| 10 Aug 2026
How trees influence the spatiotemporal dynamics of soil moisture in different forest ecosystems
Abstract. Soil moisture (SM) is a key variable in terrestrial ecosystems, regulating water and energy exchange at the soil-atmosphere interface. In forest ecosystems, SM exhibits pronounced spatiotemporal variability at the tree-scale in the topsoil of the active root zone due to complex interactions among multiple factors such as soil properties, topography, climate, and vegetation. Existing research on spatiotemporal SM characteristics in forests is limited, impeding our understanding of SM-vegetation interactions. A detailed understanding of when and to what extent vegetation influence topsoil moisture spatiotemporal variability in forests is required to quantify its consequences for (eco‑)hydrological processes and to test hypotheses about the dominating influences on spatiotemporal patterns across different ecosystems.
This study characterizes spatiotemporal dynamics of topsoil SM. Our objectives are to quantify plot-scale spatial variability and its temporal evolution from event- to seasonal scale, thereby improving our understanding of SM dynamics across wet and dry states in pure- and mixed-species forest stands. We recorded SM with a dense monitoring network of 400 sensors (SMT100, Truebner GmbH, Germany) installed in the topsoil in a tree-centered design across four plots of mixed and pure Douglas fir, Beech and Silver fir trees in the ECOSENSE forest, southwestern Germany. Statistical and Empirical Orthogonal Function (EOF) analysis together with temporal stability (TS) analysis were combined to quantify spatial SM dynamics from a 2.5-year dataset (2023–2025). Spatial SM follows a consistent annual cycle across all plots transitioning between wet- and dry-preferential states. SM shows high variability for all plots and most plots display a characteristic convex upward curve for the spatial mean soil moisture to coefficient of variation (SM̄–CV) relationship. The CV peaks at intermediate wetness and declines toward the driest and wettest states. Only the Mixed plot shows a linear, negative SM̄–CV relationship. Daily SM skewness follows the same convex relation with negative skew for wet conditions (SM̄ > 20 %) and positive skew for dry conditions. The EOF analysis identified two to three statistically significant modes per plot, together explaining > 80 % of spatial variance. The first EOF explains > 60 % of the variance at each plot and represents the dominant, stable, time-invariant spatial pattern. Seasonal wetting and drying cycles show that spatial variability differs between wet and dry phases, suggesting distinct local and non-local controls dominate the SM patterns according to the wetness state and seasonal phase of the system, increasing or decreasing spatial SM heterogeneity. The temporal stability analysis identified persistently wetter and drier locations. Mean relative difference (MRD) within each plot ranges from –75 % (dry) to +50 % (wet), with the widest range in the Silver fir plot and the narrowest range in the Beech plot. Within each plot, a subset of representative locations could be identified, but the exact location in the plot is random. Spearman rank correlations of spatial SM between the Douglas fir and Beech plot show high spatial stability (ρ ≥ 0.8) for wet phases (winter) whereas lower correlation (ρ ≈ 0.45–0.60) during dry phases. The findings of this study can help to inform sampling strategies and modelling approaches. The dataset presented here is appropriate for detailed analyses of event-based wetting-drying dynamics, and when combined with additional data, it can disentangle the relative contributions of local and non-local controls to the observed spatial patterns and pattern stability.
The authors investigate the spatiotemporal variability and temporal stability of topsoil moisture in four forest stands in the Black Forest, Germany, using a remarkably dense network of 400 soil moisture sensors and 2.5 years of observations. The study combines analyses of the relationship between spatial mean soil moisture and its coefficient of variation, skewness, empirical orthogonal functions (EOFs), temporal stability, and Spearman rank correlations. The dense tree-centered sensor network provides a very valuable dataset for investigating small-scale soil moisture variability in forest ecosystems. The study identifies several interesting patterns, including a pronounced convex relationship between mean soil moisture and spatial variability in three of the four stands, strong dominance of the first EOF, persistent wet and dry locations, and seasonal changes in spatial pattern stability.
The manuscript addresses an important topic in forest ecohydrology and makes use of an unusually detailed observational dataset. In particular, the combination of high spatial density and multi-year temporal coverage is a major strength of the study. The results on state-dependent spatial variability and temporal persistence could be useful for improving soil moisture sampling strategies and for understanding the scale dependence of forest soil moisture dynamics.
However, in its current form, the manuscript contains several weaknesses that limit the strength of some of its conclusions. Most importantly, the manuscript frequently interprets differences between the four plots as effects of tree species, although the experimental design does not provide replicated tree-species treatments and the plots differ in other potentially important characteristics. In addition, the treatment of missing data in the EOF analysis requires clarification. Some methodological and statistical descriptions should also be improved, and several interpretations in the Discussion are stronger than supported by the observations. These issues are detailed in the major and minor comments below.
Major comments:
Specific comments:
Title: “How trees influence…” implies a causal analysis that is not supported by the non-replicated plot design. Consider replacing it with “Spatiotemporal dynamics…” or “Spatiotemporal patterns and temporal stability…”. Also “different forest ecosystems” is somewhat misleading because all four plots are located at the same research site in the Black Forest. “Different forest stands” would be more precise. I suggest the following title: “Spatiotemporal dynamics and temporal stability of topsoil moisture across four forest stands in the Black Forest, Germany”
L10–12: The conceptual framing appears one-sided, as it considers vegetation primarily as a control on topsoil moisture variability. However, vegetation and soil moisture are involved in a bidirectional interaction: vegetation affects soil moisture through processes such as interception, transpiration and root water uptake, while soil moisture availability in turn constrains vegetation water uptake and functioning. I suggest revising this statement to explicitly acknowledge these feedbacks. This would also provide a more balanced conceptual framework for interpreting the observed spatiotemporal soil moisture patterns.
L15–17: Please clarify whether the 400 sensors are equally distributed among the four plots and report the number of sensors per plot in the Abstract or Methods.
L25: The wording “time-invariant spatial pattern” should be reconsidered. EOFs are statistically time-invariant basis functions; their physical interpretation as stable environmental structures requires additional evidence.
L124: The objectives should be revised to avoid claiming that the monitoring network allows “isolated” observation of species-specific vegetation influence. It provides a very detailed observation of different forest stands, but it does not isolate species effects because each stand represents a single, non-replicated site and therefore tree species/composition is confounded with site-specific factors such as soil properties and topography.
L155–156: The difference in topographic setting between the Silver fir plot and the other plots is important and should be emphasized more strongly when comparing the plots.
L166-170: Please provide more information on the accuracy and calibration of the SMT100 sensors in the specific forest soils. The use of the generic Topp et al. (1980) relationship may introduce systematic uncertainty depending on soil texture and organic matter content, the latter of which can be particularly high in forest soils. Please also clarify whether the sensors were installed through the litter layer, such that the measurements represent soil moisture in the underlying mineral soil rather than a mixture of mineral soil and organic litter.
L229: A better formulation would be: “CV is a normalized measure of relative variability, whereas the standard deviation is expressed in the original units and is therefore dependent on the magnitude of the mean.”
L242–245: The definition of the two hydrological cycles is based on observed minima and maxima. Please clarify whether the selected cycles are comparable between plots despite their different moisture regimes.
L275: Please clarify whether the North et al. criterion is appropriate when the number of independent spatial observations is estimated using Moran’s I and the observations are strongly spatially correlated.
L302: Please provide justification for the selected ITS threshold of 0.15 and discuss whether the results are sensitive to this threshold.
L399–425: Please be careful when interpreting positive and negative EOF amplitudes as increased or decreased soil moisture. EOF signs are arbitrary, and the physical interpretation should be based on the corresponding reconstructed fields.
L476-477: The statement that wet and dry locations are randomly distributed should be supported by an explicit spatial statistical test if the authors intend “random” in a statistical sense. Otherwise, use “no clear spatial relationship was apparent”.
L623-625: The mechanisms involving rooting strategies, throughfall and interception are plausible but not directly measured. Please consistently use language such as “may”, “could”, or “we hypothesize”.
L663–680: This section contains several interesting hypotheses concerning the controls of EOF1. However, because soil properties, root density, throughfall and canopy structure were not quantitatively incorporated into the analysis, these interpretations should be clearly distinguished from demonstrated relationships.
Conclusions: I recommend ending the manuscript with a short outlook identifying the measurements needed to disentangle vegetation, soil and topographic controls—for example root distribution, throughfall/stemflow, canopy structure, soil hydraulic properties and topographic metrics—and explaining how the existing sensor network could support such analyses.