Data from: Nitrogen content of herbarium specimens from arable fields and mesic meadows reflect the intensifying agricultural management during the 20th century
Data files
Dec 16, 2024 version files 86.73 MB
-
Kuehn_et_al._Resurvey_-_JEcol_Open_Data.zip
86.72 MB
-
README.md
8.76 KB
Abstract
Arable fields and mesic meadows have been affected by intensifying agricultural management and nutrient input during the 20th century, but direct evidence for the long-term impact of intensification on plant nutrient contents remains scarce. Non-destructive novel spectroscopic methods can produce such data from herbarium specimens, making it possible to investigate how contents of leaf nutrient traits, especially nitrogen and phosphorus, changed over the last century, and what role habitat type and management practices play.
We carried out a resurvey study of functional traits in arable field and mesic meadow communities. We used specimens from two German herbaria with a high coverage of their local floras: the Senckenberg Herbarium Görlitz and the Herbarium Haussknecht in Jena. Following specimen information, the same plant species were resampled in the field in 2022 at the same locations. We employed Near-Infrared Spectroscopy to predict leaf nitrogen, phosphorus, and carbon content of herbarium and field specimens. Nutrient content changes over time were compared with publicly available records of P and N fertilization.
Overall, 1270 specimens of 76 species from both herbarium and field were studied, the oldest from the 19th century. Leaf nitrogen increased significantly through time, with a corresponding increase in leaf nitrogen:phosphorus ratio in both habitats. Leaf phosphorus and carbon content decreased significantly over time, with the latter decreasing significantly stronger in mesic meadows compared to arable fields. The total amount of nitrogen or phosphorus fertilizer applied per year on a regional scale was found to be significantly correlated with the respective leaf nutrient content levels. Mesic meadow species showed a stronger response in leaf nitrogen and leaf phosphorus content over time.
Synthesis: Our study shows a long-term increase of leaf nitrogen in the studied habitats, running in parallel to increased chemical fertilizer application in Germany. Our data indicates a shift from predominantly N-limited towards more P-limited growth conditions. The stronger response of species from mesic meadows compared to species from arable fields could indicate a faster adjustment to environmental pressures. This study thus also serves to showcase the potential of the combination of herbarium collections and NIR spectroscopy.
README: Nitrogen content of herbarium specimens from arable fields and mesic meadows reflect the intensifying agricultural management during the 20th century.
https://doi.org/10.5061/dryad.z34tmpgpw
Description of the data and file structure
“Resurvey_analysis_code.R” is a commentated script for RStudio, which allows the user to load the data, calculate statistics, and create plots.
"01_resurvey_combined_table_cleaned.csv" is a data table containing data for 1270 herbarium and field specimens of 49 plant species. For each specimen, data on the species and sampling date, the locality description, GPS coordinates (if available) and the collection it is stored in are included. Leaf functional trait values are also included: leaf nitrogen content, leaf phosphorus content, leaf carbon content, and the leaf carbon:nitrogen ratio and the leaf nitrogen:phosphorus ratio were all predicted from NIR spectra using PLSR calibration models.
"02_fertilizer_values_historical.csv" consists of anorganic nitrogen and phosphorus fertilization, tracked in N or respectively P kg ha-1 a-1. The data is derived from statistical yearbooks of the German Reich, the German Democratic Republik, and the Federal Republic of Germany, and is collected either on a national or federal state level.
"03_resurvey_NIRS" is a folder containing the calibration datasets (spectral and functional trait data) and the R code needed to create PLSR calibration models as described by Kühn et al. 2024 (doi.org/10.1186/s13007-024-01146-x ) .
In all files, "NA" is used to denote missing values.
"Resurvey_analysis_code.R"
A commentated script for RStudio, which allows the user to load the data, calculate statistics, and create plots.
"01_resurvey_combined_table_cleaned.csv"
A comma-seperated table with 1271 rows, including the header. The 32 columns are as follows:
- sampleID: Unique identifier of each sample
- Area: Defines if the sample came from the region of Jena or Görlitz
- Loc: Specifies if the sample came from the herbaria JE or GLM, or the field
- speciesID: six-letter abbreviation of plant species name
- Species: taxon name with author abbreviation
- Catalogue: catalogue number for GLM specimens
- sampleno: ID number, unique within species and loc_abbrv
- Location: Description of sampling locality in German
- Habitat_type: Mesic meadow or arable field
- Latitude: Latitude in decimal degrees (WGS84)
- Longitude: Longitude in decimal degrees (WGS84)
- Year: Sampling year
- Field_visit: Was the locality from which the specimen is from visited in 2022, Y or N?
- Found_in_field: Was a herbarium specimen precisely relocated in 2022, sampling the same species from the same location?
- spectra_file: Filename for NIRS spectra file
- Loc_abbrv: 4 letter abbreviation for each sampling locality, including herbaria
- Predicted_leafN: Predicted leaf nitrogen content in % (transformed from logleafN_pred)
- Predicted_leafP: Predicted leaf phosphorus content in mg/g (transformed from logleafP_pred)
- leafC_Pred: Predicted leaf carbon content in %
- logleafCN_pred: Predicted leaf carbon:nitrogen ratio
- logleafN_pred: Predicted logarithmic leaf nitrogen content in %
- logleafP_pred: Predicted logarithmic leaf phosphorus content in %
- NP: Predicted leaf nitrogen : phosphorus ratio
- Prediction_present: Are predictions/ .dpt spectral files present for the sample?
- GDR_district: GDR administrative district ("Bezirk") in which sample would be located in, irrespective of sampling year
- Fiscal.Year: Fiscal year from which fertilization values are taken for this sample
- N_kg_ha_a: Amount of nitrogen in kg per hectare per year applied in the GDR district/FRG federal state in sampling year
- P_kg_ha_a: Amount of phosphorus in kg per hectare per year applied in the GDR district/FRG federal state in sampling year
- sampling_period: Past (historical herbarium specimens) or present (2022 herbarium specimens)
- Flowering status: Only for field specimens. What was the latest phenological stage observed at the time of sampling?
- Height: Only for field specimens. What was the maximum vegetative height observed at the time of sampling?
- Functional group: Based on taxonomy, is the species a forb, grass or legume?
"02_fertilizer_values_historical.csv"
A comma-seperated table with 785 rows, including the header. The 5 columns are as follows:
- Fiscal year: the starting year of the fiscal year in which the values were recorded
- GDR_district: The district for which the values were recorded. Includes both the regular districts and “Country_” for nation-wide values.
- Fertilization: the type and measurement unit of mineral fertilizer tracked: either nitrogen or phosphorus
- Value: the amount of fertilizer applied per district per fiscal year in kg per hectare per year
- State: the current-day federal state which the fertilizer value applies to: either Federal for nation-wide measurements, Thuringia or Saxony.
“03_resurvey_NIRS”
This folder contains the relevant files for the Near-Infrared Spectroscopy analysis.
“Spectroscopy_v_6_8.R” is a commentated R-script which can be used to calculate partial least square regression (PLSR) models to calibrate Near-Infrared spectra for leaf functional traits, based on sample NIR data and lab measurements. In the course of the script, data is cleaned, compiled, models are calculated and performance metrics assigned, and models for the prediction of leaf functional trait values based on NIR spectra alone are produced. The PLSR itself is calculated using the “plantspec” package for R.
The subfolders “Spectra_for_Calibration”, “Spectra_for_Calibration_field” and “Spectra_to_be_Predicted” contain spectral files for all samples in this study, encoded as “.dpt” files. Each “.dpt” file is a tab-separated table, with the wavelength in nanometers (nm) in the first column and the reflectance of the sample at that wavelength in the second column. These spectral files have already been spliced and averaged as described in the “Spectroscopy_v_6_8.R” script. Filenames correspond to the “sampleID” column of “01_resurvey_combined_table_cleaned.csv”, consisting of a six-letter species code (e.g. “AchMil” for Achillea millefolium), a sample number, and a four-letter code denoting the locality.
The subfolder “Prediction_models” contains the three partial least squares regression models calculated for the leaf functional traits of leaf carbon content, leaf nitrogen content and leaf phosphorus content, encoded as .RData objects for easy loading into the R programming environment.
“PTrait_reference_mg_g-l.csv” is a comma-separated table consisting of 263 rows and 5 columns with wet laboratory analysis data (nitric acid digestion and ion chromatograph analysis) for leaf phosphorus contents. The columns are as follows:
- SampleID: The unique sample identifier
- P_mg-l: the Phosphorus concentration in mg per liter of solution.
- Weight_g: The weigh of leaf powder used for the nitric acid digestion of each sample.
- Sample volume: The volume which the sample was filled up and diluted to.
- Concentration_mg-g: The concentration of leaf phosphorus in the leaf powder in mg per g.
"Reference_Samples.csv" is a comma-separated table consisting of 4045 rows and 5 columns, giving a reference for each .asd-formatted spectral data file and the corresponding sample ID. The columns are as follows:
FILES_LIST: The spectral data files. Three NIRS readings were taken for each sample.
Sample: The unique sample identifier.
“Trait_reference_field.csv” is a comma-separated table consisting of 260 rows (including header) and 4 columns, giving the wet laboratory analysis data (gas chromatograph analysis) for leaf carbon content, leaf nitrogen content and leaf carbon:nitrogen ratio for each sample. The columns are as follows:
- Sample: The unique sample identifier.
- N_perc: The leaf nitrogen content in percent.
- C_perc: The leaf carbon content in percent.
- C-N: The leaf carbon:nitrogen ratio.
Code/software
The "Resurvey_analysis_code.R" for the analyses is written in R. The RStudio IDE is also required.
The following packages are required for the analyses:
library(tidyverse)
library(magrittr)
library(cowplot)
library(RColorBrewer)
library(readxl)
library(sf)
library(lme4)
library(mgcv)
library(lmerTest)
library(performance)
library(emmeans)
library(gratia)
library(car)
Additionally, for the NIRS calibration script "Spectroscopy_v_6_8.R", the following packages are required:
library(plantspec)
library(Hmisc)
library(foreach)
library(doParallel)
Methods
The dataset consists of metadata, functional trait data, and fertilization data for herbarium specimens and plant specimens sampled in 2022.The herbarium specimens are from the collections of the Herbarium Haussknecht in Jena, Germany (JE), and the Senckenberg Herbarium Görlitz in Görlitz, Germany (GLM). The field specimens were gathered from May to August of 2022, and were press-dried in the same way as the herbarium specimens. Near-Infrared Spectroscopy (NIRS) measurements were taken from both specimen types using an ASD FieldSpec 4 and a contact probe. A subset of the field specimens was selected to create a NIRS calibration dataset, which involved conventional wet lab analyses of the leaf nutrient contents of leaf nitrogen, leaf carbon and leaf phosphorus contents for these specimens. These wet lab measurements were used in conjunction with the NIRS measurements to build partial least squares regression (PLSR) models for each trait. The PLSR models were in turn used to predict the respective trait values from NIRS measurements alone, allowing for the non-destructive functional trait measurements of the other field specimens and the historical herbarium specimens. Because some of the wet lab measurements were log transformed to improve the calibration model accuracy, some of the predicted trait values are log transformed as well. Further details on the calibration workflow and the use of NIRS in herbaria can be found in Proß et al. 2023 (doi.org/10.1111/oik.10255) and Kühn et al. 2024 (doi.org/10.1186/s13007-024-01146-x).