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

Given a seasonal operation plan, weekly or shorter-term control structure release policies can be developed constrained by the storage and system rule curves. In this case, system simulation tools that can accurately predict the stage response at target locations given a combination of upstream releases and rainfall are needed. The operator needs guidance as to how to maintain a seasonal trajectory by making weekly/daily release decisions at the control structures that allow system storages and water levels to stay on the right side of the system rule curve, while adapting to real time feedback (i.e. stage response to recent releases) and forecasts (e.g., near term weather forecasts). Since the seasonal system rule curve was derived considering a specified reliability, the daily/weekly decisions need to seek this reliability over seasonal operation. A real time system control model would have to be developed for this purpose.

A number of hydrologic system simulation models (NSM, SFWMM, 2*2) are available through the SFWMD. These models typically have a resolution that is coarser than is needed for ecological analyses. Specifically, they can exhibit strong local biases in their sensitivity to releases from the control structures. Where these biases are systematic, they can be easily dealt with in the operational setting. However, where these biases are due to an inability to model subscale processes or hydrologic features that have a large local impact (as is apparently the case in the C-111 canal area), it is difficult to make them operational. In such a setting, it may be more effective to directly build a statistical model that connects existing and emerging data on releases, rainfall and stage or flow response. Given 20 years of training data, a neural network model appears to effectively approximate the dynamics of the NSM model. This suggests that given enough real data, a statistical approach may be as effective as one of the deterministic models currently in use. However, such an approach needs careful though to enable effective pooling of information across multiple recording locations at multiple times. It may indeed be useful to consider applying the statistical model to correct the bias (for instance) the SFWMM model results. In the general case, we seek to develop a statistical approach to simulating response to release and rainfall that can effectively use both the available and evolving data and the database of concurrent simulations of existing SFWMD models.

The simplicity of just trying to attain the predicted NSM stage given current conditions and projected rainfall is very attractive as a yardstick for developing the seasonal or real time optimization plans. However, as noted earlier, particularly for real time operation, this strategy may not guarantee improvement of real outcomes in terms of the various PMs and HSIs. On the other hand, trade-offs between and interpretation of the very large number of state variables/targets/ PMs are a concern. In reviewing the IOP and HSI reports from the ENP and the SFWMD we have noted that at a given location, and for a given time period, many of the proposed PMs, HSIs and water levels are highly correlated. Indeed, many of the PMs and HSI's are monotonic functions of the same target variable (e.g., hydroperiod duration). Further, many of these measures are highly correlated spatially. Thus, at least in as far as seasonal planning and assessment is concerned it may be possible to dramatically reduce the dimensionality of the assessment process. Development of a reduced set of objective performance measures that derive from the NSM targets and the HSIs is consequently a goal.

Probabilistic considerations typically follow from the random nature of streamflow and rainfall inputs into the reservoir, i.e., weather and climate uncertainties. Consider dry season operation. The reservoir operator would like to begin the dry season with the reservoir full, or with the reservoir storage above some target level with some probability. Given this probability, and the corresponding reliability for meeting the dry season demands, a Monte Carlo simulation of climate scenarios can be used to identify the end of season target level, as well as the week by week lower bound (i.e., rule curve) on storage in the preceding wet season. Since flood control is a competing factor in the wet season, the wet season operating rules also need to consider flood storage and release policies from the control structures. Weather scenarios for these rules could be derived through system simulation using a space-time daily weather generator. In the ENP context, the climate/weather scenarios allow for a system wide probabilistic evaluation of the impacts of a specific operating rule or release policy on the performance measures selected. The development of climate/weather forecasts and long-term simulations is next.