Streamflow Reconstructions from Tree Rings: an Example from Middle Boulder Creek

Water resource planning requires knowledge of the range of natural streamflow variability over time. What is the lowest annual streamflow one might expect over a given period of time? What is the highest? What is the long-term average? How many below-average years in a row might be expected? The time frame of reference for answering these questions is typically less than 50 years, the length of time for which streamflow records have been kept. This short period of time does not allow an evaluation of the representativeness of the 20^th century in a longer context, which is needed to detect long-term trends or changes. It also does not allow us to determine whether events such as the 1950s drought are rare events, or if they are something we should expect to occur every so often.

Tree rings are natural recorders of climate variability and have proven to be useful for extending records of climate and streamflow back hundreds of years. Tree growth is related to a variety of climatic and non-climatic factors which are integrated into patterns of annual ring widths. Tree rings can be used to reconstruct streamflow because growth is often controlled by the same climate-related factors, including precipitation and evapotranspiration, that influence streamflow variability. Thus, variations in ring widths can match variations in streamflow. Trees growing in the semi-arid western United States, including the Colorado Front Range, depend on winter and spring precipitation to recharge soil moisture, and since this precipitation also influences streamflow, these trees are useful for reconstructing annual streamflow.

Reconstructions of past streamflow from tree rings

To reconstruct streamflow, trees that are sensitive to the same set of climatic conditions that influence streamflow must be sampled. Trees that are growing on open slopes and on dry, well-drained soils are good candidates, whereas those growing close to the stream channel are usually not. Trees in drainages received more moisture so are less sensitive to climate variations and do not record climate variations as well as trees growing on dry slopes. Although tree growth is also influenced by non-climate factors (such as fire, insect infestation, and competition between trees), careful selection of trees can reduce these effects. Trees are sampled with a tool called an increment borer that removes a thin core from the tree, and does not harm the tree. The photos on this page show, first, a researcher inserting an increment borer into a tree, and then, the core as it comes out of the tree. Typically 20-40 trees per site are sampled in order to screen out variations in growth that are specific to individual trees, and enhance the common climate-related signal .

After collections are complete, the samples are taken back to the laboratory, mounted in wooden core mounts and well-sanded. It is necessary to carefully sand cores so that individual cells within the rings can be discerned under the microscope. Next, all growth rings are dated, using a ring-width pattern-matching technique called skeleton plotting, and then measured with a computer aided measuring system. The resulting series of ring width measurements are averaged together to create a site chronology. A number of statistical operations are applied in compiling a site chronology to further enhance the common climate-related signal. The two most fundamental are eliminating the age-related growth trend in the tree-ring series and removing the correlation between growth in one year and the next, which is largely due to biological processes.

To reconstruct streamflow from tree rings, variations in year-to-year growth, represented by one or more tree-ring chronologies, are calibrated with the instrumental streamflow record using a regression equation, which expresses the relationship between the two records. This equation is then used to reconstruct the streamflow record back in time for the length of the tree-ring record. The skill of the reconstruction is evaluated by comparing observed instrumental streamflow values with the values produced by the reconstruction.

Middle Boulder Creek streamflow reconstruction

Middle Boulder Creek mean monthly streamflow was reconstructed using four tree-ring chronologies from the Front Range (Figure 1). The trees are not growing very close to Middle Boulder Creek, but their growth is controlled by large regional patterns of atmospheric circulation. These same patterns also influence variations in the streamflow. The four chronologies were used in a regression equation to estimate mean annual streamflow for the years 1950-1980. This equation was then applied to the tree-ring records for 1912-1949 as an independent verification of the model. The actual streamflow values and the reconstructed values for both the calibration period (1950-1980) and the verification period (1912-1949) are shown in Figure 2. The reconstruction from tree rings explains about 60% of the variation in the streamflow record.

Very dry climate conditions, limiting to both tree growth and streamflow, tend to be duplicated more accurately in the reconstructions than very wet conditions. (When trees already have abundant moisture, further precipitation may not result in more tree growth.) This effect can be seen in Figure 2, where low flow values for years such as 1922, 1925, and 1966 are closely matched by the reconstructed values, while high flow for years such as 1921, 1923, and 1957 are underestimated by the reconstruction. In general, extreme values tend to be muted as a result of the regression process and as a consequence, reconstructions are usually a conservative estimate of past variability.

Figure 3 shows the full reconstruction of Middle Boulder annual streamflow, 1703 to 1987. The reconstruction has been smoothed to facilitate comparisons of low flow events. The reconstruction captures the 20^th century periods of low flow (1930s,1950,s mid-1960s), but these events appear to be moderate compared to low flow events in previous centuries. In particular, there are several episodes of persistent and extreme low flow values in the 19^th century, during the 1840s and the 1880s. There is also a period of extreme low flow in the early 1700s. This period may be exaggerated somewhat, because fewer trees are in the earliest part of the record, but other reconstructions for the region also show this to be a dry period.

Low flow events in Middle Boulder Creek

How does the 20^th century record of flow compare to the full 300-year record? We can look at distribution of extreme low flow events to answer this question. For this analysis, reconstructed flow values for single years, 3-year averages, and 5-year averages were ranked. Those that fell into the driest 10% of flow years were grouped according to the half-century periods in which they occurred. Figure 4 shows how these dry event single years and 3- and 5-year averages are distributed over the past three centuries. The extreme low flow single years are fairly evenly distributed across time, with concentrations of extreme values in the 1840s and 1880s, and a marked lack of extremes in the first half of the 20^th century (Figure 4, top). The three-year averages show a greater clustering of low flow periods, with the bulk occurring in two 19^th century periods, along with a peak of extreme values in the early 18^th century (Figure 4, middle). The extreme events in the 20^th century are primarily associated with the 1950s drought. The five-year average extreme lowest flows occur almost exclusively in the 19^th century (Figure 4, bottom), and are concentrated around 1880, 1887, and the decade of the 1840s. This examination of extreme values suggests that multi-year low flow periods are not evenly distributed across time. Specifically, the 19^th century appears to be characterized by much more frequent extremely low three- and five-year flow conditions than either the previous or following centuries.

Summary

Tree-ring reconstructions of streamflow have proven to be extremely useful for extending records of streamflow back in time. These extended records help us better understand the range of natural streamflow variability, and are vital for assessing the representativeness of the 20^th century record. This reconstruction suggests that the 20^th century record of Middle Boulder Creek streamflow may not be representative of flow in past centuries. If the record of past variability is used as a guide to the future, then it might be wise to acknowledge the possibility of more persistent low flow events in the future.

Although tree-ring reconstructions of streamflow offer invaluable insights on the long term and low-frequency behavior of streamflow, further work is being done to increase the usefulness of these records for water resource management and planning.

For further information on tree rings and drought

More details on tree rings and the science of dendrochronology can be found within in this web page:

http://web.utk.edu/~grissino/

More information about reconstructions of past drought in North America from tree rings and other types of paleoclimatic data can be found within this web site:

http://www.ngdc.noaa.gov/paleo/drought/drght_home.html

References on tree-rings and reconstructions of streamflow in the western U.S.