Changing Streamflow in Icicle, Peshastin, and Mission Creeks

About this dataset

As part of the Icicle Work Group (IWG), a diverse set of stakeholders has been working to identify collaborative solutions to water management in Icicle Creek. Water management decisions that are made today will have implications for decades to come. Given the large changes in climate and hydrology anticipated in the coming decades, such plans will need to account for the effects of climate change if they are going to be robust.

The purpose of this project is to leverage existing hydrologic change datasets to estimate future changes in streamflow in Icicle, Peshastin, and Mission Creeks as well as the seven alpine lakes. These will be used to evaluate proposed alternatives for managing water in Icicle Creek.

This project is intended to provide screening-level estimates of future streamflow in these three watersheds. The emphasis on off-the-shelf streamflow projections made it possible to assess these changes at a much lower expense than would have otherwise been possible. Comparisons across datasets allow for a quick assessment of the consistency of the results

Figure 1. Map of the study area, including the location of the streamflow sites for which streamflow projections were derived. The map also shows the delineation of the watersheds for the three creeks.


The results, datasets, and methods are all described at length in the project report. Data files and figures are also available for download here. In short, we find that projected changes in streamflow follow the expected response to a decrease in snowpack: more precipitation falls as rain in winter, snow accumulation is reduced, and the snow melts out earlier. As a result, we see increased winter flows, and an earlier and less pronounced peak in spring flows, and a decrease in summer flows in each of the three creeks.

Since the projections will primarily be used for a screening-level assessment of proposed infrastructure and management changes, one simple way to distill the results is by considering the average projection for each dataset (see Datasets section below). This is a very simplistic approach, since it involves averaging over different numbers of models for each dataset and, in some cases, averaging results from two different greenhouse gas scenarios. Nonetheless, it is useful for getting a general sense of what the projections show. The figure below shows these results for the 2080s. For comparison, the right-hand panel shows the difference between observed flows in 2015 and the long-term average. The temperatures experienced in 2015 are about what we expect will be “the new normal” sometime in the second half of this century. The consistency between the model results (left-hand plot) and observations (right-hand plot) suggests that the projections are robust.

Figure 2. Comparing the projected changes for the 2080s (left) to the percent difference between 2015 flows and normal conditions (right). Percent differences for the 2080s (2070-2099) are calculated relative to the 1980s (1970-1999). The plot shows the average (“average of the averages”) and interquartile range for each of the five datasets. Percent differences for 2015 are calculated relative to observed monthly flows for 1950-1999. For each month, the median among all years is shown along with the 25th-75th percentile range.

Getting the Data

Interactive tool

The visualization below shows the projected changes in streamflow for each calendar month as well as for the daily extremes. Projections were obtained from five off-the-shelf climate change datasets, each produced using different approaches. Using the tool, you can look at the range of projections among datasets for different future time periods. Models are generally consistent in the direction of change (i.e.: most agree on whether streamflow is increasing or decreasing for each metric), but there is a wide range among projections for most metrics. If it is suitable to consider only the average projection for each dataset, these are likely to provide a more reliable estimate of change.

Additional detail is provided below and in the project report.

How to use this tool

Layout of the tool

The layout of the tool is illustrated in the figure below. There are two tabs that can be viewed: one showing percent changes for each metric, and another showing the seasonal cycle of monthly streamflow for each water year, for both historical and future simulations. All user selections can be made on the left-hand side of the figures, while all legends are included to the right.

Figure 3. Screenshot of the online tool. The tool has two tabs: one showing the percent changes for each metric, facilitating comparisons across all datasets, and the other showing the full seasonal cycle of historical and future monthly flows, in which only one dataset and scenario can be viewed at a time.

Menus and Options

Streamflow site

The drop-down menu on the top left side of the tool allows you to select the streamflow location you want to consider: Icicle, Peshastin, or Mission Creeks, or one of the seven Alpine lakes.


On the percent changes tab, you can select a metric (only monthly averages are shown for the monthly streamflow tab). We use the term “metrics” to describe different quantities that can be derived from daily streamflow. The three options are: monthly average streamflow, minimum daily streamflow for each water year, and maximum daily streamflow for each water year. The daily streamflow extremes are estimated in terms of the recurrence interval (e.g., 10-year event), and are only available for Icicle, Peshastin, and Mission Creeks. Once you have chosen a category of metric (e.g, Monthly Average Streamflow), then you can pick a metric from the radio buttons below (e.g., January).


On the monthly streamflow tab, you can choose which dataset you want to view. The data used in this project come from the following sources:



A set of hydrologic projections that were developed as part of the Integrated Scenarios of the Future Northwest Environment project. Climate projections stem from the statistically downscaled MACA approach (Multivariate Adaptive Constructed Analogs, Abatzoglou and Brown, 2012), and are based on the latest global climate model projections (CMIP5, Taylor et al. 2012). The MACA downscaling is applied to the 10 GCMs identified by Rupp et al. (2013), each of which include both a low and a high greenhouse gas scenario (RCP 4.5 and 8.5, respectively), for a total of 20 future climate scenarios. The projections extend from 1950-2099. Hydrologic simulations were made using VIC version 4.1.2.


A modified version of the MACA dataset in which average monthly temperature and precipitation was adjusted (or bias-corrected, hence bcMACA) to match the estimates derived from the observationally-based Parameter-Elevation Regressions on Independent Slopes dataset (PRISM, version AN81M monthly, Daly et al. 2008). Over the U.S. the monthly PRISM time series was used to apply the adjustments, while over Canada the long-term average for each month was adjusted to match the long-term average from PRISM.


A new set of hydrologic projections developed for the 2016 Columbia River Basin Long-term Water Supply and Demand Forecast (Hall et al. 2016). Hydrologic model simulations are driven by the same MACA projections described above, except that only five of the 10 GCMs are used, each again for both a low and a high greenhouse gas scenario, adding up to a total of 10 future scenarios. Hydrologic simulations are performed using VIC-CropSyst v2.0 and run for two 31-year time periods: 1981-2011 and 2020-2050. This means that future changes are only available for the 2030s, and that changes for this time period are assessed relative to 1981-2011 instead of 1970-1999 as with each of the other datasets.


A previous set of projections, developed with funding from Washington State House Bill #2860 (HB2860, Hamlet et al. 2013). Climate projections stem from the statistically downscaled BCSD approach (Bias Correction and Spatial Disaggregation, Wood et al. 2004), and are based on the previous set of global climate model projections (CMIP3, Meehl et al. 2007). The BCSD downscaling was applied to seven GCMs based on the ranking of Mote and Salathé (2010). In this project we analyze results for a moderate greenhouse gas scenario (A1B). The projections extend from 1950-2099. Hydrologic simulations were made using VIC version 4.0.7.


Regional Climate Model simulations using the WRF model (Weather Research and Forecasting; Skamarock et al. 2008, Salathé et al. 2010). Projections stem from two GCMs selected from the previous set of global climate model projections (CMIP3, Meehl et al. 2007), both for a moderate greenhouse gas scenario (A1B). The projections extend from 1970-2069, meaning that future changes are not available for the 2080s. Hydrologic simulations were performed using VIC version 4.1.2.

Greenhouse Gas Scenarios

On the monthly streamflow tab, a greenhouse gas scenario must be selected after you choose a dataset. As noted in the table above, not all datasets include the same greenhouse gas scenarios. The following greenhouse gas scenarios are included:


Scenario Scenario characteristics Description Citation
RCP 4.5 A low scenario in which greenhouse gas emissions stabilize by mid-century and fall sharply thereafter. “Low” Van Vuuren et al. 2011
A1B A medium scenario in which greenhouse gas emissions increase gradually until stabilizing in the final decades of the 21st century “Moderate” Nakicenovic et al. 2000
RCP 8.5 A high scenario that assumes continued increases in greenhouse gas emissions until the end of the 21st century “High” Van Vuuren et al. 2011

Future Time Period

Both tabs allow you to select the future time period you want to view. Results have been assessed for three future time periods: the 2030s, 2050s, and 2080s. In order to isolate just the long-term climate change signal from short-term natural variability, each of these is actually composed of the average of the 30-year period that is centered on that decade. For example, results for the “2030s” are actually calculated by taking the average for the years 2020-2049. With the exception of the WSU dataset, percent changes are all calculated relative to the average for the years 1970-1999. For the WSU dataset, changes are assessed relative to 1980-2009. The same historical time periods are used for the historical results on the Monthly Streamflow tab.


This project was funded by Chelan County’s Department of Natural Resources.


Mauger, G.S., Lee, S.-Y., Won, J.S. Effect of Climate Change on Streamflow in Icicle, Peshastin, and Mission Creeks (2017). Report prepared for Chelan County. Climate Impacts Group, University of Washington, Seattle.

Download the report here.


Please contact Guillaume Mauger ( with any questions about this project.

Back to Top