| 12 Jun 2026
Extending terrestrial water storage anomalies beyond the GRACE era using tree rings
Abstract. Satellite observations from NASA’s Gravity Recovery and Climate Experiment (GRACE) and GRACE Follow-On have transformed monitoring of terrestrial water storage (TWS) anomalies, but their short record (2002–present) limits characterization of long-term variability and extremes. Here, we evaluate whether tree-ring width indices (RWI) can be used to extend TWS variability across western North America. Using 144 chronologies, we find predominantly positive correlations between RWI and GRACE-observed TWS anomalies, with mean annual correlations of r = 0.25 (IQR: 0.07–0.50) and the strongest relationships during summer (mean r = 0.27; 37% of sites significant at p < 0.1). Correlation strength is highest in moisture-limited environments and among Pinus species. In the Upper Colorado River Basin, a two-site composite model explains 69% of the variance in GRACE-observed summer TWS anomalies (r = 0.83) and enables reconstruction back to 1567 CE. The reconstruction exhibits a standard deviation of 29.9 mm and a peak-to-trough range of 167.6 mm, and agrees well with other hydroclimatic records, including reconstructed Palmer Drought Severity Index over 1567-2000 (r = 0.79) and climate-data-driven TWS estimates over 1901-2000 (r = 0.73). Using climate-data-driven TWS estimates as a longer reference record, moving-window analyses show that calibration stability improves rapidly from 10-year to approximately 25–30-year windows, after which additional years yield more gradual gains. Together, these results suggest that the now nearly 25-year GRACE and GRACE Follow-On record is approaching the duration needed for more stable tree-ring–TWS calibration, and that targeted extension of moisture-sensitive chronology networks through the satellite era may provide a robust pathway for reconstructing multi-century TWS variability in western North America and other similar regions.
This manuscript contributes to understand of the statistical relationships between tree-ring and GRACE-gravity data. This is an interesting because of the potential for tree rings to serve as proxies of historic water storage volumes. This manuscript emphasizes statistical relationships and takes a light approach to addressing the meaning of those relationships. Although statistical understanding can be useful on its own, the links between tree growth and water storage are understood well enough—both generally and specifically in the study region—that they deserve to be integrated into the statistical modeling and interpretation. By omitting most of the literatures on those topics, robust interpretation of the results of this work is difficult, and the utility of the proxy developed using tree rings remains unclear.
The hydrological meaning of GRACE variations is highly relevant to the current work, but the manuscript does not address this issue except cursorily. The nature of the hydrological store being investigated is just as important as the nature of the trees being investigated, and there are large variations in water storage across the study region. There is a substantial literature on the hydrological meaning of GRACE data in the region of this work, and it would improve the manuscript substantially if it were added to inform the methods and discussion of the results.
The utility of tree-rings for reconstructing hydrological history depends on why those correlations exist: it matters whether GRACE-relevant storage directly affects tree growth, or whether each of them are independently responding to climatic drivers. In the first case, we can use tree-rings to infer historic regional water storage, but in the second case we can only use tree-rings to infer the drivers of storage but not storage itself, because historic storage variations in the past were not subject to the same management effects as they are today, during the calibration period for the reconstruction.
Detailed comments:
L46 the hypothesis here is that trees respond to changes in water reservoirs measured by GRACE. However, these are vague causal statements, and they are in conflict with section 4.2, which says that tree growth in energy-limited regions is not well correlated with water availability. What is the non-statistical hypothesis of the study?
L68 stiffness of the splines is a crucial control on how much low-frequency variability is retained. What was done here? At what timescales is variance large in the GRACE data and are those same timescales emphasized in the tree-ring data? Large-scale changes in water storage are of course slower than soil moisture variability.
L88 L194 presumes causation, whereas noncausal correlation is not excluded.
L183 Pinus spp. are not likely to be directly accessing groundwater because they are mainly in uplands in semi-arid climates. This suggests the tree-ring and GRACE time series are both responding to a common driver, not that groundwater is controlling tree growth.
L233, Sec 4.1 if the TWS and PDSI reconstructions were based on the same tree-ring data, then these comparisons would be much more convincing. Why not include all 80 series of Steiger instead of just two series, if the goal is to understand whether the same tree-ring data can generate contrasting reconstructions?
L311 what climate-date-driven TWS anomalies? This analysis isn’t fully described. Omitting it would not hurt the manuscript.
Quibbles:
L28 conjugation of “characterize”
Table 1 is a little rough. Alphabetization by common name is annoying, column headings “species” and “genus” (though genus is not needed as a column at all because it is already listed in the species column), only one species name should be listed, authority names should not be italic.