Jump to navigation Jump to search


Figure 1. Reconstructed Colorado River streamflow at Lees Ferry, 762-2005, data from Meko et al. (2007). Reconstructed values are shown in blue and observed values are shown in gray, both plotted as the 10-year running mean. The long-term reconstructed mean is shown by the dashed line. (Image source: TreeFlow)

Paleohydrology refers to information about streamflows that occurred prior to the existence of observed or gaged records. In the Colorado River Basin, the primary source of paleohydrology is the ring widths of moisture-sensitive trees. Tree-ring reconstructions of streamflows have been developed for several dozen gages within the basin, extending the perspective of hydrologic variability back up to 1200 years. While some information about sub-annual features of climate and hydrology can be extracted from tree rings, the streamflow reconstructions that have been developed for the basin are almost exclusively for annual (water-year) flows.

Tree-ring reconstructions of streamflow have been used by many water management entities using Colorado River water in their long-range planning, to provide system stress tests that are more extreme than the observed hydrology, but are still physically plausible, as they have already occurred in the past. Since most water system models operate on daily or monthly timescales, the annual reconstructed streamflows need to be processed to allow them to be input into system models.

The most influential and widely used reconstructions in the basin have been those developed for the Lees Ferry gage, which because of its importance to basin water supply has been reconstructed by several research teams, beginning in the 1970s. The longest recent reconstruction was developed by David Meko and others (Meko et al. 2007), and extends from 762-2005 (Fig. 1). This reconstruction was used by Reclamation in analyses to support the 2007 Interim Guidelines, and several subsequent reports.

A consistent finding from the Lees Ferry reconstruction and other basin tree-ring reconstructions is that the period prior to 1900 contains a broader range of variability than that seen in the observed records, notably drought events worse that are more severe than the worst-case droughts since 1900 (Fig. 1). The most severe and sustained of these paleo-droughts have been termed "megadroughts." The tree-ring reconstructions also show that the 20th century had overall more favorable hydrology than the preceding 3-10 centuries, with two multi-decadal pluvials or wet periods, from 1905-1930 and from 1982-1999. The earlier of those two periods may have been the wettest multi-decadal period in the past 500 years.

Data and Tools


The TreeFlow resource provides access to data, metadata, and plots of nearly all existing tree-ring reconstructions of streamflow for the Colorado River Basin, as well as other river basins throughout the U.S. Currently there are 31 gages in the Upper Basin and 4 gages in the Lower Basin for which reconstructions are available.


The Paleoflow tool provides additional analysis and plotting options for the Colorado River Basin reconstructions that are available on TreeFlow, including dynamic time-series plots, analysis of extreme events, and comparisons of distributions (probability density functions) between two periods. Make sure Time Resolution = Annual to see the full list of reconstructions.

Additional Resources

State of the Science Report

Chapter 10 of the State of the Science report (Paleohydrology) covers many aspects of tree-ring reconstructions of streamflow in detail: methodology, the multiple reconstructions for Lees Ferry, data uncertainties, and applications of the reconstructions.

New and Notable Research (2020-present)

Bias Correction of Paleoclimatic Reconstructions: A New Look at 1,200+ Years of Upper Colorado River Flow

Robeson et al., 2020, Geophysical Research Letters


Robeson et al. take the widely used tree-ring reconstruction of annual streamflow at Lees Ferry by Meko et al. (2007) and apply quantile mapping to bias-correct the reconstruction so that the reconstructed flows better fit the distribution of flows in the observed record. After bias correction, the 1100s megadrought is even more extreme than in the original Meko et al. reconstruction. Robeson et al. contend that such bias-correction is advisable when paleo-reconstructions are compared directly to observed records.


It has been long known that the regression procedures typically used in reconstructions tend to compress the range of observed variability, which can especially effect multi-year extreme events (wet and dry). Bias-correction via quantile mapping is a reasonable approach to try to correct this tendency and other biases in the reconstructed values. But it is not a panacea; users should keep in mind that all reconstructions are estimates of the past hydrology.