Understanding the Causes of Streamflow Changes in the Eurasian Arctic Jennifer C. Adam1, Fengge Su1, Laura C. Bowling2, and Dennis P. Lettenmaier1 Department of Civil and Environmental Engineering, Box 352700, University of Washington, Seattle, WA 98195 2. Department of Agronomy, Purdue University, West Lafayette, IN 47907 FWI All-Hands Meeting, Bodega Bay, California, June 6-8, 2007 Photo: http://www.globalcarbonproject.org/ (2) Role of Precipitation in Annual Streamflow Changes Potential Streamflow Controls, Modulators, and Linkages ABSTRACT Since the 1930s, the combined streamflow from the six largest Eurasian rivers discharging to the Arctic Ocean has been increasing. Much of this change is thought to be due to an acceleration of the hydrologic cycle (and associated increased precipitation), but recent studies show inconsistencies in long-term streamflow and precipitation trends, perhaps due to uncertainty in the observations. Alternatively, these inconsistencies may imply another freshwater source, such as the release of water from storage via permafrost degradation. For many of the Eurasian rivers, an increase in annual streamflow volume has been accompanied by a shift in seasonality, in which winter streamflow has been increasing and summer streamflow has been decreasing. However, this apparent signature is confounded by the fact that climate-induced changes in streamflow seasonality may be similar to those resulting from the operation of a number of large reservoirs that have been constructed on these rivers over the last fifty years. A combination of data analysis and large-scale hydrological modeling are used to explore the historical controls on annual and seasonal streamflow for eleven study basins in the Eurasian Arctic. We are interested in how the controls varied regionally as well as in time. We examine the following potential mechanisms of change: Each of these parts corresponds, according to color, to the elements in the flow chart (on the right) and to the individual panels (following the flow chart). Objective: To determine for which basins and periods streamflow trends are inconsistent with precipitation trends and to develop a hypothesis to explain mismatches. Tools: Observation-based gridded precipitation and observed and reconstructed streamflow data. Tasks: Develop ½° gridded precipitation product with bias adjustments for: long-term variability due to a changing station network in time for gauge under-catch and orographic effects on mean monthly precipitation For each study basin, perform streamflow trend analysis for ~400 periods between 1937 and 2000 and isolate the periods for which trends are 99%. For periods in which streamflow trends are significant, calculate precipitation trends. For each basin, plot streamflow trends again precipitation trends and sort basins according to climate and permafrost extent. Main Findings: Agents of change depend on mean climate and permafrost state. Hypothesis to explain divergence of streamflow trends from precipitation trends (see hypothesis below) is developed from results. Cold Permafrost Basins: positive divergence permafrost melt Non-Permafrost Basins: negative divergence increase in ET Threshold-Permafrost Basins: competing effects Publication: Adam and Lettenmaier (2007, in review) To explore the mechanisms causing observed streamflow changes, we examine the following climate and human-induced controls, the components of the system that modulate the effects of the controls, and the linkages between them. In the flow chart below, the bright colors represent each of the three mechanisms that are explored (see left panel) and relate to each of the following three color-coded panels. Gray indicates that this component or linkage is involved in more than one of the explored mechanisms. P, mm/year monthly mean long-term variability short-term variability + = Development of Gridded Precipitation (e.g. Lena) Annual Q Seasonal Q Modulator Snow ET Subsurface Storage Control Precipitation Air Temperature Artificial Reservoirs Streamflow (Q) Effects Climate Human Precipitation Streamflow (4)Aldan (cold) (2)Yenisei (threshold) (9)Chulym (non-permafrost) 1940 1960 1980 1940 1960 1980 2000 Stream Flow Trend, mm/yr2 Precipitation Trend, mm/yr2 Trend mm year-2 Influence of artificial reservoirs on annual and seasonal streamflow changes Application of a coupled hydrology/routing/reservoir model Role of precipitation changes in annual streamflow changes Application of observation-based gridded precipitation and (observed and reconstructed) streamflow data Influence of climate controls and modulators (permafrost, snow, and evapotranspiration) on annual and seasonal streamflow changes Application of the VIC land surface hydrology model (1) Reservoir Influences Objective: To understand the effects of reservoir construction and operations on long-term annual and seasonal streamflow trends. Tools: Coupled hydrology/routing/reservoir model and observed streamflow data. Tasks: Simulate influences of reservoirs on streamflow at outlet of the primary basins using the coupled model. Create a reconstructed (“naturalized”) streamflow product by subtracting the simulated reservoir influences from observed streamflow at the basin outlet. Perform trend analysis on observed versus reconstructed streamflow. Main Findings: Little influence on long-term annual trends. Percent of ’37-’97 seasonal trends explained: Lena: winter 80% spring 30% Yenisei: all seasons 100% Ob’: winter 100% summer: 35% Publication: Adam et al. (2007a) (in review) (3) Climate Change Impacts on Long-Term Streamflow Changes and the Roles of Snow, Permafrost, and ET Large Dam Objectives: To evaluate the individual roles of precipitation and temperature in observed streamflow changes. To examine the roles of the modulators (snow, permafrost, and ET) in long-term streamflow changes and their relationships to each other. Tools: Observation-based gridded precipitation and temperature and the VIC hydrology model Tasks: (in progress) Preliminary 1930-2000 streamflow simulation – does VIC reproduce long-term trends? Improvement of frozen soils simulation in permafrost regions bottom boundary specification, distribution of thermal nodes, implicit solver, permanent ground ice initialization and tracking Analysis of final simulations Quantification of change in each water balance component Sensitivity of streamflow to precipitation vs. temperature changes Identifying the roles of snow, permafrost, and ET as modulators Publication: Adam et al. (2007b, in preparation) Study Domain Observed Yenisei: threshold permafrost basin Ob’: non-permafrost basin Lena: cold permafrost basin Q, 103 m3s-1 Simulated 1940 1960 1980 We focus on Northern Eurasian basins (stream flow has been shown to be increasing and longer records exist for these basins). We chose three primary and eight secondary basins and sub-basins, all with varying extents of permafrost. 1 2 3 4 VIC Hydrology Model Reservoir Model River Routing Model Primary Study Basins Secondary Study Basins Current Status: Long-term streamflow changes are not captured in permafrost basins. Reasons include improper permanent ground ice initialization and lack of tracking. We are currently finalizing improvements to the frozen soils algorithm to handle permafrost. Lena Yenisei Ob’ Reservoir Continuous Permafrost Discontinuous Permafrost Sporadic Permafrost Yenisei Isolated Permafrost Seasonally Frozen SUMMARY Our efforts to understand mechanisms behind observed streamflow changes include three parts, each corresponding to a publication: Examination of influences of artificial reservoirs (Adam et al. 2007a) Reservoirs are responsible for much if not all of observed changes in streamflow seasonality at the outlets of the primary basins. Development of hypothesis to explain mismatches between trends in annual precipitation and streamflow (Adam and Lettenmaier, 2007) The extent and state (temperature, ice richness, etc…) of permafrost in a basin is an important factor in understanding past changes. Influences of climatic change on long-term streamflow changes and the roles of snow, permafrost, and ET (Adam et al. 2007b) Current work involves including the tracking of permanent ground ice in the VIC model frozen soils algorithm. Accurate prediction of future streamflow changes is dependant on understanding the mechanisms controlling historical changes. For example, we anticipate that streamflow increases will continue to accelerate if these are due entirely to warming-induced acceleration of the hydrologic cycle (i.e. increasing precipitation). Conversely, if past increases are also partially due to permafrost degradation, we may not expect streamflow increases to accelerate as quickly once permafrost melt is complete. Streamflow, 103 m3s-1 Annual Winter Summer Operations begin for 1st Reservoir Primary Basins Permafrost Extent (Brown et al. 1998) Area (106 km2) All Types Cont. Discont. Sporadic Isolated Lena 2.43 100% 80% 11% 6% 3% Yenisei 2.44 89% 33% 12% 18% 26% Ob’ 2.95 27% 2% 4% 9% Observed Adam et al. (2007a) McClelland et al. (2004) Reconstructed: