Southwest Climate Science Center FINAL TECHNICAL REPORT

Transcription

1 Southwest Climate Science Center FINAL TECHNICAL REPORT 1. USGS GRANT/COOP AGREEMENT G14AP PROJECT TITLE: Disentangling the Influence of Antecedent Temperature and Soil Moisture on Colorado River Water Resources 3. PRINCIPAL INVESTIGATOR 4. CO-PRINCIPAL INVESTIGATOR PERSONNEL a. Principal Investigator: Connie Woodhouse (University of Arizona) b. Co-Investigators: Greg Pederson (USGS-Bozeman), Adam Csank (University of Nevada-Reno) c. Other scientific team members: Stephanie McAfee (University of Nevada, Reno), Greg McCabe (USGS-Denver), Steve Gray (USGS-Anchorage); Becky Brice (graduate student,university of Arizona) 6. PROJECT START DATE: 09/01/ EXPECTED COMPLETION DATE: 02/28/ PURPOSE AND OBJECTIVES: A major challenge to water resource management in the southwestern US is anticipating and planning for the effects of climate change on water supplies. Changes in precipitation have obvious and direct effects, but warming temperatures can also impact water supplies in a variety of ways; by influencing snowpack amount and snowmelt timing, evapotranspiration and soil moisture, exacerbating the effects of drought, and increasing demand for water. In this project, a number of water managers and water users in the Southwest have informed the overarching goal of this project, namely gaining a better understanding of how a set of hydroclimatic factors contributes to low flows on the Colorado River. In turn, water managers and users will use this information to guide expectations for climate change impacts on future flows. It is clear that reduced snowpack is a leading cause of low flows, but recent observations and research have indicated that other factors, including antecedent soil moisture conditions and temperature effects on snowpack may play an important role in exacerbating or mitigating the impacts of low snowpack on total water year streamflow. Guided by input from the water resource community, we initially focused on three hydroclimatic factors: prior summer/fall soil moisture, cool season precipitation/snowpack, and late winter/spring temperatures. We focus on these factors because 1) prior work suggested they are key controls on runoff, and 2) they are variables for which reconstructions of past conditions from tree rings are possible. Extended hydroclimatic records from tree rings are critical because the period of instrumental data is too short to assess the full range of natural variability that is possible. Our project had three main lines of stakeholder-driven inquiry: 1. Given instrumental records for the 20th and 21st centuries, how have the contributions of antecedent soil moisture, winter/spring temperatures, and total cool season precipitation varied during the major periods of low flow in the upper Colorado River basin (UCRB)? 2. Using tree-ring reconstructions of pre-1900 antecedent soil moisture, temperatures, and precipitation, are contributions of these three factors to low flows over past centuries similar to those of the 20th and 21st centuries? Are there differences in these contributions between cooler and warmer time periods? Have contributions changed over time? 1

2 3. What is the sensitivity of UCRB low flows to different scenarios of changes in antecedent soil moisture, winter precipitation, and winter/spring temperatures projected by CMIP5 models? How do results relate to conditions during significant low flow events captured in the instrumental and paleo record? Over the course of the project, due to initial results and based on feedback from our water manager partners, several revisions were made. Regarding the first objective, we modified this line of inquiry to not just focus on periods of drought, but more generally to address the influence of antecedent soil moisture, temperatures, and total cool season precipitation on flow over the period for of the gage record (since 1906). In addition, we found that spring/early summer temperatures (not winter/spring, as we had initially thought) were the most important temperature season for streamflow, and that fall soil moisture played a somewhat unclear role. Although it appears that antecedent moisture is important, we were not able to identify a hydroclimatic variable that adequately represents this (at least not yet!). We also did these analyses for the three main sub-basins, but they ultimately seemed to be of less interest to our management partners in the end. For the second objective, we used reconstructed streamflow and cool season precipitation to infer the times during which temperature likely played an important role, either enhancing or diminishing flows, beyond that expected given the cool season precipitation. While a preliminary reconstruction of spring temperature from tree-ring widths and oxygen isotopes has been generated, the final reconstruction is still in progress. Currently we have used a network of tree-ring width and maximum density records specific to the UCRB to produce a runoff season (May-July) temperature reconstruction for comparison against the hydrologically inferred temperature influence on streamflow over the time period (see Woodhouse and Pederson, in review). The hydrologically inferred temperature history and the reconstructed temperature history compare well, and we expect with the forthcoming addition of the isotopic data into the reconstruction the relationships will improve substantially. To that end we have already had some success using the isotopic data in combination with tree-ring width and maximum density records from the UCRB to produce a reconstruction of April to July temperatures for the 1700 to 2000 time period and are waiting on the last set of isotopic data to extend the record back to 1500 before we submit the paper (Csank et al. in prep). The third objective was modified based on feedback from water manager partners to address the major sources of uncertainty on the estimated range of future water year flows. Manager expressed a strong interest in taking a bottom-up approach, wherein we evaluated the range of climate changes that could cause flow conditions that are of concern to managers. At the stakeholder meeting in 2015, water managers identified two flow conditions that would be difficult to manage: (1) periods of 7 to 10 year with no above normal flows, and (2) shorter periods where flow is substantially below normal (e.g., 75% of normal). Identifying the broad range of conditions under which these conditions could occur required compiling a very large ensemble of plausible flows, something that was not tractable using downscaled CMIP5 output. Instead, we employed a synthetic time series/statistical modeling approach that allowed us to evaluate a very large ensemble of future flow traces for concerning flow conditions and assess their drivers. The work was presented to water managers in the Upper Colorado River Basin via webinar on October 14, 2016 and at a meeting in Phoenix on November 9, Results were also presented in a poster at the American Geophysical Union meeting in San Francisco during December We currently have a manuscript in preparation (McAfee et al. in prep). 2

3 9. ORGANIZATION AND APPROACH: Research was organized base on four main questions, with the approaches as follows: 1. Can we identify variable contributions of the three hydroclimatic factors discussed above to low flows in the gage record, and have these contributions varied over time? We used correlation analysis and multiple linear regression approaches to identify the precipitation, temperature variables that best explained water year streamflow, as well as to examine the importance of antecedent soil moisture. We also identified key low flow periods from the Lees Ferry natural flow record to see how climatic conditions varied over the different droughts. We used time series graphics and analyses to evaluate changes in the importance of temperature and precipitation over time. 2. Can we develop skillful and management-relevant reconstructions of antecedent summer/fall soil moisture, and winter/spring temperatures from tree-ring data? a. Soil moisture reconstruction approach/tasks: We used gridded soil moisture values from the McCabe and Wolock (2011) water balance model for the Upper Colorado River Basin as a surrogate for the lack of in-situ measurements of soil moisture. Correlation analysis was used to determine antecedent months when soil moisture and Colorado River streamflow are highly correlated. November was selected as the representative antecedent month according to correlation statistics and the theoretical physical mechanisms during winter in the basin. Only grid cells with significant and positive correlations (p<0.01) in both split periods ( , ) were retained for regression analysis. We developed a final reconstruction of November soil moisture for the UCRB for the period using ring-width from a pool of 62 moisture-sensitive standard chronologies in and near the upper basin. Potential predictors were selected from this full pool based upon split-period positive and significant correlation (p<0.01) with November soil moisture. Stepwise regression and leave-one-out validation was used to develop a reconstruction model, with reconstruction skill (R 2 ad) = 0.492, and validation statistics root mean square error (RMSE v ) = and the reduction of error statistic = To understand the influence of soil moisture on streamflow variability, our goal was to produce a soil moisture reconstruction that could be compared with existing streamflow reconstructions. However, November soil moisture and streamflow reconstructions were derived from Upper Colorado River Basin chronologies with very similar climate information, leading to a statistical inflation of the relationship. The dependence of soil moisture on streamflow was removed using linear regression so that independent soil moisture could be evaluated. The soil moisture residuals from this regression, and the streamflow reconstruction series, were categorized by quintile and compared in two subperiods, and b. Temperature reconstruction approach/tasks: Our initial goal was to produce a temperature reconstruction that would reflect the full runoff season, March July. This proved to be challenging, as we expected, since tree-ring width and density typically reflect warm season conditions, and the tree-ring isotopes that we hoped would document winter temperatures turned out to be more sensitive to February-April temperatures. Our efforts continue, but we developed a preliminary reconstruction of May-July average temperature reconstruction for the UCRB using ring-width and maximum density data using the following methods. Potential predictor chronologies for use in the reconstruction were selected by acquiring all publicly available ring-width and maximum density records that fell within 100 km of the UCRB (n = 23) that exhibited a positive and significant (p 0.1) correlation with observed May-July average temperature. The temperature data 3

4 used for screening and calibrating the reconstruction are the PRISM 4km resolution grid points for the upper Colorado River basin above Lees Ferry. The PRISM temperature data were compared against climate division data for the UCRB to ensure integrity of 20 th century trends and variance. The final May-July average temperature reconstruction was produced by nesting successively longer, but statistically weaker, best-fit multiple linear regression models through time. Models were crossvalidated using both a leave-one-out and a k-folds leave-10-out method that was replicated 100 times. Reconstruction skill ranges from R 2 adj of to 0.217, with cross-validation statistics from the leaveone-out and leave-10-out method equivalent showing the root mean squared validation error (RMSE v ) ranging from to The nested reconstruction spans the years We used this reconstruction to 1) evaluate reconstruction improvements over the early spring (Feb-April) time interval from δ 18 O tree-ring isotopes, and 2) compare against a hydrologically inferred runoff-season temperature history (see below), Seeking to improve upon temperature relationships over the early spring (Feb-Apr) and during the runoff season (May-July) we next integrated newly developed δ 18 O tree-ring isotope data. The isotope data captures some cool-season temperature relationships because they reflect surface air temperatures during the precipitation event, hence their potential utility for improving early spring temperature reconstructions. To produce this reconstruction, we initially compared the isotopic records against temperatures during the cool season and found that the isotopes showed a significant correlation with spring (March-June) temperatures. Although correlated the R 2 for the isotope-records with temperature was too low for use to have confidence using the isotopic record alone to develop a reconstruction. Instead we incorporated the δ 18 O tree-ring isotope reconstructions into a best fit multiple linear regression model along with temperature sensitive tree-ring width and maximum latewood density data from the UCRB. By combining the two records we were able to develop a model of April-June temperatures. Models were cross-validated using a leave-one-out method that was replicated 100 times. Reconstruction skill ranges from R 2 adj of to 0.442, with our leave-one-out method showing a reduction of error (RE) and coefficient of efficiency of (RE = ; CE = ). The reconstruction so far spans the years but will be extended to 1500 in the near future, when we receive the final laboratory analyses for the isotopes. 3. What can be said about the relationship between temperature, precipitation and streamflow over past centuries? We used two approaches to address this question. 1) Because instrumental data analysis suggested the difference between cool season precipitation and water year streamflow could be used to infer to influence of temperature, we employed this approach using reconstructions. We used an existing reconstruction of water year streamflow for Lees Ferry and generated a new reconstruction of cool season precipitation for the UCRB (with independent tree-ring chronology predictors, using stepwise regression). We evaluated the role of temperature by analyzing the years with the largest difference between standardized flow and precipitation. Based on results from the instrumental period analysis, we used the difference of these two series to infer the influence of temperature on streamflow going back to We evaluated periods of drought and pluvials, based on the streamflow reconstruction, in terms of streamflow, precipitation deficits, and inferred temperature, then assessed the frequency of cool and warm pluvials and droughts over the past 4 centuries. 2) We also assessed the results from the first approach using a reconstruction of runoff season May-July temperature (above) developed for this project from tree-ring data, using simple graphical comparisons. 4. In simulations with the MWBM, what combinations of antecedent soil moisture, winter precipitation (snow and/or rainfall), and winter/spring temperatures induce flow levels at or below the extreme lows identified in the historical and paleo record, and where do these conditions fall within the range of CMIP5 projections? As noted above, the water managers were interested in using a bottom-up approach, which was not tractable with a research plan that involved feeding MWBM with a suite of downscaled 4

5 climate projections. Instead, we developed a set of 62,500 climate traces covering a range of winter and summer precipitation changes and summer temperature increases. To prepare these synthetic time series, we resampled time series of historical winter (October April) precipitation, summer (May September) precipitation and summer (May July) temperature anomalies produce 500 unique climate traces for each season. The means of each of these traces was varied such that temperatures could be 0 C, +1 C, +2 C, +3 C, or +4 C above the long-term average, and precipitation could be 80%, 90%, 100%, 110%, or 120% of the long-term average. All combinations were tested (e.g., +2 C, 80% of historical average winter precipitation, and 120% of historical average summer precipitation) to make 500 traces of each of 125 discrete climate change scenarios. These synthetic climate futures were fed into a statistical model of annual water-year flow, based on historical climate and naturalized Lees Ferry flow. We also tracked the number of low flow events (defined as 7+ years with no more than one year of above historical average flow) and very low flow events (defined as 4+ years with no more than one year with flow above 75% of the average flow). Simulated flow was evaluated to identify changes in the average flow for each of the 125 unique climate change scenarios, as well as changes in the occurrence of low and very low-flow periods that managers identified as being of concern. Given that there was substantial variability in mean flows, as well as in the frequency of low-flow events, we also investigated potential sources of that variability, since the mean climate changes were identical. The only two potential sources of variability in mean flow were in the (randomly chosen) flow in the first year of the modeled flow and the temporal characteristics of the climate time series, so we plotted and correlated flow characteristics (mean, frequency of low-flow events) against flow in the first year of series, and measures of autocorrelation in the input climate data series. 10. RESULTS: 1. Contributions of hydroclimatic factors to upper Colorado River streamflow results: a. Cool season precipitation accounts for most of the variability in UCRB flow, explaining 66% of the variability in water year streamflow between 1906 and 2012 and 70% since March-July average temperature explains about 6-8% of the variability in flow, and prior November soil moisture explains an additional 2-3%. b. In years with flows less or greater than expected given cool season precipitation (such as 2002 and 1984), temperature is a much greater influence, explaining about 40% of the variability. Spring moisture can be an important contributor to flow as. An example of this is 1957, a high flow year with an extremely wet May and June, cool spring, and moderately wet cool season. c. Major upper Colorado River droughts of the 20th and 21st centuries have occurred under various combinations of precipitation, temperature, and prior fall soil moisture. For example, The 1950s drought ( ) was the driest, in terms of precipitation (30th percentile), but the coolest (43rd percentile). The 2000s drought ( ) was the least dry (precipitation at the 48th percentile), but the warmest (79th percentile). The droughts ( ) was also only moderately dry but quite warm. The 1930s drought ( ) was dry (36th percentile for precipitation, 30th percentile for soil moisture, the driest soil of all drought periods) and warm (61st percentile), but not as warm as the most recent droughts. d. While cool season precipitation explains most of the variability in annual flows, temperature appears to be highly influential under certain conditions, with the role of antecedent fall soil moisture less clear. 2. Reconstructions of antecedent summer/fall soil moisture, and winter/spring temperatures from treering data a. Reconstruction of November soil moisture for the UCRB using ring-width tree-ring chronologies was produced and demonstrates some temporal relationship to low 5

6 streamflow in the UCRB, though the nature and strength of this relationship remains difficult to parametrize. The completed reconstruction spans a. Reconstructions of runoff season (May-July) average temperature for the UCRB were produced and show reasonably high fidelity against observational data using existing tree ring-with and maximum density records. The final reconstruction spans CE. b. Tree ring δ 18 O isotope records improved the fidelity of late spring (April-June) average temperature reconstructions, lending further insight into the role of temperature in controlling drought during the critical temperature sensitive snow accumulation months. The final reconstruction is nearly complete and will span the period Relationship between temperature, precipitation and streamflow over past centuries results: a. Runoff season temperature can be inferred from the cool season precipitation/annual streamflow difference, reconstructed from tree rings. This record, as a proxy for temperature, is supported by the May-July temperature (2b), but more importantly underpins the role of temperature in moderating or exacerbating low flow and high flow events over the past 500 years. b. Temperature has likely played a key role in conditioning the impacts of some droughts and pluvials on Colorado River flows over past centuries. On a relatively regular basis, persistent droughts contain years with runoff temperatures warm enough to further intensify low flows beyond what would be expected from precipitation deficits alone. c. While there are a much smaller number of years that suggest cool conditions may ameliorate drought impacts on streamflow, they do occur. Although not a characteristic of particular droughts, these years seem to occur most often during relatively cooler intervals of time, such as the 19 th century. d. Compared to prior centuries, the 20th century contains twice as many years with high flows that are still lower than might be expected from precipitation e. While the reconstructions of past hydroclimatic variability suggest the proportion of years in which temperature may be a more dominant factor has not changed appreciably over the past four centuries, they do indicate that some of the ways in which temperatures impact streamflow are shifting. Specifically, warming temperatures may be resulting in less efficient precipitation in terms of its contribution to streamflow, with increasing frequency. 4. Future scenarios results: a. Warmer and drier scenarios led to lower average flow and more frequent low-flow events (to absolutely no one s surprise), but we identified a surprisingly large range of future flow conditions associated with identical changes in mean climate. Within climate scenarios that had identical changes in summer temperature and summer and winter precipitation, mean flow varied by as much as 0.4 maf. b. Low- and very low-flow periods became much more frequent with decreases in winter and precipitation, but under winter precipitation increases, there was much greater variability in the frequency of these events. c. Mean flow and the frequency of low-flow (but not very low-low) events are highly correlated with flow in the first-year of the future flow trace. d. Variability in the frequency of low- and very low-flow events appears to be associated with the degree of serial correlation in precipitation, with less influence from persistence in temperature. 6

7 11. NEXT STEPS: The project s end date has passed and this is the final report for this project. Almost all the tasks outlined in the original proposal have been completed, or modified and completed based on stakeholder feedback. The one outstanding task is to complete the tree-ring isotope/ring-width temperature reconstruction, for which we are waiting the final isotopic analysis results. We anticipate the completion of this reconstruction in Spring 2017, and a publication to follow shortly after. In addition, we have two papers in revision (Woodhouse and Pederson, McCabe et al.) and two in preparation (McAfee et al., and Csank et al.). In all cases, fact sheets will be generated, in addition to the peerreviewed papers, for resource management audiences. Our final workshop with water managers resulted in a number of additional avenues of research that would be useful to resource management. Guided strongly by this and subsequent feedback, we are currently working on a proposal to continue this project s collaborative research to be submitted in response to the Southwest CSC RFP. 12. OUTPUTS: a. Peer-reviewed publications resulted from this project, including in preparation, in review, accepted, or published Brice, R. et al. In prep. Fall season soil moisture reconstruction from tree rings in the Upper Colorado River Basin, U.S.A. Journal to be determined Csank, A.Z., Pederson, G.T., Woodhouse, C.A. In prep. A 400-year reconstruction of spring temperatures in the upper Colorado River Basin using a multi-proxy approach. Journal to be determined. McAfee, S.A. et al. In prep. Initial conditions and internal variability are important drivers of diversity in Upper Colorado flow estimates. For Geophysical Research Letters McCabe, G.J. D. Wolock, G.T. Pederson, C.A. Woodhouse, S. McAfee. In revision. Contributions of Temperature and Precipitation to Trends and Variability in Upper Colorado River Flows. International Journal of Climatology. Woodhouse, C.A., G.T. Pederson, K. Morino, S.A. McAfee, G.J. McCabe Increasing influence of air temperature on upper Colorado River Streamflow. Geophysical Research Letters 43, doi: /2015gl Woodhouse, C.A. and G.T. Pederson. In revision. Inferring the influence of temperature on upper Colorado River streamflow over past centuries. Water Resources Research. b. Non-peer-reviewed publications Woodhouse, C.A., G.T. Pederson, K. Morino, S.A. McAfee, G.J. McCabe Increasing influence of air temperature on upper Colorado River Streamflow. 2.amazonaws.com/assets/palladium/production/s3fs-public/atoms/files/UCRBStudyFactSheet.pdf 7

8 c. Conference talks based on this project Warm and Cool Droughts: The Influence of Temperature on Colorado River Flow. GC 13E American Geophysical Union, San Francisco, CA, Dec , A 400-year reconstruction of spring temperatures for the upper Colorado River Basin. GC 13E American Geophysical Union, San Francisco, CA, Dec , Ensembles of 21st Century Colorado River Flow Projections Exhibit Substantial Diversity in Response to Seasonal Hydroclimatic Scenarios. GC 13E American Geophysical Union, San Francisco, CA, Dec , Fall season soil moisture reconstruction from tree rings in the Upper Colorado River Basin, U.S.A. American Quaternary Association Annual Meeting. Santa Fe, New Mexico, July Changing Station Coverage in the Upper Colorado River Basin: Is This a Problem? American Geophysical Union, San Francisco, CA, Dec , Use-Inspired Hydroclimatic Research in the Upper Colorado River Basin. Biennial Conference of Science & Management on the Colorado Plateau & Southwest Region, Flagstaff, AZ, October 6, Antecedent soil moisture reconstruction from tree rings in the Colorado River Basin. Association of American Geographers Meeting, Chicago, April 21-25, 2015 Using tree-ring isotopes to understand hydroclimate variability in the Upper Colorado River Basin. Pacific Climate (PACLIM) Meeting, Pacific Grove, CA, March 8-11, Investigating the role of temperature in mediating relationships between cool season precipitation and water year streamflow in the Upper Colorado River basin. Pacific Climate (PACLIM) Meeting, Pacific Grove, CA, March 8-11, Evaluating the Influence of Precipitation, Temperature, and Soil Moisture on Upper Colorado River Basin Streamflow and Drought. H31M-07. American Geophysical Union, San Francisco, CA, Dec , Collaborative Research on Upper Colorado River Basin Streamflow and Drought. Association of Pacific Coast Geographers meeting, Tucson, AZ. Sept , d. Data outputs, maps, decision-support or other informational tools developed as part of this project: The project web site, Drivers of Drought in the Upper Colorado River Basin, provides access to the articles generated from the project and related supporting data, as well as presentations from both professional and stakeholder meetings. 8

9 13. OUTREACH AND ENGAGEMENT:. e. Presentations, seminars, webinars, or workshops made to stakeholders, the public at large, or any other group outside the research community. Stakeholder Advisor board, Denver Federal Center, Lakewood, CO, September 1, 2015 (9 water managers from 6 agencies, a total of 16 participants) Full science team except Gray. Stakeholder Advisor board plus others, Salt River Project, Tempe, AZ, November 9, 2016 (16 water managers from 8 agencies; a total of 21 in-person and 4 remote participants) Full science team. Presentations at meetings with water manager partners, including two workshops with representatives from all our management partner agencies: Brice, R. Reconstructing antecedent soil moisture using tree rings. Water Manager Advisory Board workshop, Denver Federal Center, Denver, CO, Sept. 17, Csank, A.Z. Can a multi-proxy approach to tree ring data be used to reconstruct temperature? Water Manager Advisory Board workshop, Denver Federal Center, Denver, CO, Sept. 17, Csank, A.Z. and G.T. Pederson. Using tree-ring widths, maximum density, and isotopes to reconstruct temperature. Final Project Workshop, Salt River Project, Tempe, AZ, November 9, McAfee, S.A. Three windows on the future. Water Manager Advisory Board workshop, Denver Federal Center, Denver, CO, Sept. 17, McAfee, S.A. Future flows? Preliminary data from a synthetic flow experiment. WEBINAR Oct McAfee, S.A. Future flows. Final Project Workshop, Salt River Project, Tempe, AZ, November 9, McCabe, G. Relative Contributions of Temperature and Precipitation to Upper Colorado River Flow During the Past Century. Final Project Workshop, Salt River Project, Tempe, AZ, November 9, Pederson, G. T. Recent Temperature Bias in Gridded Datasets: minor or MAJOR issue in models? Water Manager Advisory Board workshop, Denver Federal Center, Denver, CO, Sept. 17, Woodhouse, C.A., The Role of Temperature (and soil moisture) in Mediating Relationships between Cool Season Precipitation and Water Year Streamflow in the UCRB. Denver Water Board, Denver, CO, July 21, Woodhouse, C.A., The Role of Temperature (and soil moisture) in Mediating Relationships between Cool Season Precipitation and Water Year Streamflow in the UCRB. Colorado River District brown bag talk, Glenwood Springs, CO, July 23, Woodhouse, C.A. Disentangling the Influence of Temperature and Antecedent Soil Moisture on Colorado River Water Resources. Water Manager Advisory Board workshop, Denver Federal Center, Denver, CO, Sept. 17,

10 Woodhouse, C.A., G. Pederson, K. Morino, and G. McCabe. The Role of Temperature in Mediating Relationships between Cool Season Precipitation and Water Year Streamflow in the UCRB. Bureau of Reclamation Colorado River Hydrology Work Group Meeting, Tucson AZ, May 27, Woodhouse, C.A. The influence of climate on Lees Ferry water year flow and, Streamflow minus Precipitation: A Filtered Record of Past Temperature. Final Project Workshop, Salt River Project, Tempe, AZ, November 9, f. Communications with decision-makers, including their name and agency and the date(s) and frequency of communications. Our water management partners initially included: Jim Prairie (Bureau of Reclamation, Lower Colorado District); Eric Kuhn and Dave Kanzer (Colorado River District); Laurna Kaatz and Steve Schmitzer (Denver Water); Paul Miller (NOAA Colorado Basin River Forecast Center); and Charlie Ester and Jon Skindlov (Salt River Project). These partners, our stakeholder advisory board, helped us articulate the initial proposal s research questions and deliverables, providing feedback on draft versions of the proposal. During the first year of the project, several visits were made to agencies to present preliminary results (Bureau of Reclamation, Boulder, CO, May 27, 2015; Denver Water Board, Denver, CO, July 21, 2015; Colorado River District brown bag talk, Glenwood Springs, CO, July 23, 2015). Short presentations were made to partners and their agency colleagues. These meetings were small and highly interactive. Comments, questions, and suggestions were incorporated into analyses that were presented to the stakeholder advisory board at a workshop at the Denver Federal Center, Lakewood, CO, September 1, At the workshop, the science team (PIs Woodhouse, Pederson, and Csank and partners McCabe and McAfee) presented results to date, and sought guidance for research in the next phase. Resource managers provide extremely useful feedback. On example concerned the part of the study addressing future projections of hydroclimate. In particular, we wanted to know what approach to future projections would be most useful and relevant. There are many ways of using climate change projections, some more meaningful to management than others, and we took this opportunity to explore this from water managers perspective. Steph McAfee presented a range of approaches for using climate projections to investigate the future, and water managers discussed the most meaningful approaches. This gave Steph a starting place for her analyses. At this meeting, one of the agencies, Salt River Project, volunteered to host the final workshop. Over the next year, we interacted with our stakeholder partners in several ways. One of the most useful suggestions from the workshop was to generate a fact sheet to summarize project results published in paper in Geophysical Research Letters. This fact sheet was co-produced with our stakeholders. In several cases, preliminary results in the form of annotated PowerPoint presentations from scientific meetings were also made available to our stakeholder through the project web site. After the September meeting, McAfee followed up with several water managers (Kaatz, Prairie, and Nowak) over the course of year on the future streamflow scenarios. She presented an interactive webinar to get feedback on Oct. 14, 2016 (attendees included: James Prairie, Ken Nowak, and Jon Skindlov among others) to obtain feedback, which was incorporated into the analyses she presented at the final workshop. Our final workshop, hosted by Salt River Project in Phoenix AZ on Nov. 9, 2016, was attended by representatives of all of our original partner agencies except one. Besides representatives from the original advisory board, water managers from partner agencies but who had not been a part of the project to that point also attended, and since we opened this workshop to our partner s suggestions, several new agencies were represented (Central Arizona Project and Southern Nevada Water Authority). 10

11 Besides presenting results, we had an excellent discussion of additional questions, needs, and interest from the decision maker s perspective. These ideas have formed the basis for a new co-produced proposal. Our stakeholders include a range of agencies, from federal agencies to urban water providers to rural water conservation districts. They are all fairly sophisticated water management professionals, and value scientific information. Their interest in the project has not been for a specific piece of information or product, but for furthering their understanding about the relationship between temperature and streamflow, and the impacts of warming temperatures on drought. This understanding is critical for considering how water resources are best managed under warming conditions. As such, they are not explicitly applying information to specific practices, but gaining an awareness of the conditions that may be expected in the future. While they appear to highly values this information (and well as the opportunity to engage with our science team and their peer group), it is not possible to point to any specific application of this information, at least at this point. g. Resource management decisions that have come out of this project? None that we are aware of, to date. 14. OTHER project impacts, outcomes, or communications not discussed above. None 15. BUDGET: The original budget totaled $140,782. Of this, $22,758 went to Pederson at USGS, $65,425 to Woodhouse at University of Arizona (included isotope processing costs), $49,730 to Csank, then at DRI, and $2,870 to McAfee at University of Nevada-Reno. In December 2015, I requested a 6-month no-cost extension and a budget change, in which part of Operating Costs budget was reallocated to cover McAfee s salary for two weeks. McAfee had not been funded by the project, except for some travel, but she played a major role in the part of the project concerning climate change projection with water resource managers. The importance of this part of the project and the extent of her expertise in working with stakeholders in this area became apparent in the September 2015 workshop. There were funds remaining in the operating expenses portion of the budget due to a fewer number of samples being submitted for isotope analysis at the UA isotope lab, so it made sense to reallocate these funds to support her. As of this report, the budget has been spent out. Submitted by: (Project PI name) Connie Woodhouse 03/13/2017 Reviewed by: Stephen T. Jackson, SWCSC Director, USGS / / / 11