CLIMATE AND HYDROLOGIC CHANGE ASSESSMENT FOR GEORGIA

Transcription

1 Lake Lanier 15MW Norcross Morgan Falls 16.8 MW Atlanta Whitesburg 82 MW West Point Bartletts Ferry 173 MW Goat Rock 26.3 MW Oliver 6 MW N. Highlands 29.6 MW Colubus W. George 168 MW Andrews 36 MW Woodruff Chattahoochee Blountstown Suatra Range: (ft) (bcf) Range: 62-64(ft) (bcf) (bcf) Range: (ft) Bainbridge Range: (ft) (bcf) Griffin Montezua Albany Newton Legend Reg. Reservoir Runoff Reservoir Control Point Local Inflow Withdrawal CLIMATE AND HYDROLOGIC CHANGE ASSESSMENT FOR GEORGIA F. Zhang and A. Georgakakos AUTHORS: School of Civil and Environental Engineering, Georgia Institute of Technology, 79 Atlantic Dr. Atlanta, GA REFERENCE: Proceedings of the 211 Georgia Water Resources Conference, held April 11-13, 211, at the University of Georgia. Abstract. This article describes a cliate change and hydrological ipact assessent for several basins in Georgia. First, a new statistical technique, Joint Variable Spatial Downscaling (JVSD), is developed to produce high resolution gridded hydrological datasets for the Southeast US fro 13 different Global Circulation Models (GCMs). A luped conceptual watershed odel (Georgakakos et al., 21) is then eployed to characterize the hydrologic responses under the historical cliate and the future cliate scenarios. The historical (baseline) assessent is based on cliatic data for the period 191 through 29. It consists of running the hydrological odels under historical cliatic forcing (of precipitation and teperature) for the 19 year period fro 191 to 29 (in onthly steps). The future assessent consists of running the Georgia watershed odels under all A1B and A2 cliate scenarios for the period fro 2 through 299 (1 years) in onthly tie steps. For the baseline scenarios and each of the 26 future cliate scenarios (i.e., 13 A1B scenarios and 13 A2 scenarios), this study assesses the changes of both cliate variables (i.e., precipitation and teperature) and hydrologic variables (i.e., soil oisture, evapotranspiration, and runoff) for each watershed. The results show that: (1) the 26 IPCC future cliate scenarios (2-299) do not indicate any long ter change in average precipitation; (2) the precipitation distribution is expected to stretch becoing wetter and drier than that of the historical cliate; (3) teperature and potential evapotranspiration (PET) show consistently increasing historical and future trends; (4) soil oisture storage exhibits a declining trend historically and for future cliates; and (5) watershed runoff, and thus river flow, exhibits a siilar historical decline across all Georgia watersheds. INTRODUCTION Current water resources planning and anageent practices in the Southeast US ay be vulnerable to the potential ipacts of future cliate changes on both water quantity and water quality. These are largely due to the hydrological stationarity assuption aong policy and decision akers who are often unconcerned about cliate and environental changes over the coing decades. An integrative approach to assessing cliate change ipacts on water resources by following a well-defined assessent fraework is crucial to regional water resources anagers (Georgakakos et al., 1998). GCM Scenarios Bias Correction Downscaling OBS Cells (~12 x 12 k 2 ): ly & Daily (5 Years); Teperature, Precipitation GCM Cells (~25 x 25 k 2 ): ly & Few Daily A1B: 21-21; (1 Years) A2: 21-21; (1 Years) Generate consistent cliate forcing sequences, i.e., rainfall, teperature etc. Watershed Hydrology Siulate soil oisture, evapotranspiration, runoff, etc. River/Reservoir Planning & Manageent Assess Ipacts on water uses & users. Engage Stakeholders & Support Planning & Managnt Processes. Figure 1. Integrated Modeling Fraework The ai of this study is to develop an integrated cliate change assessent fraework and to generate reliable data and inforation to support the on-going regional water resources planning and anageent efforts in Georgia and the southeast US. Figure 1 illustrates the integrated odeling fraework coprising three ain coponents: (1) processing of general circulation odel (GCM) scenarios for bias correction and downscaling (cliate coponent); (2) developing physically based conceptual odels for all ACF sub-watersheds (hydrology coponent); and (3) representing all ACF regulation infrastructure and water uses within an adaptive river and reservoir regulation and assessent odel (water resources coponent). CLIMATE CHANGE SCENARIOS Many researchers have deonstrated the physical science basis, ipact, adaptation and vulnerability of our changing cliate and environent. The cliate change issues have been addressed in a couple of literatures for the ipact on water resources (Lettenaier and Rind, 1992; Sta et al., 1994; Conway, 1998; Wood et al., 24;). Aong these studies, as an iportant tool for qualitative ipact

2 assessent, general circulation odels (GCMs) are broadly used. General circulation odels (GCMs) are scientific tools used to assess the future global cliate response associated with various greenhouse gas eission scenarios (IPCC WGI, 27). The GCMs represent (through a large syste of partial differential equations) the coupled atospheric and oceanic processes currently understood to govern the Earth s cliate. Cliate scenarios are generated by the nuerical integration of the underlying equations over space and tie. Thirteen different GCMs, selected scenarios fro which (corresponding to eission scenarios 2CM3, SRESA2, and SRESA1B) are utilized in this study (Georgakakos, Zhang, and Yao, 21). The new downscaling approach differs fro other existing statistical downscaling ethods in that ultiple cliatic variables are downscaled siultaneously and in a consistent way to produce realistic cliate projections. Figure 2 shows the flow chart of the JVSD procedure.it starts with a bias correction step: it uses a differencing process to create stationary tie series and the joint variables quantile-based appings (also known as joint epirical cuulative density functions) to adjust differentiated tie sequences for Global Circulation Model (GCM) outputs. Then, in the next step, spatial disaggregating, JVSD uses the historical analogue approach and the historical analogues are identified siultaneously for all atospheric fields being downscaled. JOINT VARIABLE SPATIAL DOWNSCALING GCM outputs are usually inadequate to capture the spatial variability at regional or local scales necessary for hydrological applications. Such conclusion is corroborated by the large uncertainties arising fro using different odels driven by the sae scenarios (Tebaldi, 25; Mitchell and Hule, 1999; Mujudar and Ghosh, 28). To overcoe the liitation of directly using GCM outputs, and to produce high resolution gridded hydrological datasets suitable for regional watershed odeling and assessents, a new statistical downscaling technique, Joint Variable Spatial Downscaling (JVSD), was developed by Zhang and Georgakakos (211). Input Historically Observed on OBS Grid 1 Bias Correction Upscale Historical Tie Series HYDROLOGIC MODELS A luped conceptual watershed odel for hydrologic ipact assessent of cliate changes has been developed by Georgakakos et al. (21). The watershed odel includes several water balance eleents with nonlinear storage-release functions at onthly tie resolution. The odel forulation is siilar to that of a luped paraeter Sacraento odel type, and is intended to siulate the hydrologic processes of infiltration/percolation, evapotranspiration, and surface and subsurface runoff (Figure 3). ET 2 ET 1 P ET S Ipervious Areas Q ip u GCM Control Run Data Difference Observed, GCM Control, and GCM Future Tie Series u 1 S 1 u 1 GCM Future Run Data Correct Biases for Differenced GCM Future Tie Series Reconstruct Bias-Corrected Future GCM Tie Series u 12 S 2 u 2 Q = Q ip + u + u 1 + u 2 2 Spatial Downscaling Expand Historical Data into 15, 1, or 5 day Sequences until Reconstructed Sequences Fall within the Range Find the Nearest Historical Analogues Over the Region and Create a ly Sequence Downscale GCM Future Data on OBS Grid Using the Historical Analogue Sequence No Output Downscaled Future GCM on OBS Grid Within Upscaled ly Historical Data? Yes Find the Nearest ly Historical Analogue Over the Region of Interest Downscale GCM Future Data on OBS Grid Based on the Historical Analogue Figure 2. JVSD Procedure Flowchart u G Available Observations: P, T, PET, Q, Area, Terrain, Land Cover. Model Calibration: Storage capacities, runoff functions, and percolation functions. Model Outputs: S, S 1, S 2, ET, ET 1, ET 2, u, u 1, u 2, Q. Figure 3. Hydrologic Modeling Syste Model inputs include precipitation and potential evapotranspiration deand (PET) averaged over the watershed area. The odel includes one surface and two subsurface oisture storage layers, with water contents S, S1, and S2 respectively. Water enters the top odel layer as precipitation, P, and, after soe losses to surface retention, it infiltrates/percolates to the lower storage layers. Precipitation falling on ipervious areas contributes iediately to runoff (QIp). Storage layers ay be depleted by evapotranspiration ET, ET1, and ET2, or run-

3 off to the strea u, u1, and u2. Evapotranspiration depends on PET as well as storage. Runoff depends on storage through the storage-runoff functions u(s), u1(s1), and u2(s2). Total runoff, Q, to the strea is the su of all runoff contributions: Q = QIp + u(s) + u1(s1) + u2(s2). The infiltration/percolation functions u1 and u12 are key odel eleents and depend on various odel variables. In addition to the evapotranspiration, storagerunoff, and infiltration/percolation functions, odel paraeters include storage capacities. These functions and paraeters are calibrated fro conteporaneous observations of precipitation, PET, and total watershed runoff. The ACF watershed odels developed in this study have a onthly tie resolution. The function fors and paraeters of the odel are data driven and they are estiated using a recursive identification ethodology suitable for ultiple, interlinked odeling coponents. By using such ethod, each watershed is calibrated by using area averaged precipitation, PET and unipaired flow sequences: (1) ly precipitation sequences for each ACF watersheds are generated by aggregating of gridded data over watershed areas. The gridded dataset were obtained fro the PRISM (Paraeter-elevation Regressions on Independent Slopes Model) cliate apping syste (Daly et al., 1997). (2) ly average ground air teperature sequences are also obtained fro the PRISM dataset. The Haon ethod discussed by Lu (25) is used. (3) Unipaired flows are the river flows that would have been observed in the absence of huan water use and regulation. The unipaired flow sequences used in this study were initially developed by the U.S. Ary Corps of Engineers (USACE) as part of the ACF Coprehensive Study for the period fro 1939 to This dataset was extended to 21 by USACE Mobile District in Septeber 23. A further extension to 27 was carried out recently by the Georgia EPD as part of the Georgia Water Plan. CLIMATE AND HYDROLOGIC ASSESSMENT The calibrated hydrologic odels are eployed to characterize the hydrologic responses under the historical cliate and the future cliate scenarios for all Georgia watersheds. We will take the Apalachicola-Chattahoochee- Flint (ACF) River Basin as an exaple for the assessent. The historical (baseline) assessent is based on cliatic data for the period 191 through 29. It consists of running the hydrologic odels under historical cliatic forcing (of precipitation and teperature) for the 19 year period fro 191 to 29. The future assessent consists of running the Georgia watershed odels under all A1B and A2 cliate scenarios for the period fro 2 through 299 (1 years). For the historical scenarios and each of the 26 future cliate scenarios (i.e., 13 A1B scenarios and 13 A2 scenarios), this study assesses the changes of cliate and hydrologic variables (i.e., soil oisture, evapotranspiration, and runoff) for each watershed. The results show that: (1) the 26 IPCC future cliate scenarios (2-299) do not indicate any long ter change in average precipitation; (2) the precipitation distribution is expected to stretch becoing wetter and drier than that of the historical cliate; (3) teperature and potential evapotranspiration (PET) show consistently increasing historical and future trends; (4) soil oisture storage exhibits a declining trend historically and for future cliates; and (5) watershed runoff, and thus river flow, exhibits a siilar historical decline across all ACF watersheds. As an exaple, Figure 4 shows the frequency curves of precipitation, teperature, potential evapotranspiration, and runoff for Buford watershed under A1B scenario. The figure leads to the following observations: (1) While on average (i.e., in the vicinity of the 5% percentile), Buford precipitation is not expected to change relative to the historical baseline, the precipitation distribution is expected to stretch becoing wetter and drier than that of the historical cliate. (2) Most future scenarios result in higher PET, evapotranspiration, and lower soil oisture storage. This effect is especially pronounced in dry years (those that fall below 75% of the distribution values). (3) In the wettest 2% of the years, runoff is expected to be higher than historical. However, the rest of the future enseble distributions portend drier than historical runoff conditions. Thus, the coing decades are likely to usher in ore severe floods and droughts than those experienced in the historical past. (4) The previous results and conclusions are typical of all watersheds. However, they are based on frequency coparison with all data. To exaine the potential changes on a onthly basis, box plots of the historical and future scenarios were developed for each onth of the year, watershed, cliate scenario type (A1B or A2), and hydrologic process (precipitation, PET, soil oisture storage, and runoff). Figure 5 shows the plots for Buford under A1B scenarios: the historical box-plots are denoted H1 through H12 while next to the are the future scenario box-plots denoted F1 through F12. The future box-plots include data fro all 13 future scenarios, while the historical box-plots include only historical data. This figure indeed shows that cliate change ipacts are not unifor across the onths of the year.

4 CONCLUSIONS This paper describes an integrated cliate assessent for river basins in Georgia. The study cobines (1) downscaling and assessent of future precipitation and teperature scenarios for six ACF sub-watersheds, (2) hydrologic assessents for each sub-watershed, and (3) water resources assessents for the entire basin (Yao and Georgakakos, sae issue). The cliate and hydrology changes are assessed based on the integrated assessent fraework. ACKNOWLEDGMENTS Figure 4. Frequency Curves for Buford Watershed under A1B Scenario H1 F1 H2 F2 H3 F3 H4 F4 H5 F5 H6 F6 H7 F7 H8 F8 H9 F9 H1F1H11F11H12F12 Blox Plots for Precipitation (Historical vs. Future) This research was funded in part by the National Oceanic and Atospheric Adinistration, Office of Global Progras, Cliate Prediction Progra for the Aericas (CPPA; NOAA contract No. NA6OAR43173), the Georgia Departent of Natural Resources, Environental Protection Division (Georgia EPD contract No ), and the State Water Institute Progra adinistered by USGS. The authors are grateful to all sponsors. The ideas and opinions presented herein do not necessarily reflect those of the sponsoring organizations. We would like to thank the two anonyous reviewers for their constructive coents on the work..25 Blox Plots for PET (Historical vs. Future) LITERATURE CITED H1 F1 H2 F2 H3 F3 H4 F4 H5 F5 H6 F6 H7 F7 H8 F8 H9 F9 H1 F1 H11 F11 H12 F12 Blox Plots for Soil Moisture (Historical vs. Future) H1 F1 H2 F2 H3 F3 H4 F4 H5 F5 H6 F6 H7 F7 H8 F8 H9 F9 H1 F1 H11 F11 H12 F12 Blox Plots for Runoff (Historical vs. Future) H1 F1 H2 F2 H3 F3 H4 F4 H5 F5 H6 F6 H7 F7 H8 F8 H9 F9 H1 F1 H11 F11 H12 F12 Figure 5. ly Historical vs. Future (A1B) Watershed Response, Buford Conway, D., (1998) Recent Cliate Variability and Future Cliate Change Scenarios for Great Britain, Progress in Physical Geography, 22(3), Georgakakos, A. P., Yao, H., Mullusky M. G., and Georgakakos, K. P., (1998). "Ipacts of Cliate Variability on the Operational Forecast and Manageent of the Upper Des Moines River Basin." Water Resources Research 34.4: Georgakakos, A., F. Zhang, and H. Yao, 21: Cliate Variability and Change Assessent for the ACF River Basin, Southeast US. Georgia Water Resources Institute (GWRI) Technical Report No. GWRI/21-TR1, Georgia Institute of Technology, Atlanta, Georgia, USA, 321 pp. Lettenaier, D. P., and D. Rind, (1992). "Hydrological Aspects of Global Cliate Change - Preface." Journal of Geophysical Research-Atospheres 97.D3: Lu, J., G. Sun, S. McNulty, and D. M. Aatya, 25. A Coparison of Six Potential Evapotranspiration Methods for Regional Use in the Southeastern United States. Journal of the Aerican Water Resources Association (JAWRA) 41 (3): Mitchell, T. D., and M. Hule, (1999). "Predicting Regional Cliate Change: Living with Uncertainty." Progress in Physical Geography 23.1 (1999):

5 Mujudar, P. P., and S. Ghosh, (28). "Modeling Gc and Scenario Uncertainty Using a Possibilistic Approach: Application to the Mahanadi River, India." Water Resources Research Sta, J. F., E. F. Wood, and D. P. Lettenaier, (1994). "Sensitivity of a Gc Siulation of Global Cliate to the Representation of Land-Surface Hydrology." Journal of Cliate 7.8 (1994): Tebaldi, C., Sith, Richard L, Nychka, Doug, and Mearns, Linda O., (25). "Quantifying Uncertainty in Projections of Regional Cliate Change: A Bayesian Approach to the Analysis of Multiodel Ensebles." Journal of Cliate 18.1: Wood, A. W., et al., (24) Hydrologic Iplications of Dynaical and Statistical Approaches to Downscaling Cliate Model Outputs, Cliatic Change 62(1-3), Zhang, F., and A. Georgakakos, (211) Joint Variable Spatial Downscaling, Cliatic Change, in subission