Assessing the Effect of Undirected Forest Restoration and Flooding on the Soil Quality in an Agricultural Floodplain

This study was conducted in the Saint Michael’s College Natural Area (73°9′1″ W 44°29′33″ N), a 350- acre area of floodplain grasslands, wetlands, and woodlands located along the banks of the Winooski River (Colchester, VT, USA). The study design and site selection were based on visual analysis of 2022, 1962, and 1942 aerial imagery. Two old growth forest zones (5.4 and 6.6 acres) were selected, which were well- developed forest patches in as of 1942. Two new growth forest zones (3.2 and 4.1 acres) were selected, which were agricultural land in 1962, but had developed into forest patches by 2022. Two recently abandoned farm field sites (3.4 and 4.1 acres), now a mixture of young trees and prairie species, were selected near two forest sites.Within each zone, ten sample sites were randomly selected using ArcGIS Pro version 3.0 random points tool, with sites not being permitted within 100 m of each other or along the edge of each zone. Pre-flood soil samples were collected be-tween 12 June 2023 and 7 July 2023. At each site and pre-flood, a one-meter-by-one-meter quadrat was set down and four surface (0–5 cm) and four bottom (25–30 cm) soil samples (one from the center of each quadrant) were taken. Sample depths approximated the surface of the A-horizon (0–5 cm) and the bottom of the A-horizon to the start of the C or B horizons (25–30 cm), based on USGS soil classification for the region. The four samples from each quadrant were then pooled to produce one surface and bottom soil sample per site. Soil moisture, pH, and temperature were taken using handheld meters (Kelway pH and Moisture tester and Taylor soil thermometer). Soil samples were dried in the lab at 60 °C until constant weight. Then, roots and course woody debris were removed, and what remained the remaining material was sieved through a 250 µm mesh to break up soil aggregates, remove rocks, and homogenize the sample. Dried and homogenized samples were further processed for extractable phosphorus (Pext), SOM content, and water- extractable SOM characteristics.After the initial soil samples were collected, a major flood event occurred 10 and 11 July 2023. Within the Winooski River watershed, 4 to 9 inches of rain fell in over 48 h causing the river to exceed major flood stage for over 24 h and crest at an estimated 23.3 feet. The flood of July 2023 submerged the Natural Area under 15 feet of water, and it took 5 to 7 days for flood waters to fully recede. Post-flood 50 of the 60 sites were resampled across the six zones. At each site, four surface soil subsamples were pooled for each site to make one composite sample. Post-flood samples were processed the same as pre-flood samples and analyzed for Pext, SOM content, and water- extractable SOM properties.SOM content (%) was determined by loss on ignition (550 °C for one hour).Extractable phosphorus (Pext) was determined by adding 50 mg of soil into 10 mL of nano-pure water and then digested in the presence of potassium persulfate in the autoclave at 121 °C for 1 h. Samples were then diluted 1 to 10 and measured using the molybdate-blue colorimetric approach using a Seal AQ300 Discrete Analyzer.SOM quality was accessed in surface soils only by extracting 1.0 g of soil in 40.0 mL of nanopure water in a 45 mL centrifuge tube. Pre-flood soils were extracted for 30 min and 24 hours using a soil shaker. Post-flood samples were extracted for 30 min. Extracted SOM was then centrifuged for 3 min (1500 RPM) and the supernatant was filtered using a 0.45 µm membrane attached to a syringe. SOM extracts were then measured for organic matter characteristics using optical chemistry methods. A Horiba Aqualog was used to make light absorbance and fluorescence excitation and emission scans of each sample. Samples were diluted to have light absorbance at 254 nm below 40, which helps eliminant eliminate matrix effects on scans. Sample scans were corrected for instrument bias, inner-filter effects, blank- subtracted, and normalized to Raman Units. Blank scans were conducted at the start and end of each day of scans scanning using nanopure water. The spectral slope ratio (SR), freshness index (BA), modified humification index (HIX), and fluoresce peaks A (terrestrial, humic-like), C (humic-like), D (soil, fulvic-like), E (soil, fulvic-like), M (microbial, humic-like), N (microbial, humic-like), and T (protein-like) were used as indicators of SOM quality. SR is inversely related to the size of the extracted organic matter. BA provides an indication of the level of organic matter degradation. HIX indicates the fraction of humic matter as compared to protein-like or aliphatic organic compounds, where a value of >0.95 would indicates fully humified organic matter.Statistical AnalysisRepeated sampling, permutation analysis of variance (ANOVA), and multivariate analysis of variance (MANOVA) were used to determine univariate (pH, soil moisture, temperature, SOM content, and Pext) and multivariate (SOM quality) differences between land-use zones, flood-period, and soil depth. 30-min and 24-h extracted SOM quality indicators were not used together or compared in the same analysis be-cause extraction time impacted the quality and concentration of extracted material. All analyses were carried out in R using R-Studio using custom functions and the tidyverse version 2.0.0 and vegan version 2.6-10 packages. Repeated measures were selected over data transformation because the assumptions of multivariate normality are challenging to meet, and the resampling approach sets the test population to that of the data. Pre-flood, a full-model linear ANOVA with two factors (soil depth and zone) was used for univariate tests (dependent variable ~ zone * soil depth). Post-flood, only surface soil samples. Post flood were analyzed using a two-factor linear model was used with zone and flood period (pre or post) as the factors (dependent variable ~ zone * flood peri-od). For MANOVA, a Euclidean distance matrix was produced using the dist function after data were scaled. Pre-flood, 24-h extracted SOM properties were compared across zone (single-factor design). Post-flood, 30-min extracted SOM properties were compared using a two-factor model with zone and flood period as factors. For each test, 9999 permutations plus the observed condition were conducted to make generate the test distribution for each comparison. When a significant test was determined, a least squares comparison approach was used to complete a resampling pair-wise comparison for univariate tests and pairwise.adonis2 for multivariate tests. Multivariate comparisons were visualized using principal components analysis (PCA) ordination of the first two components. The prcomp function was used to conduct the PCA, and DOM data were scaled and centered prior to analysis.