Proposal Cover Page Research Area Restoration Goal 1: Get the Water Right; Sub-goals (e) and (j) Program Area

A simple way to think about “getting the water right” is to reproduce the space and time patterns of water inundation and flows in a wet (dry) year to be the ones that would have been expected in a similar year under natural conditions. With this idea in mind, the South Florida Water Management District (SFWMD) developed the Natural System Model (NSM). This model has been very useful for highlighting the differences in hydrology (and ecology using coupled trophic system models) between “natural” and “modified” operations over the 1965-95 (now 2002) period. The NSM topography has been adjusted to represent current conditions, and simulations have been made to derive the corresponding hydrologic response to rainfall. If the NSM output is considered an accurate analog for the natural system, then indeed one could simulate a historical wet, dry or average year and use the results to prescribe the hydrologic targets in space and time. A strategy for operating the control structures could then be to arrive at releases that mimic these conditions, given the expectation that the current year is wet, dry or average. Several problems emerge.

First, we usually don’t know if the current year is wet, dry or average, until the wet season is well on its way. We could use probabilistic seasonal forecasts to address this shortcoming. Alternately, we could focus purely on near-real time operation, which would use daily information on rainfall and stage to update NSM hydrology and to then try to meet the resulting targets. By lagging operations by a few days or a week, one would not need forecasts, and could adapt to the desired targets. NSM requires a comprehensive input suite and a burn in time during simulations, and hence updating the projections through frequent NSM updates may or may not be feasible. However, Baldwin and Lall (2004) show that a Neural Network Model can accurately capture the relationship between rainfall and stage generated by the NSM. Hence, such a plan for “Rainfall Driven Operation” could indeed be developed.

Second, we still need a procedure or model that can relate rainfall and releases from control structures to hydrologic outcomes at specific locations in the ENP. The SFWMD has developed models like the South Florida Water Management Model (SFWMM) that can be used to predict such responses with the “real” system. Van Lent et al (2004) compared the predictions (on average) of the SFWMM and measured values at target locations as part of an assessment of the Intermediate Operations Plan (IOP). They found that (a) the model and measured flow responses under IOP typically had a consistent bias under different operational plans at many locations, and (b) there were some locations where the differences were marked. Since the SFWMM parameter calibrations sometimes carry over to the NSM, such biases can be expected in the NSM targets. Hence, using SFWMM for predicting responses, and for making decisions as to control structure releases may be problematic.

Third, as evidenced (SFNRC-ENP, 2004) under IOP, releases to meet specific goals in a certain area can lead to negative impacts in other areas. Some of these impacts relate to deficiencies in the manner the system is represented in the models, while others relate to explicit policies that address flood control or other issues. Similarly, there is considerable variety in hydrologic targets that can impact sensitive ecological populations. Meeting the hydroperiod at a given location is a seasonal target, while maintaining flow/depth in a certain section could be a shorter-term target. Reconciling such trade-offs through simple and static operating rules for a network of facilities may not be trivial.

Fourth, there is reason to expect that NSM stage and flow levels for a given weather/ climate condition may not correspond to what could result at specific locations and times, due to either model biases or because available point rainfall data are not representative of the space-time rainfall field. Consequently, meeting such targets through releases from control structures may become more difficult as the year progresses (e.g., if large releases are undertaken early in the year in an attempt to meet a NSM target, and the subsequent period is much drier).

Fifth, if only a near-real time operational strategy responsive to current, past or anticipated daily rainfall is followed, the ability to respond to longer term (including seasonal and multi-year) climatic anomalies may be lost. In this sense, while NSM targets may indeed represent reasonable analogues of natural system conditions, meeting these targets (e.g., very wet or very dry) may actually have adverse effects on marginal populations of species facing extinction. Thus, getting the water right, in the NSM sense, may not always get us to the desired ecological outcome. In such situations, an approach that actively manages the risk of species loss through appropriate operation of the storage and canal systems to deliver water where it would do the most good may actually be preferable. Given the dramatic multi-decadal climate variability in the region, using long climate records (including proxies) may be important to accurately define such risks and to develop the associated contingency plans. Such scenarios would then be potentially useful with seasonal climate forecasts, and may constitute an effective intermediate adaptation strategy to the potential anthropogenic climate change.

Sixth, a large number of performance measures (PMs) and habitat suitability indices (HSIs), reflecting different restoration goals have been posited for the restoration project. Even if we get the water right, we will need to demonstrate measurable improvement in terms of these measures for the restoration project to be a success. We do expect that many of the PMs and HSIs are highly correlated mutually and also with the NSM hydrologic targets. Thus, a mathematical decomposition of the large number of PMs/HSIs into a smaller set of indices that captures the commonality and the divergence between these measures may allow for the exploration of robust operating strategies for the system that would be “best” in the sense of making the most progress towards these objectives, as an alternative to trying to meet selected NSM targets. The PMs/HSIs could be classified into weekly (at a certain time of year, e.g., nesting opportunity), seasonal or multi-year goals prior to the decomposition, and then be used to develop “objectives” for analyses that reflect those planning and operational time scales.

• 
  Short-term, seasonal and long-term goals for hydrology, ecology and human uses.
  
• 
  Trade-offs in space, time and objectives
  
• 
  Uncertainty in and representativeness of rainfall, climate and models
  
• 
  Predictive schemes that can relate control structure releases to hydrologic outcomes
  
• 
  Correction of system trajectories through adaptive modification of the operating policy given climate/weather forecasts and realized system stage from recent releases.
  
• 
  Whether and how NSM derived targets should be met, or the system should be operated to maximize a basket of agreed upon performance measures
  
• 
  Risk based optimization of release policies and operations that makes superior choices transparent for stakeholder decision making as to seasonal and longer plans.
  

The proposed research intends to develop an operational methodology to systematically address these issues building on available data and models. Collaboration with researchers at the ENP and SFWMD will form the basis for the development of appropriate statistical and optimization tools as part of a decision-making umbrella that bridges operations, planning, assessment and monitoring. The hydrologic modifications of the region have translated into larger (flood) releases through the ENP in the wet season, and a drier dry season. These modifications are restoration targets. Recognizing that Lake Okeechobee and the WCAs can be thought of as seasonal storage reservoirs, it makes sense to think of end of season storage targets in these reservoirs, with sub-seasonal operation to meet flood contingencies and ENP water delivery targets. A natural temporal decomposition of the decision problem then follows. We intend to explore this research direction, assuming that simulation and optimization models will be assembled or developed as needed.