Pecos Basin Study Technical Service Center Support 1 Mission Statements The mission of the Department of the Interior is to protect and provide access to our Nation’s natural and cultural heritage and honor our trust responsibilities to Indian Tribes and our commitments to island communities. The mission of the Bureau of Reclamation is to manage, develop, and protect water and related resources in an environmentally and economically sound manner in the interest of the American public. 2 Pecos Basin Study Technical Service Center Support 3 Introduction This document reports work done by the Technical Service Center (TSC) to support the Pecos Basin Study (PBS), a component of the WaterSMART program. The objective of the TSC support was to provide assistance to the Albuquerque Area Office in development of climate and streamflow data used in the PBS. All data are intended for use in the Pecos Basin RiverWare Model (PBRM), Work on the data development began in April, 2013 and was completed in August, 2013. Tasks performed by TSC were: 1. Development of climate data for Maurer1 observed (gridded historical climate data developed from gage record) and 112 Bias Corrected and Spatially Downscaled (BCSD) climate change projections by PBRM nodes for computation of evapotranspiration, evaporation, and streamflows. 2. Development of VIC routing model for Maurer observed flows for PBRM nodes. 3. Development of VIC routing model for BCSD flows for PBRM nodes. 4. Computation of bias corrected monthly VIC flows for Maurer observed and 112 BCSD climate change projections for PBRM nodes. 5. Development of HDe climate data for PBRM nodes. 6. Development of HDe daily flows for PBRM nodes. 7. Development of HDe daily depletions for PBRM nodes. 8. Development of HDe daily evaporation for PBRM nodes. All Hybrid Delta Ensemble (HDe) data development used a spatial average dataset for the basin so that HDe quadrants by future were common to all HDe datasets. 1 Maurer, E. P., A. W. Wood, J. C. Adam, D. P Lettenmaier and B. Nijssen, 2002, a Long-Term HydrologicallyBased Data Set of Land Surface Fluxes and States for the Conterminous United States, Journal of Climate, 15(22) 3237-3251. 4 Data Types Three categories of data were developed for the PBS – Meteorological (Met), flow, and irrigation requirements. Developed data consisted of two types--transient Maurer observed 1950 through 1999 and 112 BCSD traces from 1950 through 2099 and Maurer observed and HDe traces for a simulation period from 1950 through 1999. The PBRM requires daily flow, evaporation and irrigation requirements. Daily meteorological data are an output of the HDe meteorological application. Daily flow require a two-step process whereby monthly HDe flows are developed from monthly bias corrected Maurer Observed and BCSD flow, then disaggregated to daily. Monthly bias corrected Variable Infiltration Capacity (VIC) flows were computed as documented elsewhere. Daily flows were computed by disaggregation of monthly bias corrected flows. Daily irrigation requirements were computed for use with PBRM using an evapotranspiration model. The same model was used to compute reference evapotranspiration that was used to compute reservoir evaporation using the historic monthly relation of reservoir evaporation to reference evapotranspiration. Monthly precipitation and monthly average precipitation are available for the Maurer observed and BCSD projections. Daily precipitation, maximum temperature, and minimum temperature are available for the Maurer observed trace. Maurer observed and BCSD projections data are available from: http://gdo-dcp.ucllnl.org/downscaled_cmip3_projections/dcpInterface.html HDe is a method to create representative subsets of the full suite of BCSD traces, with the purpose of reducing the computational intensity of the operational modeling that is based on the projections. HDe analyses are performed plotting the average difference in temperature and precipitation between the projected future period and a representative past period. As shown on Figure 1, these results are then divided into four quadrants and five clusters are created from the four quadrants as well as the “central tendency”, the area around the center of the plot in which significant changes in either temperature or precipitation do not exist. The HDe subsets for a quadrant are developed using all data in a given quadrant. HDe time series are computed using data of a BCSD cell specific to the target node. HDe quadrants can be defined using either the target node’s BCSD data or regional BCSD data. For the PBS, the latter approach was taken using the spatial average option of the BCSD web site as shown on Figure 2. For this project, the HDe analyses were performed for three thirty-year future periods which are outlined in Table 1. The projections for each quadrant (an ensemble) are processed by user-specified time periods by month to develop adjustment factors that are applied to Maurer observed trace to obtain a given HDe trace. The adjustment factors are applied as: HDe temperature = Maurer observed temperature + adjustment value HDe precipitation = Maurer observed precipitation * adjustment factor The projection subsets used in HDe analysis are listed in Table 2 and include the five categories of projections used in HEe analyses as well as the Maurer observed dataset. HDe traces are monthly average temperature and monthly precipitation, consistent with the BCSD traces. The PBRM requires monthly maximum and minimum temperatures. These data were aggregated from the daily HDe traces. HDe daily traces are computed by applying the monthly HDe adjustment factors for precipitation and values for temperatures to the Maurer observed daily trace. 5 Figure 1. Example HDe quadrants. HDe quadrant determination can use either the site’s precipitation and average temperature or an alternative precipitation and average temperature that is intended to be a spatial average of the basin. This is the approach used by the PBS. The spatially averaged Met data set was downloaded from the BCSD web site. A simplified method was used to develop bias-corrected HDe flow projections for operations modeling in URGSOM and PBRM. Rather than develop HDe climate data as input to the VIC models, which would have required considerable computer as well as human resources, HDe monthly flow traces were computed as: HDe flow = Bias corrected Maurer observed flows * adjustment factor where the adjustment factors were computed from the bias corrected flows using the meteorological quadrants. 6 Figure 2, PBS Spatial Average Specifications. 7 Table 1. HDe Setup Specifications. Time Frame Specifications Observed Historic Time Frame Output Simulation Period Start 1950 1950 End 1999 1999 Years 50 50 1950 2099 150 BCSD Period Historic Base Time Frame Future 1 Base Time Frame 1950 2010 1999 2039 50 30 Future Range Name 2020 Future 2 Base Time Frame 2040 2069 30 2050 FALSE Future 3 Base Time Frame 2070 2099 30 2080 TRUE Used FALSE Table 2 HDe Climate Change Scenarios. ID S1 S2 S3 S4 Scenario Observed Data Lower Temperature Lower Precipitation Lower Temperature Higher Precipitation Higher Temperature Lower Precipitation Higher Temperature Higher Precipitation S5 Central S0 VIC Routing Models Unrouted VIC flows exist for the Maurer observed trace and all 112 BCSD traces. PBRM and PBRM require flows to be routed to the headwater inflows points for these operational models. Two routing models and dataset were created for this purpose: 1. PBRM Maurer observed 2. PBRM BCSD PBRM flows nodes are listed in Tables 3. 8 Table 3. VIC Model Nodes. VIC Node PNSRL ECASR ECTSR PNPDL PNACM RHNR PNLA CCNLA PNART RPDAY PNLKD FMLKD NSLKD SSLKD RACAR PDCAR DCCAR BAMAL PAPCC PARED Description PECOS RIVER ABOVE SANTA ROSA LAKE LOS ESTEROS CREEK ABOVE SANTA ROSA LAKE LOS ESTEROS CR TRIB ABOVE SANTA ROSA LAKE PECOS RIVER NEAR PUERTO DE LUNA PECOS RIVER NEAR ACME RIO HONDO NEAR ROSWELL PECOS RIVER NEAR LAKE ARTHUR COTTONWOOD CREEK NEAR LAKE ARTHUR PECOS RIVER NEAR ARTESIA RIO PENASCO AT DAYTON PECOS RIVER (KAISER CHANNEL) NEAR LAKEWOOD FOURMILE DRAW NR LAKEWOOD NORTH SEVEN RIVERS NR LAKEWOOD SOUTH SEVEN RIVERS NR LAKEWOOD ROCKY ARROYO AT HWY BRD NR CARLSBAD PECOS R AT DAMSITE 3 NR CARLSBAD DARK CANYON AT CARLSBAD BLACK RIVER AT MALAGA PECOS RIVER AT PIERCE CANYON CROSSING PECOS RIVER AT RED BLUFF Bias Corrections Methods and Analysis Simulated flows from the VIC routing models are not usable directly in decision models because the flows need to be scaled to historic flows. Simulated flows for the Upper Pecos basin were adjusted using a quantile-mapping bias correction methodology for the 20 PBRM VIC nodes shown in Table 3 using historic flows from the PBRM model. For this analysis, biases are characterized using a quantile map. The quantile map features two empirical cumulative distribution functions (CDFs)-one of simulated flows during the representative historic period, referred to as the “bias identification period” (1950-1999) and another of the reference observed flows during this period. The CDFs are constructed at a given runoff location, first on a month-specific basis to characterize bias in monthly mean flows, then on an annual basis to characterize bias in annual mean flows and to preserve mass balance. All CDFs are smoothed non-parametrically using a sliding kernel approach. Figure 3 is a plot of an example CDF. After defining these maps, simulated runoff bias correction ensues. The quantile maps are interpreted to reveal VIC runoff simulation bias for a given simulated runoff magnitude. Refer to the “West-Wide Climate Risk Assessment” report for more details on the methodology. 9 CDF of Simulated Data Lobatos, Run 6 May 20,000 18,000 1950-1999 16,000 2000-2099 Flow (cfs) 14,000 12,000 10,000 8,000 6,000 4,000 2,000 0 0 0.2 0.4 0.6 0.8 1 Cumulative Probability Figure 3. Example Cumulative Density Function. Figures 4 through 7 show example bias correction outputs for two example basins with low and high biases. Low bias stations have simulation outputs that match both the annual volume and seasonal timing well. High bias stations fail to capture either the seasonal flow patterns or the annual runoff volume. Figures 4 and 6 show monthly observed (black line), simulated (red line), and bias-corrected simulated (cyan line) flow volumes for water years 1950–1999 for the Rio Blanco below Blanco site. Figure 5 shows monthly mean volume (left panel) and annual mean volumes (right panel) including biascorrected simulated flow (BCF) calculated from water years 1950–1999 for the Rio Blanco below Blanco site. Figure 7 shows monthly mean volume (left panel) and annual mean volumes (right panel) including bias-corrected simulated flow (BCF) calculated from water years 1950–1999 for the Galisteo Creek below Galisteo Dam site. In the Rio Grande basin, increased variability of bias corrected flows after year 2000 spurred a review of bias corrected flows and procedures. The review confirmed that the bias corrected flows were computed correctly. In the future period (2000-2099) however, simulated values can exist that are larger (smaller) than the maximum (minimum) historical simulated values. This point is illustrated on Figure 3 which plots CDF’s of flow for the historic and future simulated time periods, From the two CDFs we can see that, in the future simulation, about a 20% chance of having a flow that exceeds the maximum historical simulated flow (~10,000 cfs) exists. In these instances, the simulated flow cannot be directly mapped to a historical observed flow because it is outside the upper end of the historical simulated CDF. Therefore, the simulated flow is scaled by the ratio of the maximum observed flow to the maximum historical simulated flow over the same (19501-99) period. 10 20 40 0 TAF Monthly Volumes, WY 1950-1999 (n=43) 1950 1952 1954 1956 1958 1960 20 40 0 TAF Water Year 1962 1964 1966 1968 1970 20 40 0 TAF Water Year 1972 1974 1976 1978 1980 1982 20 40 0 TAF Water Year 1982 1984 1986 1988 OBS Water Year SIM 1990 1992 BCF Figure 4. Historical Simulated Runoff, Small-Bias Example: Monthly Time Series Before and After Bias Correction. 11 Annual Mean Volumes Water Years 1950-1999 80 OBS SIM BCF 60 20 40 TAF 10 0 5 TAF 15 20 Mean Monthly Volumes Water Years 1950-1999 Figure 5. Historical Simulated Runoff, Small-Bias Example: Monthly and Annual Means Before and After Bias Correction. 12 8 12 0 4 TAF Monthly Volumes, WY 1950-1999 (n=43) 1952 1954 1956 1958 1960 8 12 Water Year 0 4 TAF 1950 1964 1966 1968 1970 8 12 Water Year 0 4 TAF 1962 1974 1976 1978 1980 1982 8 12 Water Year 0 4 TAF 1972 1982 1984 1986 1988 OBS Water Year SIM 1990 1992 BCF Figure 6. Historical Simulated Runoff, Large-Bias Example: Monthly Time Series Before and After Bias Correction. 13 Annual Mean Volumes Water Years 1950-1999 4 Mean Monthly Volumes Water Years 1950-1999 0 0 10 20 TAF 2 1 TAF 3 30 OBS SIM BCF Figure 7. Historical Simulated Runoff, Large-Bias Example: Monthly and Annual Means Before and After Bias Correction. 14 Figure 8 shows the simulated and bias corrected flow for the May data for the sixth run at the Lobatos station. From the simulated data we find 21 instances (out of 100 May flow values) where there post 2000 simulated flow data exceeds the historical maximum simulated flow (~10,000 cfs). In the bias corrected data these are the only instances where the bias corrected flow exceeds the historical simulated maximum. Figure 8. Comparison of Various Projections. (1950-1999) all runs for all stations will result in flows that are at or below historic levels. However, post2000 (2000-2099) simulated flows that are greater than the historical simulated flows can result in flows that are greater than the historical observed maximum (or less than the historical observed minimum). This is expected and is done in an effort to maintain the increased variance and extreme values that are simulated with the climate projections. Extrapolating this discussion to the entire dataset, we find e that in the time period before 2000 15 PBRM Data HDe daily Met data traces were developed to support computation of evapotranspiration using the Penman-Monteith method for use in the PBRM model. The process was similar to PBRM climate data development except that daily maximum temperature, minimum temperature, and precipitation are generated from the HDe data manager. HDe Irrigation requirements were computed using the ET model. HDe evaporation was computed as a function of reference ET using an application developed for the PBS. HDe flow traces for PBRM were generated using methods previously noted. Two additional circumstances were addressed in the development of PBRM HDe daily flows. First, PBRM performs its operational calculations based on a mix of upstream total flows and downstream local inflows (aka gains or accretions). Bias corrections are performed on total flows. Therefore, total flows were computed from the upstream total flows and downstream local flows. Conversely, the computations were reversed to spatially disaggregate HDe flows before they were temporally disaggregated for use in PBRM. To support the temporal disaggregations, daily flows were computed by computing disaggregation fractions for all flows (total and local inflows) using the historic dataset. In principal, the disaggregation factors are applied to the Maurer observed and HDe monthly and HDe traces to obtain flows (temporal disaggregation). However, because HDe adjustments are monthly and the adjustment vary month to month, disaggregated daily flows can produce transition issues between months. A smoothing process was developed consisting of following steps: 1. Disaggregate Maurer observed trace 2. Apply interpolated monthly HDe adjustment factors to compute initial daily HDe flows. 3. Adjust initial daily HDe flows to maintain consistency with monthly HDe flow. The Pecos Basin is really dry. The median flow of most tributaries and local inflows was zero. These data were problematic for doing bias corrections. Therefore another approach was developed to compute HDe flows for these cases. This approach is to compute the HDe flow as a function of the historic flow relative to the next downstream station’s historic and HDe flow. For instance: Tributary HDe flow = Downstream HDe flow * Tributary historic flow / Downstream historic flow. This was done on a monthly basis. Daily disaggregations for tributaries were done the same as mainstem flow nodes. 16 Figure 9. Hydrology Data Pathways. 17 RiverWare Model Modifications Although not part of TSC work scope, sufficient resources were available to create a version of the PBRM and rules, data management interfaces (DMI’s), associated workbooks, and a data manager that could be used to support the PBS. The workbooks and data manager were adopted from the Milk Basin Study. The following modifications were made to the model and rules: 1. Added data object slots and initialization rules for all slots of previous initialization DMI. This was necessary to accommodate difference between historic (1940 through 2009) and climate change (1951 through 1998) periods. This also removed the need for an initialization DMI. 2. Modified return flow lagging initialization from routed returns after the starting timestep to unrouted flows before the starting timestep, then created an initialization rule for original return flow lagging initialization. This corrected an error in the original setup which produced incorrect routed return flows during first year and warning messages. 3. Changed CID water user object and supply rule to enable use of evapotranspiration as an input. This consisted of setting up the water user object to compute depletion requested from evapotranspiration rate and area and to compute diversion requested from depletion requested and efficiency. The water supply rule was changed to populate the “Incoming Available Water” slot after unlinking it from Avalon Reservoir. 4. Changed UpperFSID and LowerFSID supply rules to populate the “Incoming Available Water” slots instead of the “Diversion Requested” slots because the computed “Diversion Requested” caused a conflict with the original rule setup under some climate change scenarios. 5. Created six DMI’s to import climate change or historic irrigation requirements, evaporation, and hydrology. A separate DMI exist for each parameter and data source to facilitate data management by the manager. 6. Created DMI “Standard Output” to post a standard set up of output for each model run. 7. Created a manager (workbook PecosRiverWareManager.xlsm discussed in next section) that sets up batch operation of RiverWare model using historic data, Maurer observed or one HDe trace, or all HDe traces. 18 Applications Data for the PBS were developed using several applications. The applications can be categorized as generalized and specialized where the generalized applications were developed to support similar studies west-wide and the specialized applications were developed specifically for the Pecos Basin. The generalized applications consist of VIC, the Penman Monteith model and data manager written using Excel Visual Basic For Applications (VBA). The specialized applications consist of scripts and VBA data managers. The Hargreaves Samani and Penman Monteith models and documentation are available from: ftp://ftp.usbr.gov/tsc/jrieker/utilities/HargraevesSamani/hs4.0Install.zip ftp://ftp.usbr.gov/tsc/jrieker/utilities/PenmanMonteith/pm4.0Install.zip The generalized data managers used to support the PBS include NetCDFBCSDManager.xlsm, ClimateChangeAdjustmentsManager.xlsm (HDe Met data manager), ClimateChangeHDeFlowsManager.xlsm, HargraevesSamaniManager.xlsm, and PenmanMonteithManager.xlsm. Pecos versions of each of these were created. The source applications and documentation (ClimateChangeDataManagers.docx) are available at: ftp://ftp.usbr.gov/tsc/jrieker/utilities/climatechange/ Specialized applications are PecosRiverWareManager.xlsm, PecosCCHydrologyManager.xlsm, PecosEvapHSRefETManager.xlsm and PecosEvapHSEvaorationManager.xlsm. The PecosRiverWAreManager.xlsm is used to manage RiverWare historic and climate change data and model runs. The Pecos CCHydrologyManager.xlsm is used to compute HDe daily flows from HDe monthly flows. The Pecos EvapHSRefETManager.xlsm is used to compute the historic reference ET that was used to develop evaporation to reference ET regressions. The PeocsEvapHSEvaporationManager.xlsm is used to compute evaporation as a function of reference ET for Maurer observed and HDe traces. Instructions for these applications exist within the applications. The applications are available upon request. List of Acronyms BCSD – Bias Corrected Spatially Downscaled. CC – Climate Change. CDF - Cumulative density function. DMI - Data management interface. ET – Evapotranspiration. HS – Hargraeves-Samani ET Model. HDe – Hybrid Delta Ensemble. MRGCD – Middle Pecos Conservancy District. PM – Penman Monteith ET Model. PBS - Pecos Impacts Assessment. PBRM – Upper Pecos Simulation Model. PBRM – Upper Pecos Water Operation Model. RiverWare – River and reservoir modeling system used to develop operation models. VIC - Variable Infiltration Capacity. WaterSmart – Sustain and Manage America’s Resources For Tomorrow. WWCRA – West-wide Climate Risk Assessment. 19 20