Complex Drivers of Riparian Soil Oxygen Variability Revealed Using Self-Organizing Maps

doi:10.21203/rs.3.rs-1108620/v1

Download PDF

Research Article

Complex Drivers of Riparian Soil Oxygen Variability Revealed Using Self-Organizing Maps

https://doi.org/10.21203/rs.3.rs-1108620/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

Oxygen (O₂) is a key regulator of soil reduction-oxidation processes and therefore modulates biogeochemical cycles. The difficulties associated with accurately characterizing soil O₂ variability have prompted the use of soil moisture as a proxy for soil O₂, based on the low solubility of O₂ in water. Due to seasonal shifts in soil O₂ depletion mechanisms, the use of soil moisture alone as a proxy measurementcould result in inaccurate O₂estimations. For example, soil O₂ may remain high during cool months when soil respiration rates are low. We analyzed high-frequency sensor data (e.g., soil moisture, temperature, CO₂, O₂) with a machine learning technique, the Self-Organizing Map, to pinpoint suites of soil conditions that are associated with contrasting O₂ regimes. At two low-lying riparian sites in contrasting land use and topographic settings of northern Vermont, we found that soil O₂ levels varied seasonally, and with soil moisture. For example, forty-seven percent of low O₂ levels were associated with cool and wet soil conditions, whereas 32% were associated with warm and dry conditions. Contrastingly, the majority (62%) of high O₂ conditions occurred under warm and dry conditions. High soil moisture levels did not always lead to low O₂, however, as 38% of high O₂values occurred under cool and wet conditions. Our results highlight challenges associated with predicting soil O₂ solely based on soil moisture, as variable combinations of soil and site-specific hydrologic conditions can complicate the relationship between soil water content and O₂. This indicates that process-based ecosystem and denitrification models that rely solely on soil moisture to estimate O₂ availability will, in some cases, need to incorporate other site and climate-specific drivers to accurately predict soil O₂.

Environmental Chemistry

Self-organizing map

clustering algorithm

riparian zone

soil O2

soil moisture

high-frequency data

Oxygen (O₂) is the quintessential electron acceptor and therefore drives biogeochemical cycling on Earth. Its availability within soil pores strongly modulates soil oxidation-reduction (redox) potential, thereby controlling which energy yielding soil biogeochemical reactions proceed (Hefting et al., 2004; Silver, et al., 1999). For example, decreases in soil O₂ reduce redox potential, causing facultative and obligate anaerobic microorganisms to shift their energy-yielding respiration processes to utilize alternative electron acceptors. Oxygen availability therefore impacts the capacity of soils to transform nutrients, such as carbon (C) and nitrogen (N) via processes like heterotrophic aerobic respiration and denitrification, respectively (Liptzin et al., 2011). For example, the transformation of nitrate (NO₃⁻), a soluble form of N that is limiting to primary productivity in freshwater and marine ecosystems, to gaseous forms of N (N₂O and N₂) via denitrification, is a major N removal pathway (Hefting, 2003). This process can reduce N loading to water bodies, but it will not proceed if O₂ is abundant. Oxygen availability therefore controls denitrification rates and efficiency (Groffman et al., 1988; Bouwman et al., 2013), playing a critical role in creating conducive environments for soil N removal. As such, soil O₂ regulates N₂O emissions and N mobility (Groffman et al., 1988), highlighting the importance of accurately characterizing shifts in soil O₂.

Our ability to predict soil O₂ concentrations across spatial and temporal gradients is limited, however. This is a result of the complex network of biotic and abiotic soil factors, as well as climatic conditions, that interact to modulate soil O₂ dynamics and create widespread spatial and temporal soil O₂ variability (Silver et al., 1999). Soil O₂ levels are regulated by the diffusion of O₂ into and displacement of O₂ out of soil pores by water (physical processes), and the consumption of O₂ via soil respiration (a biological process, i.e., aerobic microbial, plant root, and faunal respiration) (Moyano et al., 2013; Neira et al., 2015; Ponnamperuma, 1972). Because O₂ has a very low solubility in water (Moldrup et al., 2000), the presence of water inhibits O₂ diffusion from the atmosphere to soil pores (Skopp et al., 1985). Thus, the combined effects of inhibited O₂ diffusion and displacement, and soil respiration typically result in O₂ depletion if reaeration of soil pores is prevented (Neira et al., 2015; Ponnamperuma, 1972). Furthermore, variability in soil O₂ is difficult to manually monitor in-situ (i.e., using handheld soil probes or gas chromatography), and the collection of high spatial and temporal resolution O₂ data is costly, as it requires soil probes and data logging capabilities.

The challenges associated with measuring gaseous O₂ in soils have led to the use of soil moisture as a proxy measurement for O₂ under the assumption that soil moisture is inversely proportional to O₂ concentration (Heinen, 2005; Ridolfi et al., 2003; Rubol et al., 2013). This assumption has been implemented in many simplified process-based denitrification sub-models embedded in N-cycling and ecosystem models (i.e., those that do not account for microbial processes or gaseous diffusion). Some of these models utilize bivariate nonlinear power functions that are modeled after an inverse relationship between O₂ and soil moisture to predict O₂ depletion based solely on water-filled pore space. Examples include the NEMIS model (Hénault & Germon, 2000), the LEACHMN model (Sogbedji et al., 2001), and the SHETRAN model (Birkinshaw & Ewen, 2000). Simplified process-based denitrification models that exclude direct O₂ measurements have been found to exhibit high sensitivity to formulations that represent soil moisture (Hénault & Germon, 2000), which highlights that the relationship between these variables must be validated and defined using empirical data.

Indeed, due to seasonal shifts in the mechanisms that control soil O₂ depletion (Silver et al., 1999), the use of soil moisture as a proxy measurement for O₂ could result in inaccurate O₂ estimations. For example, soil water inputs (precipitation or groundwater) vary seasonally and are modulated by hydraulic conductivity. Water demand (i.e., vegetation water uptake) also fluctuates seasonally and varies by plant species (Ewe et al., 2007). Furthermore, O₂ depletion by plant and microbial respiration is primarily controlled by soil temperature and soil water content and thus exhibits seasonal fluctuations (Kang et al., 2003; Lavigne et al., 2004; Chen et al., 2010).

Because riparian zones are located at the interface of terrestrial and aquatic ecosystems, they can function as hot spots for anaerobic biogeochemical soil processes (Vidon et al., 2010) and are therefore ideal study systems for soil O₂ dynamics. Due to their unique position on the landscape, riparian soils experience frequent hydrologic changes that alter soil moisture content, which can modify O₂ availability. Changes in soil moisture are triggered by hydrologic fluctuations, and the magnitude of these shifts depend on site-specific riparian zone characteristics, such as topography, proximity to surface and groundwater flows, the size and depth of the upland aquifer, and soil hydraulic properties. Riparian zones can also experience seasonal hydrologic fluctuations resulting from changes in connectivity with the upland aquifer (Vidon & Hill, 2004) and variability in water inputs due to seasonal precipitation patterns. Furthermore, site-specific dominant vegetation types have unique water requirements, which could impact the physical soil wetting process. The diverse potential combinations of soil O₂ drivers suggest that the response of soil O₂ to soil moisture fluctuations is a result of multi-variate interactions that are highly dependent on site-specific soil conditions, seasonal fluctuations in environmental conditions, and ecosystem water and O₂ demands.

Recent advancements in sensor technologies facilitate the simultaneous collection of multiple soil parameters at high temporal resolution, including O₂ and its relevant controls (e.g., soil moisture, soil temperature, redox, CO₂, precipitation). This enables us to comprehensively assess potential drivers of soil moisture variability and the related O₂ response. While high-frequency data for multiple parameters is advantageous for ecosystem monitoring, it requires the use of tools that are specifically suited to analyze multivariate and nonlinear data. The Kohonen unsupervised self-organizing map (SOM), a type of artificial neural network, is a powerful clustering tool that can reliably analyze such multivariate and nonlinear data (Rivera et al., 2015), making it an ideal approach for detecting patterns in large environmental datasets. The SOM can overcome limitations of traditional statistical methods, as it can tolerate outliers, non-normally distributed, non-continuous data, and multicollinearity (Kundu et al., 2013; Merdun, 2011).

The SOM maps multivariate data to a two-dimensional map/lattice, where similar data points are situated in close proximity. In contrast to other, more traditional clustering algorithms, (e.g., k-means), the SOM approach enables visualization of variables that drive clustering, and thus, is a potentially powerful statistical tool for leveraging the capacity of high frequency sensor networks to monitor physical and biogeochemical parameters. SOMs have been successfully applied to resolve spatial and temporal heterogeneity in complex systems within large soil and water quality databases (Obach et al., 2001; Wu et al., 2008; Xiaoyong et al., 2019), as well as to classify sediment (Alvarez-guerra et al., 2008) and soil types (Tissari et al., 2007). Additionally, the SOM approach has been utilized to address questions concerning water resources and hydrology, such as rainfall-runoff relationships (Lin & Chen, 2006), precipitation dynamics (Kalteh et al., 2008), and links between physical soil properties and hydrologic soil processes (Merdun, 2011). However, to our knowledge, these tools have not yet been used to detect patterns in high frequency soil sensor time series.

We applied SOMs to test our overarching hypothesis that shifts in riparian soil O₂ levels are predictively driven by combinations of key environmental controls, such as ecosystem water delivery, and water and O₂ demand. We hypothesize that, while these controls are ubiquitous, the way they interact to impact soil O₂ will be modulated by site-specific characteristics and seasonal variability in their relative importance. To test our hypotheses, we used the SOM to identify key drivers of variability in riparian soil O₂ dynamics and combinations of conditions that lead to low and high O₂ levels. We clustered high frequency soil and meteorological data collected over three years from a poorly drained wetland position within two riparian sites located in northeastern Vermont, USA. We studied two riparian soil environments with contrasting site characteristics (e.g., adjacent land use, vegetative cover, site elevation), allowing us to test our hypotheses in two different riparian lowland settings.

2.1 Study sites

To investigate and better characterize O₂ variability within riparian soils, we studied two riparian soil transects with contrasting catchment characteristics. Both transects are located within Lake Champlain’s Missisquoi Watershed in Vermont, USA (Figure 1) and are part of a larger soil monitoring network. This high frequency soil sensor network continuously measures physical and chemical soil conditions 15 cm below the soil surface along a gradient of landscape positions (i.e., spanning upland, wetland, and near-stream locations). Included in this study are data collected from one low-lying, poorly drained position within each transect, where soil O₂ concentrations ranged from anoxic to near atmospheric. One transect is situated within the Champlain Valley physiographic province (hereafter referred to as “CV” site), a primarily agricultural catchment in Sheldon, VT. The other transect is located within a 95% forested catchment with minimal anthropogenic impact that is located within the Northern Green Mountain physiographic province (hereafter referred to as “GM” site), approximately 7 km north of the town of Montgomery, VT (Landsman-Gerjoi et al., 2020).

The elevation range of the CV site spans 101 to 106 m above sea level. Vegetation at this site includes American beech (Fagus grandifolia) trees, and various fern species (group Pteridophytes) and nettles (Urtica dioica) (Landsman-Gerjoi et al., 2020). Soil types at the CV site include Inceptisols (Aquic Dystric Eutrudepts) and Entisols from glaciolacustrine and glaciofluvial fluvial material (i.e., Aquic Udipsamments and Fluvaquentic Dystrudepts) (Ross 2019; Soil Survey Staff, 2019). The GM site is higher in elevation (350-365 m above sea level) and its vegetation is characteristic of a secondary growth northern hardwood forest, including sugar maple (Acer saccharum), yellow birch (Betula alleghaniensis), white ash (Fraxinus Americana) and red spruce (Picea rubens) (Landsman-Gerjoi et al., 2020). Soils at the GM site are Inceptisols (i.e., Fluvaquentic Dystrudepts and Fluvaquentic Endoaquepts) (Ross, 2019; Soil Survey Staff, 2019). Both study sites experience a temperate climate with four distinct seasons, including snow-dominated winters (22% and 24% of annual precipitation at CV and GM site, respectively), a snow melt period, temperate summers with occasional rains, and a fall season with high litter input (Landsman-Gerjoi et al., 2020). Differences in elevation result in contrasting meteorological conditions between the two sites (Table 1).

Groundwater levels at the CV sampling location generally decrease during the growing season (early May - early October; average = 0.82 m below the soil surface) and increase during cooler months (average = 0.38 m below the soil surface). Contrastingly, average groundwater levels at the GM sampling location are significantly higher (p<0.001; average = 0.05 m below the soil surface) and remain elevated throughout the year. The combination of distinctive elevation, atmospheric characteristics (Table 1), and groundwater characteristics between sites could alter water inputs, water demand, and O₂ demand in ways that result in contrasting O₂ dynamics.

Table 1

Meteorological conditions observed throughout the sampling period at the Champlain Valley (CV) and Green Mountains (GM) site, separated by meteorological season. The data were collected from meteorological stations installed at each site. PAR stands for photosynthetically active radiation
Season	Months	Site	Cumulative precip. (cm)	Mean Ambient T (°C)	Mean PAR (uE)	Mean Solar radiation (W/m²)	Mean Relative humidity (%)
Winter	Dec. – Feb.	CV	15.06	-5.93	265.53	162.91	80.50
		GM	18.34	-6.48	46.89	22.64	85.44
Spring	Mar. – May	CV	23.52	5.23	326.26	256.78	75.12
		GM	27.86	4.00	196.72	100.30	75.78
Summer	Jun. – Aug.	CV	25.65	19.66	345.58	305.67	79.86
		GM	31.82	18.02	241.16	137.78	84.25
Fall	Sept. – Nov.	CV	33.55	7.87	276.45	201.72	83.80
		GM	38.16	6.63	71.47	40.50	88.19

2.2 Soil monitoring network

We assessed ecosystem water delivery by measuring soil volumetric water content (VWC) and precipitation. Water demand (i.e., vegetation water uptake) was assessed by measuring ambient temperature (TA), and O₂ demand was measured via soil carbon dioxide (CO₂). Soil volumetric water content (VWC), temperature (T), and electrical conductivity (EC) were monitored at 15-minute intervals using 5TE sensors (Meter Group, Pullman, WA). Soil CO₂ was measured using GMT221 sensors (Vaisala, Helsinki, Finland), and soil O₂ was monitored using a Soil Response O₂ sensor (Apogee instruments, Logan, UT) (15-minute intervals). Each site (CV and GM) was equipped with a meteorological station that measured TA, precipitation, photosynthetically active radiation (PAR), solar radiation, relative humidity, and dew point at 5-minute intervals. Data included in this study were collected from July 2017 to June 2020.

Data analysis: Approach

To identify suites of soil conditions (clusters) associated with various soil O₂ regimes, we employed a Kohonen unsupervised Self-Organizing Map (SOM) approach using the kohonen package in R (Wehrens & Buydens, 2007). We used an “unsupervised approach”, meaning we fed independent variables to the model and excluded the response variable, O₂. Furthermore, we did not constrain the number of outcome clusters (i.e., soil condition descriptors), so that we could empirically determine the most suitable number of clusters for our dataset. We first used exploratory data analysis to identify independent variables with a potential to be linked to O₂ variability within our sites, and to examine meaningful ranges of O₂ as the response variable. We also selected input variables by examining component planes generated by early iterations of the SOM, which allow for the visualization of clustering according to each independent variable. We then ran an SOM analysis approach, which involved several iterative steps, to optimize SOM execution and validate clusters, as described below. The SOM mapped our multivariate dataset to a two-dimensional map/lattice, where observations linked to similar combinations of values for input variables were situated in close proximity. Finally, we compared O₂ values across clusters, post hoc, to better understand drivers of specific O₂ ranges.

2.3 Data analysis: Overview of Kohonen Unsupervised Self-Organizing Map

A detailed description of the SOM algorithm can be found in Kohonen (2013) and Underwood and others (2021). To summarize, the method clusters multivariate observations onto a reduced-dimension lattice. Each lattice node is first assigned a vector of random values (weights) ranging from 0 to 1. The length of this vector is equal to the number of input variables in each observation. A single vector of input values (observed data) is simultaneously presented to each node’s weight vector. The vector of input values is compared to the weight vector using the Euclidian distance formula, and the lattice node with the closest matching weight vector is designated as the best matching unit (BMU). The weight vector of the BMU, in addition to nodes surrounding the BMU (defined as a “neighborhood”), are updated to more closely resemble the input vector. The neighborhood function is unique to the SOM, as other clustering tools (e.g., K-means) only update the weight vector of a single node (Merdun, 2011). The user customizes the learning rate, α, which controls the amount by which weights are adjusted for both the BMU and nodes within the neighborhood around the BMU. This process is repeated until all observations have been presented to the lattice, which constitutes one iteration of the SOM algorithm. Both the learning rate, α, and the neighborhood size are decreased as subsequent iterations of the algorithm are executed, by relying on user-defined functions. The size of the neighborhood is eventually reduced to one node – the BMU. Multiple iterations are executed until the algorithm converges. Once the algorithm converges, the adjusted weight vectors will have self-organized across the lattice such that similar observations will be aggregated together. To define clusters of observations (i.e., nodes of the lattice containing similar weight vectors), the distance between weight vectors is calculated using a hierarchical agglomerative clustering method (Underwood et al., 2021).

2.4 Data analysis: Initial data conditioning, exploratory analysis, and selection of independent variables

To prepare our high frequency soil and meteorological observation data for input to the SOM, data conditioning steps were required. We first replaced missing observations (except for precipitation) with the overall median value for each independent variable using the impute function in the R package Hmisc (Harrell et al., 2020). We then reduced the volume of data by downscaling our 15-minute observations to an hourly frequency using hourly median values for each variable. We derived additional time series to potentially include in our dataset by computing rolling averages (for VWC, soil T, and ambient T) and rolling sum (for precipitation) ranging from 12 hours to 2 weeks prior to each observation to investigate the effects of antecedent soil and meteorological conditions on soil O₂ dynamics. We then performed Principal Component Analyses (PCA), a dimensionality reduction tool, on a correlation matrix of all available independent variables to confirm that the O₂ data clustered distinctly, based on the available parameters. PCA analyses were also used to establish the dimensions of the SOM best suited to our particular data set (Underwood, et al., 2021).

We selected a suite of independent variables to include in our final SOMs by first running provisional SOMs with all available independent variables. We observed the resulting component planes and plotted the distribution of each variable, separated by cluster, using box and whisker plots. These component planes (Figure s-1) and box and whisker plots were visually inspected to exclude variables from the input dataset, such as wind speed, relative humidity, and EC, that lacked distinct variability across clusters.

2.5 Data analysis: SOM data preprocessing

We prepared the input data for the SOM by normalizing each independent variable to a value between 0 and 1 using a range normalization technique. Normalization improves model performance by ensuring different measurement units and magnitudes do not influence the weight of each independent variable. In particular, range normalization has been found to result in optimal SOM performance, as it minimizes topographic error and quantization error (Alvarez-guerra et al., 2008). The response variable, O₂, was not presented as an input to the SOM. However, we did examine O₂ data during data preprocessing to subdivide the associated multivariate timeseries observations into sets associated with distinct ranges of O₂ values, using the Jenks Natural Breaks optimization method via the BAMMtools package in R (Rabosky et al., 2014). We ran separate SOMs for independent multivariate data observations associated with high and low O₂ for a more refined assessment of factors associated with O₂ dynamics, and to ensure that factors associated with high O₂ conditions could be parsed from factors associated with low O₂ conditions.

2.6 Data analysis: SOM computation and model optimization

After selecting a suite of independent variables to include in each dataset (CV high O₂, CV low O₂, GM), we adjusted the number of lattice nodes, the lattice dimensions, and the value of k to maximize between-cluster variance and minimize within-cluster variance. It is important to optimize the dimensions of the SOM lattice, as an unsuitable lattice configuration can distort the distribution of the input variables across the map. We followed Vesanto’s rule (Vesanto et al., 2000) to pinpoint the optimal number of lattice nodes. For each dataset, we approximated the column-to-row ratio of the lattice as the ratio of the two largest eigenvalues from PCA using the original, non-normalized values (Cereghino & Park, 2009). A hexagonal lattice arrangement was used. The lattice for the CV high O₂, CV low O₂, and GM SOM contained 17 rows and 34 columns, 13 rows and 33 columns, and 15 rows and 27 columns, respectively. The CV high O₂, CV low O₂, and GM SOM lattice configuration contained 578, 429, and 405 nodes, respectively.

SOM training was performed over 20,000 iterations, and α was set to decrease linearly from 0.05 to 0.01. The neighborhood size started at a radius from the BMU of two-thirds the lattice size and decreased linearly to a value of 0, at which point only the weights of the BMU were being updated. For a given dataset, the SOM iteration that maximized the nonparametric F statistic (ratio of within-cluster to between-cluster variance) and minimized quantization error (QE) (measure of map resolution) was selected as the final model run (Underwood et al., 2021). The nonparametric F-statistic was calculated using the adonis function in R’s Vegan package (Oksanen et al., 2019). The F-statistic values for the CV high O₂, CV low O₂, and GM SOM were 30891.5, 31395.1, and 32232.6, respectively. Quantization error values for the CV high O₂, CV low O₂, and GM SOM were 0.000505, 0.000305, and 0.000337, respectively.

For a given dataset, to identify unique attributes of each cluster, we plotted the (range normalized) intra-cluster mean, relative to the overall mean, for each input variable. This metric was used to identify suites of variables and their value ranges that constituted different classes of environmental conditions associated with various O₂ levels. Kruskal-Wallis tests and post hoc Dunn’s tests (with a Bonferroni adjustment) were completed to compare inter-cluster medians of original (i.e., not range-normalized) values of each input variable across clusters. The cluster assignments for each observation were then plotted onto an O₂ time series figure (original values) to display temporal fluctuations in classes of environmental conditions.

3.1 Champlain Valley site

For the CV dataset, we performed separate SOMs for observations associated with high (12.9%-21.5%, n = 12,593) and low (0-4.3%, n = 7043) O₂. The SOMs (i) identified sets of input variables that accounted for variability within the dataset and (ii) classified suites of conditions that led to different O₂ regimes. After visually inspecting component planes of all available input variables, key input variables selected for the CV high O₂ SOM (mean O₂ = 18%) were soil T (mean = 10.3°C), 2-week cumulative antecedent precipitation (hereafter referred to as “2-week precipitation”) (2.74 cm), VWC (0.46 m³/m³), CO₂ (3215.2 ppm), and Julian date (192) (Figure s-1). Four clusters maximized the nonparametric F statistic and minimized quantization error.

3.2 Champlain Valley site: High O₂ SOM

The High O₂ SOM identified four distinct clusters of multivariate time series observations (Figure 2c). The clustering analysis grouped CV high O₂ data points into two main categories: warm and dry (i.e., above average soil T and below average VWC), and cool and wet (below average soil T and above average VWC). Two-week precipitation and CO₂ values drove sub-clustering within the warm and dry category, while Julian date, along with two-week precipitation, drove sub-clustering within the cool and wet observations. Oxygen at the CV site ranged from 0 to 21.5%. High O₂ events, which made up 49% of the CV O₂ values, were somewhat evenly distributed among winter (26% of data points) spring (17%), summer (25%), and fall (32%). Oxygen values were consistently high from May 2018 through April 2019 (Figure 2b), and further research is required to identify what prevented O₂ depletion under cool and wet soil conditions during this period.

Soil conditions within clusters 1 and 4, which made up 62% of the CV high O₂ data points, were generally warm and dry (Table 2). Cluster 1, which includes 53% of data points, is associated with average (compared to the overall mean) 2-week precipitation and average CO₂ values. Contrastingly, cluster 4, which encompassed 8.6% of data points and had below average 2-week precipitation, was associated with the highest average CO₂ of all four clusters. There was overlap of the Julian date ranges of clusters 1 and 4 (Table 2). Cool and wet soil conditions (clusters 2 and 3) described 38% of data points within the CV high O₂ dataset and were associated with the highest average O₂ of all clusters. Key differences between clusters 2 and 3 include Julian date range (Oct.-Jan., and Jan.-Apr., respectively) and 2-week precipitation (39.9 cm and 19.2 cm, respectively) (Table 2).

Table 2

Mean value of each input variable and O2 across four clusters identified by the Champlain Valley high O2 SOM. Within the same column, different letters represent significant differences between clusters (p<0.001)
Cluster	O₂ (%)	Soil temp. (°C)	VWC (m³/m³)	CO₂ (ppm)	2-week cumulative antecedent precip. (cm)	Conditions	Julian date ranges
1	17.8a	15.2a	0.4a	3570.4a	30.2a	Warm/dry	Apr. – Oct.
2	18.2b	3.2b	0.6b	493.5b	39.9b	Cool/wet	Oct. – Jan.
3	18.6c	1.3c	0.6c	525.7b	19.2c	Cool/wet	Jan. – Apr.
4	16.8d	17.7d	0.4a	13077.4c	15.6d	Warm/dry	Jul. – Sept.

3.3 Champlain Valley site: Low O₂ SOM

Low O₂ values were distributed somewhat uniformly among the four seasons (winter = 32%, spring = 35%, summer = 12%, and fall = 21%). Low O₂ values represented 27% of the O₂ dataset and occurred intermittently throughout fall 2017 and winter 2018. Low O₂ values were observed consistently from fall (September, October November) 2019 through spring (March, April, May) 2020. There were consistent periods of low O₂ during the spring snowmelt period (April-May) of 2018, 2019, and 2020 (Figure 3a).

Based on a visual inspection of the SOM component planes (Figure s-2), we included the same suite of input variables in the CV low O₂ SOM as the CV high O₂ SOM (soil T, 2-week precipitation, VWC, CO₂, and Julian date). Average values for these variables were 5.8°C, 3.52 cm, 0.57 m³/m³, 1635.2 ppm, and 175.2, respectively. Median values of each input variable were significantly different between the CV high and low O₂ datasets (p <0.001). The SOM identified 5 distinct clusters for the CV low O₂ SOM: like the CV high O₂ dataset, observations fell into warm and dry, or cool and wet categories (Table 3). The majority of low O₂ data points (69%) could be categorized as cool and wet (clusters 1 and 3), in contrast to results for the high O₂ SOM. Data points within clusters 1 and 3 had below average 2-week precipitation and CO₂, and they differed in Julian date ranges (Figure 3c). Clusters 4 and 5 included 9.3% of data points, which fell into a warm and dry category, with above average CO₂. Two-week precipitation differed between these two clusters (Figure 3c), as cluster 4 had average, and cluster 5 had above average 2-week precipitation. Unique to the CV low O₂ SOM, an additional cluster was identified (cluster 2), which encompassed 21.5% of data points. Cluster 2 occurred throughout October and November, and May and June, and can be described as warmer than average, with average soil moisture, despite above average antecedent precipitation and below average CO₂ (Table 3).

Table 3

Mean value of each input variable and O2 across five clusters identified by the Champlain Valley low O2 SOM. Within the same column, different letters represent significant differences between clusters (p<0.001)
Cluster	O₂ (%)	Soil temp. (°C)	VWC (m³/m³)	CO₂ (ppm)	2-week cumulative antecedent precip. (cm)	Conditions	Julian date ranges
1	1.1a	2.8a	0.59a	611.0a	22.6a	Cool/wet	Nov. – Jan.
2	0.8b	10.6b	0.58b	864.0b	64.5b	Warm/high precip.	Oct. – Nov. & May – Jun.
3	0.5c	2.9c	0.60c	470.4c	27.1c	Cool/wet	Jan. – May
4	1.4d	15.3d	0.42d	4247.5d	33.3d	Warm/dry	Jun.
5	3.5e	17.6d	0.37d	18378.9e	42.5e	Warm/dry	Jun. – Jul.

3.4 Green Mountains site SOM

We included all values from the GM dataset in one SOM, because O₂ values at the GM site ranged from 0 to 0.6% (n=6921). Based on a visual inspection of the component planes (Figure s-3), a unique set of input variables was chosen for the GM model: cumulative 2-week antecedent ambient temperature (hereafter referred to as 2-week TA), VWC, 1-week cumulative antecedent precipitation (hereafter referred to as 1-week precipitation), and Julian date. Average values for these variables were 11.0°C, 0.53 m³/m³, 1.99 cm, and 186.6, respectively. The low O₂ SOM grouped the observations into 5 different clusters. Similar to results of the CV low O₂ SOM, data points within the GM dataset can be described as warm and dry, cool and wet, or warm and wet (Table 4).

Table 4

Inter-cluster means of each input variable and O2 across five clusters identified by the Green Mountains SOM. Within the same column, different letters represent significant differences between clusters (p<0.001)
Cluster	O₂ (%)	2-week TA (°C)	VWC (m³ /m³)	1-week cumulative antecedent precip. (cm)	Conditions	Julian date ranges
1	0.02a	-5.4a	0.5a	6.2a	Cool/wet	Oct. – Dec.
2	0b	17.1b	0.5b	15.5b	Warm/dry	May – Sept.
3	0b	17.3c	0.5c	44.2c	Warm/wet	May – Oct.
4	0b	-9.9d	0.5d	8.6d	Cool/wet	Jan. – Feb.
5	0b	18.1c	0.6e	8.7d	Warm/wet	May – Jul.

Encompassing 25% of data points, clusters 1 and 4 can be described as cool and wet, with differing Julian date ranges and VWC (Figure 4c). Clusters 3 and 5 (warm and wet) included 28% of data points, which can be summarized as warm and wet conditions. However, they differed in terms of 1-week precipitation and Julian date, which implies increased soil moisture was caused by differing mechanisms (Figure 4b and c). Cluster 2 encompassed 46.8% of the data points, which can be described as warm and dry, with below average 1-week antecedent precipitation and average Julian date.

Oxygen values included in the GM SOM remained consistently at or below zero percent throughout the entire sampling period, with the exception of an O₂ event in late December 2017 that reached 0.6% (Figure 4c). Clusters 1 and 4 occurred throughout winter 2017/2018 and were associated with cool and wet soil conditions. Due to the interference of winter weather with our instrumentation, we were not able to monitor winter 2018/2019 or 2019/2020. Clusters 2 (warm/dry), and 3 (warm/wet) occurred intermittently from May to October 2018, and from June to September in 2019. O₂ values within cluster 5 (warm/wet) occurred as isolated events each summer during the months of July (2018 and 2019) and May and June 2020.

4.1 SOM results confirm multivariate and complex controls on O₂

Using a unique combination of high frequency, multiparameter soil sensor data collected over multiple years, coupled with machine learning methods, we were able to pinpoint key drivers of riparian soil O₂ variability and identify the complex combinations of variables that control soil O₂ levels. Importantly, and in contrast to traditional ecological assumptions, low O₂ levels did not correspond solely with increasing soil moisture. Indeed, high O₂ levels persisted in both high and low moisture conditions. As hypothesized, low O₂ conditions instead depended on temporally varying combinations of water inputs, water demand, and O₂ demand. As a result, the output of our unsupervised clustering analysis from both field sites (i.e., all three SOMs) could be placed into general categories of “warm and dry”, “cool and wet”, and “warm and wet”. Within those distinct categories, antecedent precipitation, soil CO₂, and Julian date (i.e., season) drove further clustering.

At the CV site, the majority (69%) of low O₂ values were associated with cool and wet soil conditions (i.e., below average soil T, above average VWC) and occurred from November to May (Figure 3). Under these conditions, a physical saturation process (as opposed to biological O₂ consumption) dominated, as the combined effects of decreased water demand (Sevanto et al., 2006) and sufficient water inputs from precipitation likely prevented the reaeration of soil pores (Neira et al., 2015). Furthermore, in Northern temperate climates, ground frost can persist during this period (November -May), especially in areas of open land without significant canopy cover (Shanley & Chalmers, 1999), which could have precluded the reaeration of soil pores. All low O₂ values at the CV site that were observed under cool and wet conditions were associated with below average subsurface CO₂ (used here as indicator of aerobic soil respiration), indicating that O₂ consumption rates were relatively low under the majority (69%) of low O₂ conditions. This finding is consistent with those of Davidson and others (1998) and Moyano and others (2013) who observed decreased soil respiration rates under cool and wet soil conditions during the non-growing season, which resulted from reduced plant respiration and high soil moisture levels impeding O₂ diffusion, and thus, decomposition and CO₂ production (Doran et al., 1990; Moyano et al., 2013; Skopp et al., 1985). It is therefore possible that a restriction of air exchange between the atmosphere and soil pores is necessary in order for low levels of biological soil respiration to markedly deplete O₂ before it is replenished. We note that the effects of subtle shifts in CO₂ on O₂ depletion under cool and wet soil conditions may have occurred at time scales that were finer in resolution than our hourly SOM input data. Therefore, analyzing a dataset of finer temporal resolution, or one that encompasses a shorter time period, may help detect a more significant impact of biological O₂ consumption under cool and wet conditions.

Contrastingly, most (62%) high O₂ values from the CV site were associated with warm and dry conditions that occurred during warmer months (Apr.-Oct., Figure 2). This is consistent with our hypothesis that warm and dry soil conditions would inhibit soil O₂ depletion by allowing O₂ to readily exchange with the atmosphere via increased air-filled pore space. However, as 62% of high O₂ values were associated with above average soil CO₂ levels (clusters 1 and 4), our findings suggest that water-limitation did not suppress soil respiration, which contradicts our hypothesis. Furthermore, these findings contradict previous studies by Doran and others (1990) and Orchard and Cook (1983) that documented decreased soil respiration rates resulting from elevated soil temperatures and low soil matric potential during warm months. It is therefore likely that, under warm and dry soil conditions at our sites, sufficient soil moisture is required to block O₂ diffusion in order for elevated soil respiration rates to sufficiently deplete O₂. We note that VWC may need to decrease below field capacity (not measured in this study) in order for water limitation to significantly reduce soil respiration rates (Davidson et al., 1998). Additionally, warm and dry soil conditions coincide with the growing season in temperate systems where the combined effect of elevated plant water uptake and negative soil matric potential can lower hydraulic conductivity (i.e., inhibit additional water inputs from percolating through the soil matrix) (Hardie et al., 2012). Under this scenario, additional precipitation inputs may not have resulted in increased VWC and therefore, despite elevated CO₂ levels (i.e., high O₂ demand) at this time, O₂ usually remained near atmospheric levels.

4.2 High soil moisture levels do not always lead to low O₂

Although the majority of high O₂ levels occurred under warm and dry conditions, and most low O₂ levels occurred under cool and wet soil conditions, we also intermittently observed the opposite behavior. These exceptions hold key insights into the important role of antecedent conditions and lag effects in determining soil O₂ regimes. For example, low O₂ events occurred in June and July under warm and dry (below average VWC) soil conditions with above average 2-week antecedent precipitation inputs. This is evidenced by clusters 4 and 5 from the CV low O₂ SOM, which accounted for 9.3% of low O₂ observations (Figure 3). In this case, high antecedent precipitation inputs could have temporarily saturated the soil, thus stimulating soil respiration, while simultaneously blocking O₂ diffusion. This could have triggered a significant O₂ depletion that persisted even after soils dried back down. This scenario is in agreement with the preceding warm and above average 2-week precipitation conditions typical for cluster 2. Similar to our findings, Silver and others (1999) found that forest soil O₂ concentrations were negatively correlated with cumulative rainfall for up to four weeks preceding O₂ measurements. However, clusters 4 and 5 were present only once throughout the entire sampling period (in June-July 2019) and occurred together in quick succession, which indicates that these conditions were unusual for our site, at least within our two study years.

Another example of counterintuitive patterns are occurrences of observed high O₂ levels during fall, winter, and spring months (Oct.-Apr.), when conditions were cool and wet (38.3% of high O₂ data points). Two distinct near-atmospheric O₂ events occurred (one in Dec.-Feb. 2017/2018, the other in Nov.-Apr. 2018/2019) under such conditions (clusters 2 and 3 from CV high O₂ SOM, Figure 3), and the latter event had a relatively long duration. This suggests that the high O₂ levels were not a result of a brief O₂ transition period, but instead reflect the absence of an O₂ depletion mechanism in response to increased soil moisture. Interestingly, these conditions mirror those that resulted in low O₂ levels (Table 3), indicating that decreased soil respiration rates characteristic of clusters 2 and 3 from the high O₂ SOM likely did not prevent O₂ depletion. The only difference between cool and wet conditions that resulted in low O₂ (Table 3) and clusters 2 and 3 from the high O₂ CV SOM (Table 2) were higher 2-week precipitation values associated with low O₂ during January-May, which further emphasizes the important role of antecedent conditions in O₂ depletion. It is also possible that the distinctive O₂ levels resulted from oxygenated subsurface water inputs originating from oxygenated groundwater recharge (Nelson, 2002). Regardless, our results show that very similar soil conditions can result in distinctive O₂ levels. A better understanding of the drivers of soil O₂ is therefore required to investigate such heterogeneity.

Our results also illustrate the utility of high frequency in-situ time series in capturing infrequent and unanticipated events, especially in cases when antecedent conditions may alter the O₂ response. Indeed, intermittent manual sampling campaigns could either miss these events entirely, mischaracterize the commonality of their occurrence, or have limited capacity to identify event drivers. As in-situ sensing networks become more commonplace in soil science research, we expect increased detection of these counterintuitive events. This will ultimately change how we understand the drivers of fluctuating O₂ conditions in the soil environment, and in particular, the role of antecedent conditions.

4.3 Site-specific controls on the drivers of O₂ regimes

We also hypothesized that, while the key controls on soil O₂ are constant across sites, site-specific characteristics modulate the relative rates and impacts of ecosystem water inputs, water demand, and O₂ demand, which could lead to variable O₂ regimes across sites. Indeed, the constant anoxia (O₂ = 0-0.6%) and elevated VWC values (0.48-0.61 m³/m³) observed at the GM site were likely due to unique site-specific features. The combined effects of topography and groundwater hydrology dynamics provided a steady water supply that created consistently saturated soil conditions. This is further evidenced by warm and wet soil conditions unique to the GM site, indicating that soils did not dry out under increased ambient temperatures. This finding is consistent with those of Silver and others (1999), who found that soil O₂ levels were sensitive to hydrologic inputs and were significantly correlated with a topographic gradient spanning ridge, slope, and valley locations. The consistently high water inputs at the GM site generated constant anoxia by preventing the re-aeration of soil pores, and/or displacing O₂. As expected, we observed seasonal fluctuations in key O₂ controls (2-week antecedent ambient T, 1-week antecedent precipitation, VWC), but in contrast to the CV site, this resulted in steadily low O₂ concentrations. These findings highlight a major disconnection between the controls on O₂ and O₂ dynamics. This suggests that a physical soil wetting process is the primary mechanism controlling O₂ dynamics at the GM site, and that the prevention of soil pore reaeration or O₂ diffusion prevails, thus creating a low O₂ environment, regardless of seasonal fluctuations in O₂ controls.

The topography, groundwater hydrology, and vegetation characteristics unique to the CV site resulted in seasonal VWC fluctuations. For example, low VWC values observed throughout the growing season at the CV site were likely the result of depleted groundwater levels in combination with high plant water uptake by abundant sedges and nettles (water demand), and elevated soil respiration rates (O₂ demand). These conditions facilitated O₂ diffusion and reaeration of soil pores, thus restoring soil O₂ to near-atmospheric levels. Under warm and dry conditions, antecedent precipitation plays an important role in O₂ depletion, as soils at the CV site can become too dry to displace O₂ or block O₂ diffusion. The significant impact of antecedent precipitation conditions on soil O₂ is also highlighted by Silver and others (1999), who found soil O₂ levels at ridge locations to be significantly correlated with cumulative 4-week antecedent precipitation.

Low O₂ values (n = 7043) occurred less frequently than high O₂ values (n = 12593) at the CV site, which has important implications for nutrient cycling. This finding suggests that the process of O₂ depletion requires the convergence of a more specific suite of soil conditions than high O₂ levels do. Due to seasonal fluctuations in the controls on soil O₂ at our sites, low O₂ values also occurred less frequently (25% of low O₂ values) during the growing season (generally early May–early Oct. in Vermont), compared to high O₂ values (61%). Nutrient cycling processes that require anaerobic soil conditions or anaerobic microsites, such as denitrification, will not proceed if the soil environment is well aerated (Sexstone et al., 1982). Furthermore, the growing season is a critically important time for nutrient sequestration and transformation within agricultural watersheds (Wang et al., 2014), as fertilizer is generally applied to agricultural fields in early spring. High soil O₂ levels during this critical period for N mobilization could reduce soil denitrification rates, which could have detrimental effects on nearby aquatic ecosystems.

4.4 Complications associated with predicting soil O₂ based solely on soil moisture

The results of our clustering analysis suggest that riparian soil O₂ dynamics are controlled by a network of seasonally variable, rate-dependent, and location-dependent parameters, and as such, the relationship between O₂ and soil moisture is more complex than represented by traditional ecological models. In contrast to traditional ecological thought, high soil moisture does not always result in low O₂ levels, and vice versa. Reliance on these more traditional reduction functions would have yielded contrasting results at our two study sites. If we had predicted soil O₂ levels at the CV site solely based on the commonly assumed negative correlation between moisture and O₂ (i.e., VWC values that typically result in low O₂; VWC= 0.5-0.6 m³/m³), 30.6% of O₂ values would have been incorrectly predicted as low, and 6.7% would have been incorrectly predicted as high (VWC ≤ 0.4 m³/m³). In contrast, consistently high VWC observed at the GM site resulted in consistently low O₂. Therefore, although we did not observe a significant negative correlation between O₂ and soil moisture (data not shown), our predictions based on soil moisture conditions alone would have been reasonably accurate at this site.

Our findings have important implications for nutrient cycling models that rely solely on soil moisture measurements to predict soil O₂, and for empirical studies that make inferences about soil biogeochemical processes based on O₂ estimations (Rubol et al., 2012). Soil O₂ dynamics strongly modulate the rate and efficiency of microbially-mediated soil elemental (e.g., C,N,S) cycling through shifts in redox potential. Incorrect estimations of soil O₂ can therefore result in inaccurate predictions of critical N, C, etc., process rates. For example, much of the literature involving the measurement of soil O₂ and its relationship to soil moisture is within the context of climate-change driven shifts in soil moisture regimes and the subsequent effects on C storage and soil respiration (O’connell et al., 2015; Santiago et al., 2005). These changes in C storage are modulated by confounding effects of seasonally variable soil moisture and temperature, as well as O₂ (Davidson et al., 1998). Critical soil biogeochemical processes not only impact watershed nutrient mobilization and downstream water quality, but also soil greenhouse gas production. It is therefore imperative to continue to improve our understanding of soil O₂ dynamics, as they are likely to increase in complexity as we face complications linked to a changing climate.

Our analysis uniquely incorporated multivariate data of high temporal resolution, which allowed us to investigate and provide new insight about the mechanisms controlling O₂ dynamics within our study sites. However, the limitations of our analysis are highlighted by our spatially constrained dataset, as we included observations from one landscape position within two different riparian soil sites of contrasting adjacent land use. Therefore, our results cannot be directly scaled up to predict O₂ regimes across more expansive ecosystem scales. However, our results provide information about riparian soil O₂ dynamics that can be used for larger scale pattern analysis. As the factors that control soil O₂ were similar across the two sites, the seasonal variability we observed in the key O₂ controls may also apply to other riparian soil environments located in temperate climates. This seasonal O₂ framework could be an effective tool as a first pass prediction of whether O₂ conditions are conducive to aerobic or anaerobic soil processes. However, we also must consider that various site-specific characteristics will likely affect water inputs, water demand, and O₂ demand in ways that uniquely affect O₂ regimes. We posit that a fruitful next step would be to conduct comparable analyses that leverage more spatially expansive soil sensor networks across variable climate, topographic, hydrologic, and geologic riparian soil environments (e.g., NEON, Critical Zone Observatories, LTER) to improve our understanding of the drivers of soil O₂ dynamics and our capacity to systematically model soil O₂ behavior and associated soil biogeochemical cycles.

We used a self-organizing map approach to address the widespread spatial and temporal variability exhibited by riparian soil O₂ levels. Our results show that, in contrast to traditional ecological assumptions, O₂ cannot reliably be predicted solely based on an inverse relationship between O₂ and soil moisture. Soil O₂ is instead controlled by a diverse set of seasonally variable parameters (antecedent precipitation, soil T, VWC, soil CO₂) and location-dependent conditions (topography and groundwater hydrology) that interact to result in a complex and nonlinear relationship between O₂ and soil moisture. Importantly, our results reveal that increases in soil moisture do not always trigger O₂ depletion, indicating that process-based ecosystem and denitrification models that rely on soil moisture alone to estimate soil O₂ availability may over-estimate denitrification, or other anaerobic process rates (e.g., iron reduction or methanogenesis). A more nuanced understanding of soil O₂ dynamics would therefore lead to improved predictions of temporal variability in redox-controlled nutrient cycling processes.

Funding

This work was supported by the National Science Foundation under Vermont EPSCoR Grant No. NSF OIA 1556770, as well as grants EAR 2012123, and EAR 1724171.

Competing interests

The authors have no relevant financial or non-financial interests to disclose.

Author contributions

Conceptualization, and review and editing of the manuscript: all authors. Writing-original draft preparation: Brittany Lancellotti. Methodology: Kristen Underwood and Brittany Lancellotti.

Data availability

The datasets generated and analyzed during the current study are available in the Environmental Data Initiative repository (https://portal.edirepository.org/nis/mapbrowse?scope=edi&identifier=1034&revision=1)

Acknowledgements

This work was supported by the National Science Foundation under Vermont EPSCoR Grant No. NSF OIA 1556770, as well as grants EAR 2012123, and EAR 1724171. We thank Dustin Kincaid and Saul Blocher, as well as the Vermont EPSCoR summer interns, for help installing and maintaining the soil sensor network. We also thank the contributors of Vermont EPSCoR’s Basin Resilience to Extreme Events (BREE) project for their expertise and assistance with this study.

Alvarez-Guerra M, González-Piñuela C, Andrés A, Galán B, Viguri JR (2008) Assessment of self-organizing map artificial neural networks for the classification of sediment quality. Environ Int 34(6):782–790
Birkinshaw SJ, Ewen J (2000) Nitrogen transformation component for SHETRAN catchment nitrate transport modelling. J Hydrol 230(1–2):1–17
Bouwman AF, Beusen AHW, Griffioen J, Van Groenigen JW, Hefting MM, Oenema O, Stehfest E (2013) Global trends and uncertainties in terrestrial denitrification and N2O emissions. Philosophical Transactions of the Royal Society B: Biological Sciences 368(1621):20130112
Céréghino R, Park YS (2009) Review of the self-organizing map (SOM) approach in water resources: commentary. Environmental Modelling Software 24(8):945–947. https://doi.org/10.1016/j.envsoft.2009.01.008
Chen Q, Wang Q, Han X, Wan S, Li L (2010) Temporal and spatial variability and controls of soil respiration in a temperate steppe in northern China. Glob Biogeochem Cycles 24(2). https://doi.org/10.1029/2009GB003538
Davidson EA, Belk E, Boone RD (1998) Soil water content and temperature as independent or confounded factors controlling soil respiration in a temperate mixed hardwood forest. Glob Change Biol 4(2):217–227
Doran JW, Mielke LN, Power JF (1990) Microbial activity as regulated by soil water-filled pore space. In Transactions 14th International Congress of Soil Science, Kyoto, Japan, August 1990 3:94-99
Ewe SM, da Sternberg SL, Childers L D L (2007) Seasonal plant water uptake patterns in the saline southeast Everglades ecotone. Oecologia 152(4):607–616. https://doi.org/10.1007/s00442-007-0699-x
Groffman PM, Tiedje JM, Robertson GP, Christensen S (1988) Denitrification at different temporal and geographical scales: Proximal and distal controls. Advances in nitrogen cycling in agricultural ecosystems 174-192
Hardie MA, Doyle RB, Cotching WE, Mattern K, Lisson S (2012) Influence of antecedent soil moisture on hydraulic conductivity in a series of texture-contrast soils. Hydrol Process 26(20):3079–3091. https://doi.org/10.1002/hyp.8325
Hefting MM, Bobbink R, de Caluwe H (2003) Nitrous oxide emission and denitrification in chronically nitrate-loaded riparian buffer zones. J Environ Qual 32(4):1194–1203
Heinen M (2006) Simplified denitrification models: overview and properties. Geoderma 133(3–4):444–463. https://doi.org/10.1016/j.geoderma.2005.06.010
Hénault C, Germon JC (2000) NEMIS, a predictive model of denitrification on the field scale. Eur J Soil Sci 51(2):257–270
Kalteh AM, Hjorth P, Berndtsson R (2008) Review of the self-organizing map (SOM) approach in water resources: Analysis, modelling and application. Environmental Modelling Software 23(7):835–845. https://doi.org/10.1016/j.envsoft.2007.10.001
Kang S, Doh S, Lee D, Jin VL, Kimball JS (2003) Topographic and climatic controls on soil respiration in six temperate mixed-hardwood forest slopes, Korea. Glob Change Biol 9(10):1427–1437
Kohonen T (2013) Essentials of the self-organizing map. Neural Netw 37:52–65
Kundu P, Debsarkar A, Mukherjee S (2013) Artificial neural network modeling for biological removal of organic carbon and nitrogen from slaughterhouse wastewater in a sequencing batch reactor. Advances in Artificial Neural Systems
Landsman-Gerjoi M, Perdrial JN, Lancellotti B, Seybold E, Schroth AW, Adair C, Wymore A (2020) Measuring the influence of environmental conditions on dissolved organic matter biodegradability and optical properties: a combined field and laboratory study. Biogeochemistry 149(1):37–52. https://doi.org/10.1007/s10533-020-00664-9
Lavigne MB, Foster RJ, Goodine G (2004) Seasonal and annual changes in soil respiration in relation to soil temperature, water potential and trenching. Tree Physiol 24(4):415–424
Lin GF, Chen LH (2006) Identification of homogeneous regions for regional frequency analysis using the self-organizing map. J Hydrol 324(1–4):1–9. https://doi.org/10.1016/j.jhydrol.2005.09.009
Liptzin D, Silver WL, Detto M (2011) Temporal dynamics in soil oxygen and greenhouse gases in two humid tropical forests. Ecosystems 14(2):171–182. https://doi.org/10.1007/s10021-010-9402-x
Merdun H (2011) Self-organizing map artificial neural network application in multidimensional soil data analysis. Neural Comput Appl 20(8):1295–1303. https://doi.org/10.1007/s00521-010-0425-1
Moldrup P, Olesen T, Gamst J, Schjønning P, Yamaguchi T, Rolston DE (2000) Predicting the gas diffusion coefficient in repacked soil water-induced linear reduction model. Soil Sci Soc Am J 64(5):1588–1594
Moyano FE, Manzoni S, Chenu C (2013) Responses of soil heterotrophic respiration to moisture availability: An exploration of processes and models. Soil Biol Biochem 59:72–85. https://doi.org/10.1016/j.soilbio.2013.01.002
Neira J, Ortiz M, Morales L, Acevedo E (2015) Oxygen diffusion in soils: understanding the factors and processes needed for modeling. Chilean journal of agricultural research 75:35–44
Nelson D (2002) Natural variations in the composition of groundwater. Drinking Water Program, Oregon Department of Human Services
O’Connell CS, Ruan L, Silver WL (2018) Drought drives rapid shifts in tropical rainforest soil biogeochemistry and greenhouse gas emissions. Nat Commun 9(1):1–9. https://doi.org/10.1038/s41467-018-03352-3
Obach M, Wagner R, Werner H, Schmidt HH (2001) Modelling population dynamics of aquatic insects with artificial neural networks. Ecol Model 146(1–3):207–217
Oksanen J, Blanchet FG, Friendly M, Kindt R, Legendre P, McGlinn D, Wagner H (2020) vegan: Community Ecology Package. R package version 2.5-6. 2019
Orchard VA, Cook FJ (1983) Relationship between soil respiration and soil moisture. Soil Biol Biochem 15(4):447–453
Ponnamperuma FN (1972) The chemistry of submerged soils. In Advances in agronomy 24:29–96
Rabosky DL, Grundler M, Anderson C, Title P, Shi JJ, Brown JW, Larson JG (2014) BAMM tools: an R package for the analysis of evolutionary dynamics on phylogenetic trees. Methods Ecol Evol 5(7):701–707
Ridolfi L, D'Odorico P, Porporato A, Rodriguez-Iturbe I (2003) Stochastic soil moisture dynamics along a hillslope. J Hydrol 272(1–4):264–275
Rivera D, Sandoval M, Godoy A (2015) Exploring soil databases: a self-organizing map approach. Soil Use Manag 31(1):121–131. https://doi.org/10.1111/sum.12169
Ross DS (2019) Soil classification for BREE sites. Plant and soil science. University of Vermont, Vermont
Rubol S, Manzoni S, Bellin A, Porporato A (2013) Modeling soil moisture and oxygen effects on soil biogeochemical cycles including dissimilatory nitrate reduction to ammonium (DNRA). Adv Water Resour 62:106–124. https://doi.org/10.1016/j.advwatres.2013.09.016
Rubol S, Silver WL, Bellin A (2012) Hydrologic control on redox and nitrogen dynamics in a peatland soil. Sci Total Environ 432:37–46. https://doi.org/10.1016/j.scitotenv.2012.05.073
Santiago LS, Schuur EA, Silvera K (2005) Nutrient cycling and plant–soil feedbacks along a precipitation gradient in lowland Panama. J Trop Ecol 21(4):461–470
https://doi.org/10.1017/S0266467405002464
Sevanto S, Suni T, Pumpanen J, Grönholm T, Kolari P, Nikinmaa E, Vesala T (2006) Wintertime photosynthesis and water uptake in a boreal forest. Tree Physiol 26(6):749–757
Sexstone AJ, Revsbech NP, Parkin TB, Tiedje JM (1985) Direct measurement of oxygen profiles and denitrification rates in soil aggregates. Soil Sci Soc Am J 49:645–651
Shanley JB, Chalmers A (1999) The effect of frozen soil on snowmelt runoff at Sleepers River, Vermont. Hydrol Process 13(12–13):1843–1857
Silver WL, Lugo AE, Keller M (1999) Soil oxygen availability and biogeochemistry along rainfall and topographic gradients in upland wet tropical forest soils. Biogeochemistry 44(3):301–328
Skopp J, Jawson MD, Doran JW (1990) Steady-state aerobic microbial activity as a function of soil water content. Soil Sci Soc Am J 54(6):1619–1625
Sogbedji JM, Van Es HM, Hutson JL (2001) N fate and transport under variable cropping history and fertilizer rate on loamy sand and clay loam soils: I. Calibration of the LEACHMN model. Plant Soil 229(1):57–70
Soil Survey Staff (2019)National Resources Conservation Service United States Department of Agriculture (Web
Soil, Survey)
Tissari S, Nykänen V, Lerssi J, Kolehmainen M (2007) Classification of soil groups using weights-of-evidence-method and RBFLN-neural nets. Nat Resour Res 16(2):159–169
https://doi.org/10.1007/s11053-007-9040-y
Underwood KL, Rizzo DM, Dewoolkar MM, Kline M (2021) Analysis of Reach-scale Sediment Process Domains in Glacially-conditioned Catchments Using Self-Organizing Maps. Geomorphology 382:107684. https://doi.org/10.1016/j.geomorph.2021.107684
Vesanto J, Himberg J, Alhoniemi E, Parhankangas J (2000) SOM toolbox for Matlab 5. Technical Report 57:1-59
Vidon P, Allan C, Burns D, Duval TP, Gurwick N, Inamdar S, Sebestyen S (2010) Hot spots and hot moments in riparian zones: Potential for improved water quality management. J Am Water Resour Assoc 46(2):278–298. https://doi.org/10.1111/j.1752-1688.2010.00420.x
Vidon P, Hill AR (2004) Denitrification and patterns of electron donors and acceptors in eight riparian zones with contrasting hydrogeology. Biogeochemistry 71(2):259–283
Wang Y, Li Y, Liu X, Liu F, Li Y (2014) Relating land use patterns to stream nutrient levels in red soil agricultural catchments in subtropical central China. Environ Sci Pollut Res 10481–10492. https://doi.org/10.1007/s11356-014-2921-9
Wehrens R, Buydens LM (2007) Self-and super-organizing maps in R: the Kohonen package. J Stat Softw 21(5):1–19
Wu T, Chellemi DO, Graham JH, Martin KJ, Rosskopf EN (2008) Comparison of Soil Bacterial Communities Under Diverse Agricultural Land Management and Crop Production Practices. Microb Ecol 55(2):293–310. https://doi.org/10.1007/s00248-007-9276-4
Xiaoyong L, Huan TAO, Xuegang G, You LI (2019) Exploring the database of a soil environmental survey using a geo-self-organizing map:A pilot study. J Geog Sci 29(10):1610–1624

SI.docx

Download PDF

Version 1

posted

You are reading this latest preprint version

Complex Drivers of Riparian Soil Oxygen Variability Revealed Using Self-Organizing Maps

Status:

Version 1

Abstract

Figures

1 Introduction

2 Methods

2.1 Study sites

2.2 Soil monitoring network

2.3 Data analysis: Overview of Kohonen Unsupervised Self-Organizing Map

2.4 Data analysis: Initial data conditioning, exploratory analysis, and selection of independent variables

2.5 Data analysis: SOM data preprocessing

2.6 Data analysis: SOM computation and model optimization

3 Results

3.1 Champlain Valley site

3.2 Champlain Valley site: High O₂ SOM

3.3 Champlain Valley site: Low O₂ SOM

3.4 Green Mountains site SOM

4 Discussion

4.1 SOM results confirm multivariate and complex controls on O₂

4.2 High soil moisture levels do not always lead to low O₂

4.3 Site-specific controls on the drivers of O₂ regimes

4.4 Complications associated with predicting soil O₂ based solely on soil moisture

5 Conclusions

Declarations

References

Supplementary Files

Status:

Version 1

Complex Drivers of Riparian Soil Oxygen Variability Revealed Using Self-Organizing Maps

Status:

Version 1

Abstract

Figures

1 Introduction

2 Methods

2.1 Study sites

2.2 Soil monitoring network

2.3 Data analysis: Overview of Kohonen Unsupervised Self-Organizing Map

2.4 Data analysis: Initial data conditioning, exploratory analysis, and selection of independent variables

2.5 Data analysis: SOM data preprocessing

2.6 Data analysis: SOM computation and model optimization

3 Results

3.1 Champlain Valley site

3.2 Champlain Valley site: High O2 SOM

3.3 Champlain Valley site: Low O2 SOM

3.4 Green Mountains site SOM

4 Discussion

4.1 SOM results confirm multivariate and complex controls on O2

4.2 High soil moisture levels do not always lead to low O2

4.3 Site-specific controls on the drivers of O2 regimes

4.4 Complications associated with predicting soil O2 based solely on soil moisture

5 Conclusions

Declarations

References

Supplementary Files

Status:

Version 1

3.2 Champlain Valley site: High O₂ SOM

3.3 Champlain Valley site: Low O₂ SOM

4.1 SOM results confirm multivariate and complex controls on O₂

4.2 High soil moisture levels do not always lead to low O₂

4.3 Site-specific controls on the drivers of O₂ regimes

4.4 Complications associated with predicting soil O₂ based solely on soil moisture