-fluorescing components, Chl a daytime nonphotochemical quenching, and high-scattering environments that can affect the strength of the signal (Oestreich et al. 2016;Cremella et al. 2018;Carberry et al. 2019).
In this essay, we share a case study from the York River estuary (henceforth referred to as the York), a subestuary of the Chesapeake Bay. This location illustrates a clarity measurement puzzle, the likes of which occurs in many other water bodies globally. The dataset includes coincident measurements of Chl a, turbidity, K o , and Z SD from Fall (2020); coincident measurements of Chl a, turbidity, and K d from the Chesapeake Bay National Estuarine Research Reserve in Virginia (CBNERR-VA); and coincident measurements of K d and Z SD from the Chesapeake Bay Program (CBP) Water Quality monitoring program, all from 2014 to 2016 (Turner et al. 2022). Fall (2020) data were collected irregularly in time at eight stations within the York ( Fig. 2A). CBNERR-VA and CBP data were collected once or twice per month at the Goodwin Islands and WE4.2 long-term monitoring sites, respectively ( Fig. 2A).

Light attenuation is often estimated from other water clarity metrics
Estimating light attenuation from Secchi depth is problematic The simple hyperbolic relationship between K d and Z SD is widely represented as K d = α/Z SD such that the product of K d Â Z SD = α (Holmes 1970). However, often K d and Z SD do not adhere to a consistent relationship described by a constant α. The value of α has been found to vary widely in estuaries, lakes, and other aquatic environments across many latitudes, hydrologies, and climatic conditions (Lee et al. 2018;Bowers et al. 2020). Consequently, in turbid environments it is often disadvantageous to calibrate α (Preisendorfer 1986). In the present study, the York serves as an extreme example of this variability (Fig. 2B).
Instead, what information can be gained from the decoupling of K d and Z SD ? First, if the goal is to understand light penetration, measuring K d directly will be most useful (Table 1). Second, if the goal is instead to understand transparency or visibility, measuring Z SD alone may suffice. Finally, simultaneous measurement of K d and Z SD can be used to gain insight into how dissolved and particulate constituents interact with light, since the mismatch between K d and Z SD yields more information about light-blocking substances in the water.
This mismatch between K d and Z SD provides useful information about the constituents that limit light penetration. In moderately turbid waters, K d often has a smaller value (i.e., indicates clearer water) than that predicted based on a simple relationship with Z SD (α < 1.45). Smaller K d values than expected based on Z SD are often attributed to the properties of the suspended particles (Hou et al. 2007), such as reduced visibility of the disk due to increased forward scattering by small organic particles (Hern andez and Gocke 1988; Armengol et al. 2003;Effler and Peng 2012). For example, changes in particle scattering may contribute to the long-term shallowing of Z SD in the Chesapeake Bay while K d indicates minimal change or even an improvement in clarity (Gallegos et al. 2011;Harding et al. 2016;Testa et al. 2019;Turner et al. 2021). In the other direction, high quantities of CDOM can cause deeper Z SD compared to what K d would predict, due to high visibility yet rapid light absorption (α > 1.45) (Pedersen et al. 2014).
Water clarity is critical for SAV, which requires light penetration to depth for photosynthesis. During SAV restoration presented here include Secchi depth (Z SD ), light attenuation of photosynthetically active radiation (K d (PAR); referred to in this paper as K d ), turbidity, and beam attenuation. Metrics identified with specific components of the water column include colored dissolved organic matter (CDOM), total suspended solids concentration (TSS) (also known as suspended particulate matter, or SPM), and chlorophyll a concentration (Chl a). Some symbols are adapted from IAN UMCES media library.
work, the use of one water clarity metric to estimate another can over-or under-estimate depth limits of habitats. For example, in a fjord in Denmark, Z SD deepened over time, but K d remained relatively high due to large CDOM concentrations, causing Z SD to overestimate the potential habitat for SAV (Pedersen et al. 2014). Consequently, K d should be used rather than Z SD as a proxy for light penetration depths to infer SAV habitat quality, since the plants collect plane irradiance (Zimmerman 2003(Zimmerman , 2006.

Estimating K d from multiple metrics
Researchers and monitoring programs frequently estimate K d from a subset of other metrics. In oligotrophic waters, these relationships are based on the contributions to K d mainly from phytoplankton; thus, K d is most commonly derived from Chl a (Smith and Baker 1978;Baker and Smith 1982;Kim et al. 2015). In coastal waters, estuaries, and many lakes and rivers, K d is estimated from not only Chl a, but also CDOM (or salinity) and TSS (Woodruff 1996;Gallegos 2001;Fear et al. 2004;Xu et al. 2005;USEPA 2008;Feng et al. 2015;Cerco and Noel 2017;Turner et al. 2021). Other estimations of K d from multiple metrics employ semianalytical relationships (e.g., Gallegos 2001;Lee et al. 2005Lee et al. , 2007Zimmerman et al. 2015), enabling the use of satellite remote sensing to estimate water clarity at high spatial resolutions relevant to lakes and estuaries (Lee et al. 2015(Lee et al. , 2016.  (2008), where S = salinity, T = turbidity (NTU), and Chl = chlorophyll a concentration (mg m À3 ). Black lines in (C) and (D) indicate a 1 : 1 relationship between observed K d and predicted K d . In all subplots, Fall (2020) blue circles indicate scalar light attenuation (K o ) measurements in place of downwelling light attenuation (K d ). Values for K d and K o differ minimally in turbid, optically deep waters (Kirk 1994;Tilzer et al. 1995).
In the Chesapeake Bay and its tributaries, monitoring programs map K d spatially in shallow waters to assess habitat potential for SAV, making use of an empirical equation with turbidity, salinity (as a proxy for CDOM), and Chl a. These latter three metrics are collected with a flow-through method, increasing the temporal and spatial coverage. K d is measured directly at a few validation stations, but it is also often calculated from regionally determined empirical relationships with turbidity, salinity, and Chl a. This approach groups multiple subregions and time periods together to generate a relationship that describes a wider distribution of conditions (Dennison et al. 1993;USEPA 2003USEPA , 2007USEPA , 2008Moore et al. 2009;Reay 2009;Tango and Batiuk 2013).
Best practices for estimating water clarity using available resources

Report the metric that was measured
Perhaps the most important practice in measuring water clarity is to report the actual metric used. Some studies use  Wang et al. (2013) describe patterns in K d when the metric measured was actually Z SD , from which K d was derived using the conventional K d = 1.45/ Z SD relationship that is inherently less useful in turbid waters (Fig. 2B,C). However, if an empirical equation between metrics is required due to cost, sampling resolution needs, or other factors, then the method should be clearly communicated (e. g., CDOM estimated from salinity), and the cross-calibration data used should be made available. In the case of light attenuation, the use of a scalar (K o ) or downwelling (K d ) coefficient should be reported explicitly.

Measure K d with deep light profiles
In some cases, K d may be over-or under-estimated due to measurement error when light profiles do not extend deep enough into the water column. Collecting downwelling or scalar irradiance depth profiles over varying depth ranges can result in inconsistent estimates of the best-fit K d most relevant to the full photic zone, particularly when irradiance is not measured to a deep enough light penetration depth (Lee et al. 2018). Whenever possible, light profiles should be measured to the depth of 1% illumination to avoid measurement error.

Locally calibrated
Empirical models for K d need to be locally calibrated because the characteristics of the water's dissolved and particulate matter vary greatly, sometimes at a fine spatial scale. In the Chesapeake Bay, a single K d relationship applies only to some subregions, but not all. In smaller tributary rivers such as the York and Elizabeth Rivers, there is relatively more CDOM, while in larger tributary rivers such as the Potomac and Susquehanna Rivers, there is relatively less CDOM than salinity would predict (Cerco and Noel 2017). The diversity in contributions to K d likely results from the variety of river inputs; while the largest rivers have mountainous uplands and deliver relatively more sediment, the smaller rivers drain coastal plains and wetlands and deliver relatively more CDOM (Najjar et al. 2020;Henderson and Bukaveckas 2022). In addition, the response of K d to TSS may vary strongly with distance along a given estuary due to systematic variations in suspended sediment floc size, density, and organic content (Yard 2003;Fall et al. 2020).

Temporally representative
Ideally, a relationship used to estimate K d should incorporate measurements representative of different times and conditions, so that the variability over the targeted dataset is captured. A calibration performed during one season or tidal stage will likely not apply to the entire dataset of interest. In the York, an empirical relationship developed during a certain set of years (pre-2008) underestimates clarity compared to observations collected years later (Fig. 2D). The disagreement may be in part because 2014-2016 were hydrologically dry years in the Chesapeake Bay with lower nutrient and TSS concentrations, and generally clearer water than the early 2000s. These types of discrepancies have implications for management and restoration of important habitats. In the York example (Fig. 2D), directly measured K d would predict a greater spatial area suitable for SAV, while the empirical relationship from USEPA (2008) underestimates light availability. It could be argued that this somewhat conservative underestimation of habitat is a minor problem. However, overestimating light availability would result in negative ecological implications such as overpredicting the amount of suitable habitat for SAV (e.g., Pedersen et al. 2014).
Choose wisely: Select a water clarity metric targeting the research or management goal When planning water clarity measurements, it is recommended to select the most useful metric or metrics according to the specific application (Table 1). For example, if K d can be measured directly, it should be measured using a light sensor rather than estimated from other metrics. If an empirical relationship or a simple sensor is needed due to cost or other factors, use of best practices is recommended. When relevant to the goal, even the simplest water clarity measurements are valuable for environmental monitoring and restoration, whether by citizen scientists, non-profit organizations, or local sampling programs.
K d is the most relevant measure of water clarity for most research in aquatic ecosystems. K d is well-suited to research involving benthic autotrophs such as SAV in estuaries (Zimmerman 2003;Moore 2004) and lakes (Schwarz et al. 1996;Borowiak et al. 2017), benthic microalgae (Newell et al. 2002), kelp forests (Graham et al. 2007;Tait et al. 2021), and coral reef habitats (Baird et al. 2016;Jones et al. 2016). Scalar or downwelling light attenuation (K o or K d ) may be more appropriate for different applications. For phytoplankton photosynthesis, K o better represents the amount of total light energy available to cells from all directions. For benthic macrophytes, K d is more suitable because plants' flat leaves collect downwelling light (Zimmerman 2003(Zimmerman , 2006. For citizen scientists or non-profit organizations wishing to measure K d directly, low cost light intensity loggers are available as an alternative to expensive traditional sensors (Long et al. 2012).
Z SD is a representative measurement of visibility. Z SD applies to human perception of water clarity (Keeler et al. 2015;West et al. 2016) and water clarity's effect on property values (Klemick et al. 2018). Z SD is also relevant for sighted animals and their trophic interactions, such as visual foraging efficiency of zooplankton and fish (Aksnes 2007;Aksnes et al. 2004;Goździejewska and Kruk 2022) and interactions between predators and mesopredators (Benfield and Minello 1996;Baptist and Leopold 2010;Lunt and Smee 2014;Reustle and Smee 2020). Z SD also serves an important role in citizen science and community engagement (Crooke et al. 2017;Pitarch 2020) and in maintaining especially long time-series (Jassby et al. 2003;Opdal et al. 2019).
TSS is a representative measurement of particles that block light, directly affecting water clarity. However, TSS is truly a measurement of the mass of suspended particles rather than light penetration. Therefore, TSS may be most useful for research applications that benefit from quantifying the mass of sediments present in the water column, such as questions involving sediment resuspension, shoreline erosion, or river inputs (Fall et al. 2014;Palinkas et al. 2019;Tarpley et al. 2019;Moriarty et al. 2020). TSS can be a useful metric for shellfish research, since high concentrations of sediments can clog oyster gills and can blanket oyster reefs via deposition (Luckenbach et al. 1999;Beck et al. 2011;Gernez et al. 2014).
Chl a, turbidity, and beam attenuation are useful in that recent technologies allow sensors to be deployed at relatively low cost for long periods of time. Platforms like buoys, moorings, and floats are well suited for optical in situ sensors, thus many programs use these sensors for continuous long-term monitoring (Boss et al. 2018). Chl a and turbidity are often used in the field as stand-alone metrics. Chl a provides more information about the effects of phytoplankton on the underwater light climate, while turbidity provides more information about light scattering by suspended particles (Boss et al. 2009).
Looking to the future, the ability to collect data at high spatial and temporal resolution by a wider diversity of researchers is critical. While these measurements may be less directly representative of K d in dynamic systems, their importance should not be diminished. Provided that calibrations are well-performed, these simple longer-term measurements represent a fruitful way forward in water clarity research and monitoring. When factors contributing to light attenuation are not well understood, multiple measurements are needed to evaluate the relative magnitude and importance of the factors affecting light reduction. Use of multiple metrics is especially important in management of the causes of changes in water clarity.
River estuary. Kenneth Moore and Bridget Deemer provided helpful comments on early versions of this work. This manuscript was substantially improved by comments from two anonymous reviewers.