Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Chapter 3 Model Limitations and Other Recommendations Several issues affect the ability of the models to accurately predict the hydrologic and hydrodynamic changes that will accompany proposed water withdrawals into the future, in particular climate change and limitations imposed by the calibration process, as discussed below. Not all the Districtâs studies related to these topics have been completed, and the following discussion thus focuses on suggestions for addressing them. In addition, the increased amounts of urban stormwater runoff that would be a major source of water compensating for the WSIS withdrawals have important implications with regard to both water quality in the St. Johns River and on water quantityâflow and stage, as predicted by the hydrologic modeling. These implications and the role of stormwater management in influencing future conditions in the river are discussed below. CHANGING CLIMATE Some comments are warranted on the likely limitations of the models in a changing climate, given its growing importance. The hydrologic model for the 2030 land-use conditions will be tested using rainfall/potential evaporation data based on historic records. Hydrological scientists presently are grappling with how to handle the non-stationarity of climate, and this remains an open and important research question. While it is unlikely that simply assuming future rainfall will be the same as past rainfall is valid, models or methods to predict reasonable future rainfall distributions do not exist. Indeed, one cannot even adequately discern the direction of the change, that is, whether rainfall will decrease or increase in the St. Johns River basin over the next 20 years. The District thus can do little directly to address impacts of climate change in the WSIS. District managers should keep in mind, however, that decisions based on model predictions using historic rainfall conditions will need to be revisited as the state-of- the-science improves. Water supply levels that have had demonstrably low impacts under historic conditions could prove to be ecologically problematic under conditions of a changing climate. Conversely, it is also possible that the climate change could make more water available without ecological harm. Adaptive management and ongoing scientific studies are a necessary part of making sure that future withdrawals from the St. Johns River are ecologically sustainable. 24 P R E P U B L I C A T I O N C O P Y
Model Limitations and Other Recommendations 25 CALIBRATION LIMITS The hydrologic model was calibrated using observed meteorology and observed rainfall with fixed (1995) land use over the decade 1995 to 2006. Land-use changes that occurred over the decade of calibration and affected the measured runoff thus were subsumed into the calibration rather than represented explicitly in the calibration process. In effect, the model calibration knob for process âAâ was tweaked to account for changes in process âBâ. Inherently, this limits the reliability of the model outside its calibrated time span. There is little that the District can do about this problem, as developing year-by-year land use maps for use in calibration probably is impractical. The District, however, should acknowledge the ambiguity that this approach introduces into modeling results for the 2030 land use scenarios. There is no reason to expect that a calibration based on 1995 land use and 1995-2006 runoff, when applied to 2030 conditions, will result in a model with the same level of accuracy as that for the calibration period itself. The 2030 conditions move the model outside of its calibration range, and consequently the level of error is unknown. This was discussed in the outside peer review of the modeling approach (INTERA, 2009, page 7), where it was noted that: Given the uncertainty in the calibration,... caution should be used in predicting contributions of future land use. This would be especially true given large swings in impervious land use... Exaggerating a small error when up-scaling to the future land use condition would erroneously allow the model to simulate more water is available. Some insight can be obtained by a quantitative evaluation of the model outside its calibration range using newer data. It is recommended that the District apply the model (without further calibration) to 2009-2010 land use conditions and 2009-2010 observed rainfall and streamflow to provide a basic understanding of how the model behaves for a case outside the calibration range that has observed data. DEPENDENCY ON URBANIZATION AND RESULTING STORMWATER FLOW A key problem inherent in water budget studies on urbanizing catchments is that increasing impervious surfaces tends to decrease water fluxes through the vadose zone and increase surface water flow. The net effect is to decrease transpiration losses and thus increase the surface water available for withdrawal. Taken to its logical (but absurd) extreme, an entirely paved catchment has the maximum harvestable water. It follows that the predicted land use may be the critical driver of the study results for flow and level impacts. The predicted land use is taken as a âgivenâ in the Districtâs study, however, rather than as a variable subject to study and analysis. The District needs to evaluate how errors in land-use predictions affect the allowable withdrawals. More broadly, a review of how stormwater is considered in the WSIS is warranted. Post- development stormwater management requirements are specified in Chapter 40C-42 of the Florida Administrative Code and Section 10.2.1 of the St. Johns River Water Management District Applicants Handbook Management and Storage of Surface Waters. Future changes in land use are subject to specific requirements regarding the management of stormwater for quality and quantity impacts on receiving waters. At present, stormwater quality is primarily managed by the requirement of wet or dry detention or retention systems designed (depending on site conditions) to retain specified depths of rainfall/runoff from pervious and impervious areas, with P R E P U B L I C A T I O N C O P Y
26 Review of the St. Johns River Water Supply Impact Study: Report 3 the resulting storage volume being released later or recovered through soil infiltration within specified time limits. The effectiveness of detention/retention systems in removing the pollutants found in urban stormwater is variable (see Chapter 5 in NRC, 2009c). In general, such systems are moderately effective in removing suspended matter by simple settling processes, but they do a poor job removing pollutants present in the dissolved state or sorbed onto colloidal material. The chemical composition and quality of urban stormwater is highly variable (e.g., Brezonik and Stadelmann, 2002; NRC, 2009c Chapter 3) and depends on numerous factors, including characteristics of the rainfall event (e.g., duration, intensity), nature of the land cover and land use activities in the drainage area, and geographic and seasonal factors. Six major categories of pollutants occur in urban stormwater: nutrients, suspended matter, heavy metals, organic contaminants (especially pesticides from lawns and gardens and polynuclear aromatic hydrocarbons from paved surfaces), chloride salts from street deicing (in northern climates), and (potentially pathogenic) microorganisms. Among the heavy metals, copper and zinc are of growing concern with regard to urban runoff quality because of the widespread nature of urban sources for these metals and their toxicity to aquatic organisms. Several of these categories of pollutants (especially nutrients) are already of concern with regard to water quality in the St. Johns River, and expansion of urban land with the consequent increase in stormwater runoff loading to the river could exacerbate water quality problems in the future. Stormwater flow is managed by the implementation of temporary storage facilities for the attenuation of peak discharge to near pre-development rates (see Figure 3-1). Specific requirements are mandated for storms of various return periods and durations, depending on site conditions and location within the overall St. Johns River basin. Although the rate of stormwater runoff is constrained by regulation to predevelopment values (and then only for certain sized storms), the volume of runoff from post-development land uses (residential, industrial, and commercial) is greater than that of the land covers (agriculture, forest, and range) they replace. This is because of lower infiltration and evapotranspiration from FIGURE 3-1 Example Predevelopment and Post-Development (Managed) Stormwater Hydrographs. SOURCE: Tom Bartol, SJRWMD, personal communication, June 2010 P R E P U B L I C A T I O N C O P Y
Model Limitations and Other Recommendations 27 impervious surfaces than from surfaces covered with vegetation. The altered hydrologic regime that accompanies urbanization leads to a variety of well-known geomorphologic changes in streams, such that they have been named the Urban Stream Syndrome (NRC, 2009). The increase in stream flow and volume tends to mobilize sediment both on the land surface and within the stream channel, the latter leading to channel encision and degradation of riparian wetlands. The higher flow volumes and peak discharge caused by urbanization also tend to preferentially remove fine-grained sediment, leaving a lag of coarser bed material. The geomorphic outcome of these changes is a mix of enlargement of some stream reaches, significant sedimentation in others, and potential head-ward downcutting of tributaries. There is consequent deterioration of stream biogeochemical function and declines in species diversity and indices of biotic integrity in such streams. These effects are compounded by human actions to improve drainage, such as channel straightening and lining to reduce friction, increasing flow capacity, and stabilizing channel position. The potential deleterious impacts of these increased flows and volumes and changes in the temporal distribution of flows are not currently considered in the WSIS. The net impact of modeled land-use change on mean flow of the river at Cocoa and Christmas would be approximately 40 and 60 MGD, respectively (compare the second and third set of rows in Table 2-2). By itself, this contribution to flow would more than replace water taken under the half withdrawal scenarios at both of these locations, as well as the full withdrawal scenario at Christmas. Moving in the downstream direction, the cumulative amount of population growth-driven land-use changes and the resulting runoff increase. In total, the mean daily flow increase due to modeled changes in land use between 1995 and 2030 would be 90 MGD in the upper St. Johns River basin, and additional increases of 86 and 193 MGD would occur in the middle St. Johns River basin and lower St. Johns River basin, respectively (derived from Table 73, Cera et al., 2010). The modeling results thus indicate that future stormwater additions to mean flow of the river more than offset the proposed withdrawals. For model development, the stormwater management conditions in place in the 1995 base year are incorporated in HSPF implicitly via the calibration. The modeled stormwater impact on river flow as a result of forecast changes in land use during the planning period is based on those same calibrations. As a result of concerns expressed by this Committee with respect to quantity and quality of increased stormwater runoff, the District performed a comparative case study (Cera et al., 2010). To create the case study, the HSPF model of the Little Econlockhatchee River was modified to explicitly incorporate the quantitative stormwater best management practices (BMPs) specified by Florida statute. The difference between using the 1995 calibration relationship and the case study is only 1 percent, which suggests that use of the prior calibration is an acceptable simplification of the explicit modeling of BMPs relative to water quantity issues. To obtain an outside review of the Districtâs use of HSPF for the WSIS, the firm Intera was retained to review the watershed modeling effort. Among the recommendations for improvement, several were concerned with the way stormwater management was considered (INTERA, 2009), and these concerns were addressed in subsequent model runs. In particular, the percentages of âdirectly connected impervious areaâ of residential and industrial land use were reduced to near the recommended values, and these results were made consistent throughout the entire basin. Impervious area retention storage was universally increased to 0.1 inch, and storage attenuation for both pervious and impervious land areas was taken into account. P R E P U B L I C A T I O N C O P Y
28 Review of the St. Johns River Water Supply Impact Study: Report 3 Given that the WSIS demonstrates that the population growth-driven changes in land use will more than offset the proposed withdrawals, the Committee urges that as much attention be given to potential water quality and other environmental impacts of future increases in flows and levels and in the temporal distribution and routing of flows as to the potential for decreases in flow and levels. In particular, water quality impacts of increased urbanization on downstream reaches need further study, with particular attention to effects of increased loadings and concentrations of nutrients, heavy metals, and pathogenic organisms. In addition, the effects of urbanization, including increases in impervious areas and changes in stream morphology resulting from altered flows, on the biotic integrity of streams in the drainage basin also need more attention. CONFOUNDING PROCESSES Whether the hydrological model can adequately quantify confounding processes outside the calibration range is doubtful. âConfounding processesâ are processes whose effects are large but in the opposite direction such that they tend to cancel out, leaving a smaller residual effect. One might accept, for example, that the hydrologic model, calibrated for existing land use and rainfall, can reasonably predict impacts associated with a scenario for increased water withdrawals (reducing flows and levels). Likewise, one might accept that the calibrated model can be used to estimate the effect of a scenario for increasing urbanization (increasing flows and levels) for the existing withdrawals. These scenarios have confounding processes, however, because they have effects in opposite directions. Both scenarios have model errors that should be a fraction of the modeled impact. There is no reason to think, however, that the errors will be of similar magnitudes or opposite directions. It follows that the model error for a scenario combining withdrawal and urbanized land use may be much different than the predicted net impact. To illustrate this problem, assume that a measurable impact of water removal on some flow condition is â5 Â± 10%, and the impact of urbanization is +6 Â± 10%. The net impact is not likely to be +1 Â± 10%, but is instead likely to scale on the sum of absolute errors of the individual scenarios; e.g. â5 Â± 0.5 and +6 Â± 0.6, resulting in a net impact of +1 Â± 1.1 or +1 Â± 110%. Consequently, even though a model might predict the major impacts in a given direction by Â± 10%, when processes are confounding, the error generally scales on the magnitude of the individual processes, not on the net effect. The error thus may be large relative to the estimated impact. Even with the best possible model, confounding processes outside the calibration range can lead to uncertainty in the prediction that is larger than the magnitude of the predicted impact. The Districtâs analysis of model results across the scenarios should carefully consider which scenarios have confounding processes and which do not. APPLICABILITY OF HSPF TO WETLANDS HYDROLOGY The HSPF model that the District is using to predict hydrological changes for the different water withdrawal scenarios has limited value for modeling wetlands because HSPF output does not include water table elevation data. Essentially HSPF models water infiltration to the active groundwater, and this water either becomes base flow to the river or is âsunkâ to the P R E P U B L I C A T I O N C O P Y
Model Limitations and Other Recommendations 29 inactive groundwater (i.e., it leaves the system). The model does not include changes in water table elevations, it does not describe how the surficial groundwater interacts with wetlands as part of its output, it does not take into account water storage in wetlands, and it does not simulate wetland interactions with the river or with the unconfined water table aquifer. All of these are important components of the wetlandâs hydrologic signature. Floodplain wetlands in the watershed tend to be hydrologically âvertical,â i.e., water moves up and down through the soil more rapidly than it moves horizontally. Groundwater inflows thus are an important water source when the floodplain is not inundated, and HSPF does not model this process adequately. There are also issues of spatial resolution regarding use of HSPF for wetland impacts. The model is run for each of 411 subbasins that have been delineated in the watershed, but this does not provide enough spatial resolution to investigate wetland impacts. A key question that the wetlands workgroup has identified is how the duration of wetland inundation might be altered. To fully address this, spatially referenced wetland elevation and plant community data are needed. HSPF provides surface water level data in the subbasins but does not provide results on wetland hydrodynamics because the model cannot predict spatially variable changes in groundwater levels. To help address the above issues and shortcomings of HSPF with regard to wetlands, the wetlands workgroup is adopting the Hydroperiod Tool from the South Florida Water Management District. This tool estimates daily water depth over an area by subtracting the ground surface elevation (obtained from a digital elevation model or DEM) from an interpolated water surface elevation model based on river stage, and it can be used to estimate hydroperiods. In essence the wetlands workgroup will take the output from the HSPF model and distribute it over the landscape. Where adequate DEMs are available, this approach should aid the workgroup in accomplishing its goal of analyzing the correspondence between river stage and wetland hydroperiod. District scientists have made little progress in analyzing the empirical water level data available from transects established for previous work to establish minimum flow and levels (MFL) rules, and this lack of progress is hindering the progress of the wetlands workgroup. Having requested this analysis since May 2009, the committee is at a loss to understand the delay and urges the workgroup to undertake this analysis soon. These data have the potential to provide considerable insight into the response of the different wetland types to water withdrawals. When analyzed, these data should give the workgroup a much-improved understanding of floodplain wetland hydrodynamics. Finally, the Committee has several concerns about the proposed plans for data analysis by the wetlands workgroup. First, there are inconsistencies in the data needs presented and the plans for using the data. For example, an informal District document outlining the hydrologic data needs of each workgroup (and requested by this committee) states that salinity data are required by the wetlands workgroup to assess the potential for changes to wetland plant communities due to altered salinity levels, but no plan is presented for how these data will be employed. To tackle this question, the wetlands workgroup could focus an analysis of the impacts of increasing salinity on different wetland community types where information on salinity stress exists (i.e., literature values that indicate at what salinity level will a given community, such as hardwood swamps, begin to show stress). In contrast, they might focus on the dominant species in a given community (for example, cypress trees in hardwood swamps). It should be noted that much about the plantâs ability to tolerate salinity shifts will be tied to when the salinity peaks are experienced (season), for how long, and at what recurrence frequency. P R E P U B L I C A T I O N C O P Y
30 Review of the St. Johns River Water Supply Impact Study: Report 3 Second, questions remain about the GIS and STELLA modeling work planned by the wetlands workgroup in conjunction with the biogeochemistry workgroup. The groups propose to use DEM and vegetation maps in a STELLA model to project water elevations at certain locations. It is unclear whether the proposed STELLA model will be linked directly with the GIS database. In the committeeâs view, this is the only way to make the STELLA modeling output spatially explicit. Because the committee views this as an ambitious undertaking, it is unlikely to be an optimal use of staff resources at this point in the project, and may distract from the other efforts described above. P R E P U B L I C A T I O N C O P Y