Agricultural plastic pollution reduces soil function even under best management practices
Data files
Oct 16, 2024 version files 44.62 GB
-
CDFA_Metadata_all.xlsx
34.37 KB
-
ENVI_DAT_File1.zip
10.45 GB
-
ENVI_DAT_File2.zip
9.45 GB
-
ENVI_DAT_File3.zip
8.92 GB
-
ENVI_DAT_File4.zip
9.77 GB
-
Microplastic_images.zip
6.04 GB
-
README.md
4.86 KB
Abstract
Soil plastic contamination is considered a threat to environmental health and food security. Plastic films which are widely used as soil mulches are the largest single source of agricultural plastic pollution. Growing evidence indicates that high concentrations of plastic negatively affect critical soil functions. However, the relationships between agricultural plastic accumulation and its biogeochemical consequences in regions with relatively low levels of soil plastic pollution remain poorly characterized. We sampled farms across the California Central Coast (a region of global agricultural importance with extensive plastic mulch-based production) to assess the degree and biogeochemical consequences of plastic pollution in fields subject to ‘best practice’ plastic mulching application and removal practices over multiple years. All farms exhibited surface soil plastic contamination, macroplastic positively correlated with microplastic contamination levels, and macroplastic accumulation was negatively correlated with soil moisture, microbial activity, available phosphate, and soil carbon pool size. These effects occurred at less than 10% of the contamination levels reported to degrade field soils, but were relatively subtle, with no detectable relationship to microplastic concentration. Identifying declines in soil quality with low levels of macroplastic fragment accumulation suggests that we must improve best management plasticulture practices to limit the threat to soil health and agricultural productivity of unabated plastic accumulation.
https://doi.org/10.5061/dryad.fn2z34v3q
README: Description of Data and File Structure
Overview
This dataset includes experimental parameters, analysis scripts, images, and processed data files for studying the relationship between macroplastic and microplastic accumulation in soil, along with their effects on soil properties. Below is a detailed description of the files and their contents.
1. CDFA_Metadata_all.xlsx
- This Excel file contains all the experimental parameters, including response and predictor variables from 12 fields nested in 5 farms. For data presentation, the identity of each field and farm has been anonymized.
2. CDFA_final_script.Rmd
- This R Markdown file includes all the R code used for data analysis and generating the plots that represent the final findings of the study.
3. Microplastics_images.zip
- This zipped folder contains a directory named
MP_images
, which includes extracted microplastic images from all fields.
4. ENVI_DAT_File1.zip
- This zipped folder contains
.ENVI
and.dat
files exported from micro-FTIR, used to process the IR spectra for scanned filters from fields a to d.
5. ENVI_DAT_File2.zip
- This zipped folder contains
.ENVI
and.dat
files exported from micro-FTIR, used to process the IR spectra for scanned filters from fields e and f.
6. ENVI_DAT_File3.zip
- This zipped folder contains
.ENVI
and.dat
files exported from micro-FTIR, used to process the IR spectra for scanned filters from fields g to i.
7. ENVI_DAT_File4.zip
- This zipped folder contains
.ENVI
and.dat
files exported from micro-FTIR, used to process the IR spectra for scanned filters from fields j to l.
8. Polymer_characterization_script.Rmd
-
This R Markdown file contains the code used to process the spectra from the
.ENVI
and.dat
files, identifying microplastic polymer types using the OpenSpecy library.Parameters used:
- Signal to noise threshold (
sn_threshold
): Set as the lowest intensity from particles compared to the background. - Correlation threshold (
cor_threshold
): Particles with a correlation < 0.7 to reference spectra were excluded. - Area threshold (
area_threshold
): Artifacts from the background were removed.
The output
.csv
file, namedparticle_details_all
, provides information such asmax_cor_val
- maximum correlation value,polymer_class
- polymer type,area
- particle area,perimeter
- particle perimeter etc. This file was used to extract polymer identity for visually counted particles. Polymers identified (excluding organic matter and minerals) were normalized to the number of visually counted particles on the filters to represent the polymeric composition of the extracted microplastic particles. Any particle with an area ≤ 3750 µm² was excluded to avoid particles smaller than 150 µm. - Signal to noise threshold (
Code/Software
- Data analysis was performed using R (version 4.3.1) and RStudio (version 2023.09.0-463) with the following packages:
lme4
andlmerTest
for statistical modelingggplot2
for generating plots
Statistical Modeling
- The relationship between macroplastic and microplastic accumulation in soil was tested using a linear mixed-effect model, with ‘field’ nested within ‘farm’ treated as a random effect using R script named as
CDFA_final_script.Rmd
. - Macroplastic (count per ha - plastic particle number count per hectare of field, surface area per ha - surface area of plastic per hectare field, mass per ha - mass of plastic per hectare field, and mass per surface area - mass per unit surface area of field) and microplastic (count per kg soil - microplastic particle number count per kg of soil) influence on soil properties were tested with similar random effects.
- Clay content was included as a covariate due to its strong influence on soil biogeochemical properties and retention of microplastics.
- Log transformation was applied to response variables when assumptions of normality were not met (e.g., gravimetric soil moisture, P content - Phosphorous content in mg per kg of soil, POM - particulate organic matter, and MAOM - mineral associated organic matter). Untransformed response variables included TIN - total inorganic nitrogen, soil respiration, microbial biomass, and microplastic count per kg soil, particularly when values fell below detection.
- Predictor variables such as macroplastic count per ha and macroplastic mass/surface area were log-transformed to improve model fit.
- The
ggeffect
function was used to predict response variables with a 95% confidence interval based on the linear mixed-effect model.
Macroplastics and surface soil samples have been collected from 12 fields belongs to 5 different farms. Macro and micro plastics contamination was measured in form of number and mass concentration per ha, surface area per ha, mass per surface area etc., only number concentration (coung per kg soil) was estimated for microplastics from collected soil samples.
Different Biogeochemical properties such as gravimetric moisture (soil_moist), total inorganic nitrogen (TiN), Olsen-P, particulate organic matter (POM), minera associated organic matter (MAOM), soil respiration and microbial biomass was also estimated.
The relationship between macroplastic and microplastic accumulation in soil was tested using a linear mixed effect model, with ‘field’ nested within ‘farm’ treated as a random effect. The influence of macroplastic (count per ha, surface area per ha, mass per ha, mass per surface area) and microplastic (count kg-1 soil) on soil properties was also tested with ‘field’ nested within ‘farm’ treated as a random effect. Because clay content strongly influences soil biogeochemical properties and can increase the retention of microplastics in the soil even after typical mulch removal practices , average clay content at the field level was included as a covariate. Response variables were log transformed when assumptions of normality were not met and the transformation improved the residuals (gravimetric soil moisture, P content, POM and MAOM). Response variables that had values which fell below detection (e.g., ‘0’) were left untransformed (observed for TIN, soil respiration, microbial biomass, and microplastic count kg-1 soil). Predictor variables (macroplastic count per ha, and macroplastic mass/surface area) were rescaled by log transformation to improve model fit. The “ggeffect” function was used to predict the response variable and associated 95% confidence interval based on the “lmer” model to fit the obtained data points. Data analysis was completed using R (version 4.3.1) and RStudio (version 2023.09.0-463) using the packages ‘lme4’ and ‘lmerTest’. All plots were made using the ggplot2 package . Field and farm identity is anonymized for data presentation.