- About Us
- SW Climate
7 things western water managers should know about tree-ring reconstructions of streamflow
Published May 23, 2007
1) The science behind streamflow reconstructions has a long history. In the 1930s, researchers first began to quantify the close relationship between treering growth and the amount of water flowing in rivers and streams (streamflow) in the western United States. In the 1960s, researchers began to employ computers and modern multiple linear regression techniques to develop tree-ring reconstructions of streamflow. Techniques have been progressively refined since then.
2) Tree growth in the West is closely associated with moisture variability, leading to high-quality streamflow reconstructions. In semi-arid climates, the same two climate factors generally control both the growth of moisturelimited trees and the amount of runoff trickling into streams. Precipitation is obviously important. The other important climate factor is evapotranspiration, which refers to water evaporated from the landscape and transpired through plants. Several widespread conifer species such as ponderosa pine, pinyon pine, and Douglas-fir are particularly responsive to the variability of moisture from one year to the next. This sensitivity is even greater when they grow on dry, rocky sites like those found on many western mountainsides (Figure 4). Thus, the trees that are most likely to show annual changes in tree-ring size from annual changes in moisture levels are not the ones growing closest to rivers, but the ones eking out a living on steep slopes in the surrounding watersheds. Because of this, the relationship between tree growth and streamflow is not direct. Instead, tree growth and streamflow are robustly linked by the regional climate that influences both.
3) Combining samples from many trees into one “chronology” improves the moisture signal from a site. At each site, researchers collect pencil-sized core samples from living trees (usually 20 to 30) to maximize the common climate signal. After preparing and sanding the cores so every ring is visible under a microscope, researchers use sophisticated equipment to measure each annual growth ring. Next, they compare the growth patterns among the treesfor a given site, crossdating them to account for any missing or false rings and assigning an exact year to each annual ring. Then, measured ring widths from multiple trees for each site are averaged into a timeline showing the ups and downs of annual growth, which serves as the site chronology. Finally, multiple tree-ring chronologies from the region are combined to reconstruct streamflows for a particular stream gage.
4) The reconstruction assumes the documented relationships between specific trees’ growth and streamflow extends back in time. Researchers use several statistical methods to find the chronologies that best reflect streamflow measurements of a specific gage on the river in question. The chronologies that perform the best in estimating the gaged flows are selected to reconstruct earlier flows. The multiple linear regression equation derived from the relationship between tree growth and streamflow serves as the reconstruction model. After creating the model, researchers evaluate its skill by testing it on independent data or on data that had been left out of the model specifically so it could be used for these calibration purposes. Scientists then apply the model to the full tree-ring record, using the reconstruction to extend the streamflow record back hundreds of years.
5) Trees generally do well at estimating streamflow, but there is always uncertainty around the reconstructed flow. Streamflow reconstructions in the West generally explain about 50 to 80 percent of the variance observed in the gaged record. They also capture the important features, particularly droughts, of the gaged record. But trees are imperfect recorders of streamflow. About 20 to 50 percent of streamflow typically relates to factors that are not reflected by the growth of trees in the sampled areas. Researchers can assess the statistical uncertainty in the model by comparing the differences between the reconstructed flows and the gaged flows. They use this information to generate “confidence intervals.” For example, an 80 percent confidence interval suggests there is an 80 percent chance the values fall within the illustrated range (Figure 1). In effect, this represents each year’s reconstructed flow as a range of plausible flows, with the most probable value in the middle. In addition to the uncertainty shown by the confidence intervals, there is an undefined amount of uncertainty relating to the choices made in data treatment and modeling approaches.
6) By providing a longer window into the past, the tree-ring reconstructions describe the natural variability of climate more completely than gaged records. The tree-ring record clearly shows that the streamflow variability of the 20th century does not simply repeat itself moving back in time. Reconstructions indicate the existence of longer and more severe droughts than those measured in the gaged record—and longer and more pronounced wet periods, too. They also demonstrate that the mean annual streamflow has changed over past centuries. While human activities exert a stronger influence on climate, the influence is superimposed on natural variability. Climate models project that the range of hydroclimatic variability will likely increase in the future relative to the recent past as seen in the instrumental record. Thus the greater variability seen in the multi-century tree-ring reconstructions of streamflow may be a useful analogue for increased future variability. Using the reconstructed flows rather than just the gaged record as the frame of reference for water management planning can help reduce the number of “surprises” that will arise as we head into a climatically uncertain future.
7) Water managers can apply the streamflow reconstructions in different ways, depending on their needs and capabilities. The uses of treering reconstruction of streamflow fall into three general categories:
- An informal guide for water managers, stakeholders and decision makers.
- A quantitative assessment of long-term hydrologic variability. For example, assessing the reconstructed frequency of droughts of a given duration and/or severity.
- A direct input into hydrologic models of a water system. This allows water managers to model system performance using the reconstructed streamflow as they would the gaged measurements. This typically requires additional processing of the reconstruction, which provides annual values, into the monthly, weekly, or daily time steps required by the system model.