By: Steve Culberson;
Ecologists are quick to identify the importance of usable, scale-relevant data when attempting to characterize and predict biological and ecosystem responses to physical changes in the environment. In particular, collection of ambient environmental data describing relevant habitat conditions (physical, biological; static, and dynamic) is the only way to understand biological condition and organism response to the environment. It’s also imperative that we use correct-scale data and information to inform our ecosystem management decision-making.
The frequent mismatch in the need for, and the availability of, environmental data is no more acute than when trying to discern climate effects (decadal time steps or more) to the well-being of biological species of concern (daily time steps or less) like Delta Smelt. Managing water systems on annual timescales for species that depend upon momentary conditions is perhaps not entirely sound.
Analytical Mismatch in Categorical Terms
Whereas climatic changes are (relatively) slow, an organism must achieve existence in every moment. Ecological models and survivorship forecasts for predicting biological performance frequently takes more resolute data (in time and place) than are normally, or historically, available. Furthermore, how an organism fares tomorrow and next week may have a lot to do with how it is feeling today and this week. A life history accumulates as existence through time.
One way engineers and ecologists keep tabs on the physical state of the San Francisco Estuary and associated watersheds is through documenting Delta Outflow (OUT) as reported in DAYFLOW (California Department of Water Resources). Additionally, it has long-been customary to characterize Estuary Annual Water Year Types according to 5 annual categories (Wet, Above Average, Below Average, Dry, and Critical) as calculated via the Sacramento and/or San Joaquin river indices (California Department of Water Resources), particularly where system management decisions and allocations are needed.
Unfortunately, this method of summarizing water yield from the watersheds is deficient when trying to build causal ecological links between the physical and biological components of the Estuary ecosystems. Many important water management decisions rely on annual, categorical descriptions of water yield (e.g., Delta Outflow, or OUT in DAYFLOW terminology), and yet even a cursory examination of this commonly used metric (Annual Water Year Type) leads me to think we need an alternative metric (or metrics) for describing the physical character of our watersheds if we want to improve our ecosystem of the Estuary and its management. Annual categorization of watershed condition can only do so much to inform the real needs of organisms that survive on a minute-to-minute basis.
There are examples (Mac Nally et al. 2010; Null and Viers 2013; Hammock et al. 2019; Hamilton et al. 2022) where analysis using categorical data is meaningful and useable, of course, but sometimes the use of convenient humanly-meaningful categories is found less than adequate (see discussions in, for example, Latour 2015; PPIC 2022), resulting in low P-values, poor predictive model performance issues, and inability to “prove” causation. It’s my intent in this blog to urge a change in how we approach and use this type of “physical condition leads to ecological outcome” analysis for management into the future. Improvements to our data repositories and computational abilities over the last decade support such a change.
An Example of How Categorical Descriptions for Management Fail Ecologically
Annual Water Year Type is too crude a metric (particularly in timescale) to usefully inform our need for improved understanding of ecosystem function when addressing Estuary water system management. We have available alternatives, and I’ll get to those after illustrating my point about mismatch with a simple example.
Relying on five water year types (categories) to describe our Estuary implies that we believe there are only five states of the Estuarine ecosystem. I don’t think we’ve demonstrably shown this, but for the sake of this argument and this blog I accept this degree of “lumping” for now. Importantly, another risky assumption of lumping water years into these categories asserts that all “within category” year are similar – that is, all “Wet” years are equal, all “Above Normal” years are equal, etc. Alarmingly, existing published ecological models attempting to be managerially-relevant rely on this long-standing historical decision regarding useful ecological Water Year categories. I suggest that this decision to treat all within-category years (“Wet,” for example) as equivalent as an input parameter for ecological models is faulty, and because of the prevalence of alternative, modern computing resources available everywhere, is no longer computationally necessary, as it once was. Why would I say this? How does this cloud our understanding and our subsequent attempts at predicting ecosystem performance based upon annual hydrologic summaries?
A few simple charts can show that this latter “lumping” characterization is problematic (that all “Wet” years are the same, for example, is NOT true), and the organisms we are attempting to manage conditions to support may well depend on the vagaries in conditions that this lumping causes us to overlook. I think this last statement (that of “overlooking nuances in environmental conditions”) is especially important given that many of our native and endangered species have evolved to exploit these ephemeral environmental vagaries. Such a systematic lumping of monitoring information leads us to ignore what is really happening at the individual and species-specific level, even while we explain to those interested that we have “included hydrology” in our ecosystem approaches.
My comments rely on visual inspection of graphical displays of all Water Years as mean daily outflow (OUT) in cubic feet per second (CFS) retrieved from DAYFLOW during July 2022 for the years on record from 1997-2016 (Figure 1). I previously conducted an undocumented exercise using the DAYFLOW record from 1930 to 2004 that supported a publication by Enright and Culberson (2010) that displays similar characteristics to the more modern excerpt of years included herein. This excerpt is used to facilitate ease of graphical display and a more general blog-appropriate discussion.
Three example “Wet” years were chosen to illustrate (Figure 2) my simple point that lumping masks important flow and stage characteristics (and likely many others as well) that can ecologically be quite different in years we describe or label as “equal.” As I mentioned above, I believe this can hinder the accuracy and precision of any ecological models constructed using this characterization and may cause us to reach erroneous conclusions about physical and biological drivers in the Estuary and to confuse what we may discuss as ecological “causation.”
Figure 2a is the mean daily outflow (OUT) for 2006, from the first day of the Water Year (October 1st, 2005) to day 365 of the Water Year (September 30, 2006); Figures 2b and 2c are for 2017 and 2019 Water Years, respectively. All of these years have been classified (categorized) as “Wet” years according to the Water Year Classification Index used by the California Department of Water Resources (accessed July 21, 2022). The Y-axis indicates the value Delta Outflow (OUT) for any particular day of the Water Year. As a reminder, for classification purposes, and as a variable used in many models designed to examine correlation or causation with/by stream flow in the Estuary, all “Wet” water years are deemed equal, ignoring the water flow features we can discuss using these individual outflow profiles.
For example, in 2006 we see an early outflow peak at approximately 95 days into the water year, or at about January 5, 2006, when the flows were more than 350,000 cfs. In contrast, peak flows in 2019 (another “Wet” year) did not occur until about 154 days into the water year (mid-March, 2019) and did not surpass 200,000 cfs.
It is likely that these different volumes and timings of peak flows (and build-ups and fall-offs to and from these flows, etc.) have different effects on the ecosystems of the San Francisco Estuary. There are organisms that time their spawning behaviors according to perceptions of “high flows,” for example. A difference of 100,000 cfs peak flow through the Delta may mean the difference between some floodplains getting inundated for longer periods of time (important to, say, Sacramento Splittail spawning success) or maybe not being inundated at all. It would be interesting to catalog all the differences we can plausibly find among all of these “Wet” years regarding inundation frequency, depth, duration, and water quality as a way to appreciate how all “Wet” water years are indeed not at all the same.
Fixing the Problem
With our ever-increasing need to construct ecological forecasting tools (for climate change reasons if for nothing else), and the now widely available computing resources that allow incorporation of continuously-collected data streams (as opposed to annual categorical data), why do many of our management-directed analytical efforts persist in characterizing hydraulic years as conforming to five pre-determined categories (even as we are already collecting flow data at 15-minute intervals, or more frequently!)? This doesn’t make ecological sense to me – organisms don’t respond to “categories,” they experience actual moment-to-moment conditions, every day – and our ecological understanding suffers when we predicate ecosystem behavior upon artificially-established categories. The problem becomes compounded when we choose management options that treat the entire system as having a single, categorical, “condition.”
I suggest we urge analysts who are interested in how hydrology influences the performance of estuarine biology to move away from older notions of “hydrologic performance,” and focus more fully on actual conditions that are already capably recorded on our data records. There’s no need to summarize data into categories when we have tools that can ingest and incorporate analysis and forecasting using an entire data stream of continuously collected variables. We are capable of this type of analysis these days, and in fact what our ecology is a product of – conditions in the field at every relevant time step – measured in days, hours, or minutes.
Perhaps relooking at the hydrological-ecological condition as part of an “ecological flows” perspective is what really is called for (PPIC 2020; Stein et al., 2021). What would a retrospective analysis reveal if the hydrodynamic conditions of the Estuary were treated more as states of a continuum rather than categories of yield estimations?
Closing Thoughts
At bottom, what I’m really on about is trying to use the best data we can (much of which is already available) and retaining resolution in our data collection and analysis when discussing management options rather than relying on summary-related practices that may have become out-of-date over time. Where we can improve our observational networks within the IEP and throughout the Estuary, we should. Where we can avoid summarizing annual conditions and include real-world measurements of real-ecology states-of-being, we should. Where our management discussions can appreciate the intricacies of ecology in our natural surroundings, they should. The citizens of California deserve this, particularly where it means we transform our analytical methods to reflect improved computational resources, better data management tools, and increasingly-informed conceptual relationships within and among ecosystem components. It’s time for us to think (and manage) ecologically!!
References
- Enright, C., and Culberson, S.D. 2010. Salinity trends, variability, and control in the northern reach of the San Francisco Estuary. San Francisco Estuary and Watershed Science, 7(2).
- Hamilton, S.A., Murphy. D.D., Weiland, P.S. 2022. An evaluation of the effectiveness of Fall Outflow Actions for delta smelt. A Technical Report from the Center for California Water Resources Policy and Management.
- Hammock, B.G., Moose, S.P., Solis, S.S., Goharian, E., and Teh, S.J. 2019. Hydrodynamic Modeling Coupled with Long-term Field Data Provide Evidence for Suppression of Phytoplankton by Invasive Clams and Freshwater Exports in the San Francisco Estuary. Environmental Management (2019) 63:703–717 DOI 10.1007/s00267 019 01159 6
- Latour, R.J. 2016. Explaining Patterns of Pelagic Fish Abundance in the Sacramento-San Joaquin Delta. Estuaries and Coasts (PDF) 39:233-247. DOI 10.1007/s12237 015 9968 9
- Mac Nally, R., Thomson, J.R., Kimmerer, W.J., Feyrer, F., Newman, K.B., Sih, A., Bennett, W.A., Brown, L., Fleishman, E., Culberson, S.D., and Castillo, G. (2010). Analysis of pelagic species decline in the upper San Francisco Estuary using multivariate autoregressive modeling (MAR). Ecological Applications (20)5: 1417-1430
- Null, S. E., and J. H. Viers (2013). In bad waters: Water year classification in nonstationary climates, Water Resources Research, 49, DOI10.1002/wrcr.20097.
- Public Policy Institute of California (2020). Making the Most of Water for the Environment. Report (PDF), August 2020.
- Public Policy Institute of California (2022). Tracking Where Water Goes in a Changing Sacramento–San Joaquin Delta. Policy Brief (PDF), May 2022.
- Stein, E. D., Zimmerman, J., Yarnell, S.M., Stanford, B., Lane, B., Taniguchi-Quan, K.T., Obester, A., Grantham, T.E., Lusardi, R.A., and Sandoval-Solis, S. 2021. The California Environmental Flows Framework: Meeting the Challenges of Developing a Large-Scale Environmental Flows Program. Frontiers in Environmental Science.