Evaluation of impact of climate change and anthropogenic change on regional hydrology

Abstract. General circulation models (GCMs) have been widely used to simulate current and future climate at the global scale. However, the development of frameworks to apply GCMs to assess potential climate change impacts on regional hydrologic systems and compliance with water resource regulations is more recent. It is important to predict potential impacts of future climate change on streamflows and groundwater levels to reduce risks and increase resilience in water resources management and planning. This study evaluated future streamflows and groundwater levels in the Tampa Bay region in west-central Florida using an ensemble of different GCMs, reference evapotranspiration (ET 0 ) methods, and water use scenarios to drive an integrated hydrologic model (IHM). Eight GCMs were bias-corrected and downscaled using the Bias Correction and Stochastic Analog (BCSA) downscaling method and then used, together with three ET 0 methods, to drive the IHM for eight different human water use scenarios. Results showed that changes in projected streamflow were most sensitive to GCM selection, however, projections of groundwater level change were sensitive to both GCM and water use scenario. Projected changes in streamflow and groundwater level were relatively insensitive to the ET 0 methods evaluated in this study. Six of eight GCMs projected a decrease in streamflow and groundwater level in the future regardless of water use scenario or ET method. These results indicate a high probability of a reduction in future water supply in the Tampa Bay region if environmental regulations intended to protect current aquatic ecosystems do not adapt to the changing climate.

The no pumping scenario (scenario 1) assumed that there was no human water use in the 253 region. For this scenario, irrigation demand, agricultural pumping, and urban pumping (including 254 CWF pumping, industrial and mining) were set to zero. For the no agricultural pumping scenario 255 (scenario 2) irrigation demand and agricultural pumping were set to zero however, urban 256 pumping assumed equal to the business as usual scenario. For the no urban pumping scenario 257 (scenario 3) urban pumping including CWF pumping, industrial and mining was set to zero and 258 irrigation demand and agricultural pumping were assumed to the business as usual scenario.

259
The agricultural adaption scenario (scenario 4) assumed that alternative water sources for 260 agricultural irrigation replaced 40 MGD (6 mm/year) of groundwater pumping for agricultural 261 irrigation. The other conditions were same as the business as usual scenario. The increased 262 agricultural demand scenario (scenario 6) assumed that irrigation demand increased by 40 MGD Variance-based sensitivity analysis is a global sensitivity analysis (GSA) method (Saltelli   281   et al., 2008, 2010) used to apportion the total model output uncertainty simultaneously onto all 282 the uncertain input factors, and thus is preferred over the local, one factor at a time, sensitivity 283 analyses (Homma and Saltelli, 1996;Saltelli, 1999). In this research the sensitivity of changes 284 between future and retrospective streamflow and groundwater levels projected by IHM 285 simulations was evaluated using the variance-based GSA method described in Chang et al.

287
Using the variance-based GSA method the first order sensitivity coefficient is expressed 288 as: where ( ) the total variance of Y over all . is a normalized index varying between groundwater levelaverage seasonal retrospective groundwater level, respectively).

307
Change in streamflow was much more sensitive to the choice of GCM than to the choice 308 of ET0 method or water use scenario for all river gages, both seasons, and both future periods 309 (Table 3). Similarly, projected changes in groundwater level were generally more sensitive to the 310 choice of GCM across monitoring wells and seasons. However, unlike the projected changes in 311 streamflow, changes in groundwater level at the monitoring wells were also quite sensitive to the 312 choice of water use scenario, except for well NWH-RMP-13s which is located the furthest from 313 the consolidate well fields (Table 4 and Fig. 1 index of groundwater level to water use scenario decreased in future period 2 over future period 319 1, due to the increased variability of GCM precipitation projections in future 2 versus future 1.
using nine GCMs, ten ET0 estimation methods, and three RCP scenarios.

335
The INTB was run to compare hydrologic response due to retrospective climate and 336 human water use scenarios to historical data and model predictions generated with NLDAS-2 337 data, as well as to evaluate alternative future climate change and human water use scenarios.  projections and future water availability in the Tampa Bay region.

477
The results that future streamflow projections are not strongly sensitive to water use  Groundwater pumping for water supply in the Tampa Bay region is regulated to maintain 487 groundwater levels that promote environmental protection of lakes and wetlands near wellfields.

488
In this study the relative importance of water use scenario and GCM selection on the percent of 489 the time that groundwater levels were above the target levels for four monitoring wells was (marked as † in the Table 8).

509
The mean change in percent of the time that groundwater is above the target level in the 510 monitoring wells for different GCMs (averaged over human water use scenarios) showed similar 511 results to those discussed previously. Two GCMs (GFDL-CM3 and MRI-CGCM3) projected a 512 statistically significant increases in the mean percent of the time that groundwater is above the 513 target level for both future periods ( Fig. 8b and Table 8). Four GCMs clearly projected a 514 decrease in percent of the time that groundwater level is above the target level in the future.

515
More GCMs showed significant differences in future period 2 than in future period 1 because the 516 differences among climate model projections increase in the later future. These results show that, 517 depending on how rainfall changes in the future climate, groundwater level regulations may be 518 difficult to achieve regardless of groundwater pumping scenario, and thus may have to change 519 with the changing climate.