============================================================================================================
START readme
============================================================================================================

This file focuses more on the details of the data package. 


============================================================================================================
## General
============================================================================================================

+ Author(s): Luuk Croijmans, Fogelina Cuperus, Dirk F. van Apeldoorn, Felix J.J.A. Bianchi, Walter A.H. Rossing, Erik H. Poelman
+ Project: Data underlying the publication: Strip cropping designed for maintaining productivity increases ground beetle biodiversity
+ Contact: erik.poelman@wur.nl OR luuk.croijmans@wur.nl


============================================================================================================
## Title
============================================================================================================

Data underlying the publication: Strip cropping designed for maintaining productivity increases ground beetle biodiversity

[ADD DOI Article]


============================================================================================================
## Methods
============================================================================================================

# Abstract

Global biodiversity is declining at an unprecedented rate, with agriculture as a major driver. There is mounting evidence that intercropping can increase insect biodiversity while maintaining or increasing yield. Yet, intercropping is often considered impractical for mechanized farming systems. Strip cropping is pioneered by Dutch farmers as it is compatible with standard farm machinery. Here, we show that strip cropping systems that are designed for retaining productivity, can also enhance insect biodiversity, without incurring major yield loss. Strip cropped fields had on average 15% more ground beetle species and 30% more individuals than monocultural fields. The increase in field-level beetle species richness in organic agriculture through strip cropping approached increases found for other readily deployed biodiversity conservation methods, like shifting from conventional to organic agriculture (+19% - +23%). This makes strip cropping a useful tool for bending the curve of biodiversity loss without compromising food production. 



# Measurements and data collection

Study area
A multi-location study was conducted on four organic farms across the Netherlands. Three experimental farms were managed by Wageningen University & Research (Lelystad, Valthermond, Wageningen) and one commercial farm was managed by Exploitatie Reservegronden Flevoland (ERF B.V.) located in Almere. All four locations contained both strip cropping and monocultural crop fields, but differed in soil type, establishment year of the strip cropping experiment, number of crops grown, length of the crop rotation, number of sampled crops and sampling years, and farm and landscape characteristics such as percentage of on-farm semi-natural habitat (SNH), mean field size, and landscape configuration. The locations Almere and Lelystad were located in a homogeneous, open polder landscape characterized by intensive arable crop production and non-crop habitats consisting of grass margins, tree lines and watercourses. Valthermond was located in an open, reclaimed peat landscape with intensive arable crop production characterized by long and narrow fields separated by grassy margins and ditches and limited areas of woody elements. The site at Wageningen was located in a more complex landscape with smaller field sizes, and non-crop habitat consisting of woodlots, hedgerows, tree lines, ditches and farmyards. 

Experimental lay-out
At Almere and Valthermond the crops in strip cropping were all grown alongside each other, whereas at Lelystad and Wageningen two alternating crops (crop pairs) were grown alongside each other. At each location, strip cropping and monoculture fields were always paired on the same experimental field. At Almere eight different crops were grown in alternating strips of 6 m width, including celeriac (Apium graveolens var. rapaceum), broccoli (Brassica oleracea var. italic), oat (Avena sativa), onion (Allium cepa), parsnip (Pastinaca sativa), faba bean (Vicia faba), potato (Solanum tuberosum) and a mix of ryegrass and white clover referred to as grass-clover (Lolium perenne/Trifolium repens). At Lelystad four different crop pairs were grown in alternating strips of 3 m width, including carrot (Daucus carota subsp. sativus) and onion, white cabbage (Brassica oleracea var. capitata) and wheat (Triticum aestivum), sugar beet (Beta vulgaris) and barley (Hordeum vulgare), and potato and ryegrass. At Valthermond eight different crops were grown in alternating strips of 6 m width, including potato, barley, barley mixed with broad bean (Vicia faba) in 2020, barley mixed with pea (Pisum sativum) in 2021, sweet corn (Zea mays convar. saccharata var. rugosa), sugar beet, common bean (Phaseolus vulgaris), and grass-clover. At Wageningen three different crop pairs were grown in alternating strips of 3 m width, including white cabbage and wheat (2019 – 2021) or oat (2022), barley and pumpkin (Cucurbita maxima, 2020 - 2022) or bare soil due to crop failure (2019), and potato and ryegrass. The crop combinations and neighbors were selected based on literature, expert knowledge and experience of functionality in terms of expected advantages for yield and pest and disease control. Large scale monoculture plots (0.25 ha to 2.30 ha) served as reference, hereafter referred to as monoculture. At Lelystad, Valthermond and Wageningen not each crop grown in strips was present as monoculture in each year, but only those crops for which a monoculture was present were sampled. All fields were managed according to organic regulations, yet at each location fertilization and weed management reflected regional practices and were adjusted to local soil conditions. Flower strips were sown within the experimental fields at Almere and Valthermond.

Sampling 
The ground beetle community was sampled using pitfall traps in all crops for which both a monoculture and strip cropping field were present, at each location and in multiple rounds per year between March and September. The sampled crops, number of rounds and number of pitfalls differed per year and per location. Pitfall traps consisted of a plastic cup (8.5 cm diameter) placed in the soil so the top of the cup was level with the soil surface. Pitfalls were filled with water mixed with non-perfumed soap and covered with a black roof (12.5 cm diameter). Pitfall trap type was similar in all years and locations and traps were placed at the same specific location per year. For all analyses, year series were made, in which all ground beetle catches from the same pitfall trap were pooled per year.


Statistical analyses
We used R, version 4.2.2 for all statistical analyses. 

Effect of crop configuration on field-level richness and activity density
To analyse the difference in species richness and activity density between monocultures and strip cropping configurations at field level, we used rarefaction of samples within the same field. To rarify to an equal sampling intensity, we calculated the average cumulative number of species or individuals within x year series, where x is the largest number of year series available for the crop configuration comparison. Next, we calculated the relative change due to strip cropping by subtracting the number of species or individuals found in the monoculture field from the number found in the strip cropping field, and then dividing the result by the number of species or individuals in the monoculture. This gave the relative change centered around zero, where negative values indicated higher richness in monocultures and positive values higher richness in strip cropping. We then analysed this data using Generalized Linear Mixed Models (GLMM) with a Gaussian distribution, and assessed whether the intercept deviated significantly from zero. As random variables, we used location and year, with year nested in location. We ran these analyses using a dataset that included all comparisons among monocultural fields and strip cropping fields of all locations. To test whether strip cropping was more beneficial for beetle diversity than the most beetle rich monoculture, we separately analysed a dataset that only included the monocultural fields with the highest taxonomic richness or activity density among the constitutive crops of the strip cropping field. Generalized linear models were run using the glmmTMB package and tested for model fit using the DHARMa package.

Effect of crop configuration on crop-level biodiversity
To quantify biodiversity we used five variables: (1) activity density, the total number of ground beetles found per year series; (2) taxonomic richness, the total number of species or genera (lowest taxonomic level available) found per year series; (3) the inverse Simpson index, the inverse of the sum of proportions of different species over the total abundance (Simpson, 1949);  (4) absolute evenness, the number of effective species calculated by dividing the inverse Simpson index by the taxonomic richness (Williams, 1964); and (5) Shannon entropy (Shannon and Weaver, 1949). We chose absolute evenness as our measure for evenness, as this method removes the richness component from the inverse Simpson index and adheres to all requirements for an evenness index (Smith and Wilson, 1996; Tuomisto, 2012). We included both the inverse Simpson index and Shannon entropy as the former is more sensitive to changes in eveness and the latter to species richness (DeJong, 1975). 

To analyze the effect of crop configuration on ground beetle activity density, taxonomic richness, evenness, inverse Simpson index and Shannon entropy we used GLMM. We constructed models for each response variable, using data from all four locations. In these models we included crop configuration (monoculture or strip cropping) as a fixed factor. We included location, year and crop as nested random variables in these models. To quantify and visualize the variation in responses between locations, years and crops, we ran generalized linear models (GLM) with a variable that combined these three variables into one, which was also included as a fixed factor. Here, we also included the interaction between crop configuration and the combined variable for crop, location and year. For the model on activity density we used negative binomial distribution (log link function); for richness, evenness and Shannon entropy we used Gaussian distribution; and for inverse Simpson index we used gamma distribution (inverse link function). All models were tested for model fit using the DHARMa package.

Community composition of crop configurations and crops
To assess whether crop configurations have distinct ground beetle communities, we used permanova with a Hellinger transformation (with 999 permutations). We only used data from locations and years where pitfall catches had been identified to species level (Almere 2021/22, Lelystad and Wageningen). We considered four models, one for all locations combined and one for each location separately. In all models we included crop configuration and the nested variables of location (whenever applicable), year and crop as fixed factors. We also analyzed the interaction between crop configuration and this nested structure of location, year and crop. To visualize crop-configuration-dependent ground beetle communities, we used redundancy analysis (RDA) on Hellinger transformed data. Here, we again conducted four analyses, one for all locations combined and one for each of the three considered locations. We only used crop configuration as a predictor, to force RDA to show any change in ground beetle community associated with crop configuration. As such, only one RDA axis was created per model, which was plotted against the first principal component describing the residual variation. 

Due to the large influence of location and year on ground beetle communities, visualizing any general effects of strip cropping on these communities using all available data was challenging. To address this, we conducted RDA and visualized the effect of crop configuration on a subset of the data from one location and one year. We chose the data from Wageningen in 2021 and 2022 for this analysis because it provided a set-up where strip cropping of two crops could be compared with their constituent monocultures within the same experimental fields. Here, we used both crop configuration, crop and their interaction term as explanatory variables for each crop pair separately. To analyse whether ground beetle communities significantly differed among combinations of crops and crop configurations, we ran pairwised permanova on all three fields and two years separately, using the “pairwiseAdonis” package.  

Crop configuration specific species
To assess whether there are crop-configuration-specific species we used indicator species analyses (ISA) (indicspecies package). Here, we excluded data from Almere in 2020, as they only contained genus-level identification. As this method does not allow for conditioning the data on location, we analyzed the data for the different locations separately. To test whether species mainly occur in a monoculture or in strip cropping, we ran an ISA over the whole dataset per location. ISA uses two indicators: specificity and sensitivity of species to a certain crop configuration. Specificity refers to the predictive value of the species as indicator of the specific crop configuration, whereas sensitivity refers to the percentage of samples of the respective crop configuration that include the species. Both indicators do not take the abundance of species into account (de Cáceres, 2023). 



============================================================================================================
## FolderStructure
============================================================================================================

"There is only one parent directory present containing all files."


============================================================================================================
## FolderContents
============================================================================================================

- parent_folder/
Data, scripts, codebooks




============================================================================================================
## Software
============================================================================================================

SoftwareRequired: 

R, Rstudio, spreadsheet software


OtherSoftwareRequirements: 

Excel


============================================================================================================
## FileFormats
============================================================================================================

.csv / .R


============================================================================================================
## CodeBook
============================================================================================================

Please view GB_4loc_codebook.csv (where this readme file is also found) for documentation of abbreviations, 
column names, datapoints, etc. The file codebook.csv uses the columns:

+ index = a number used to distinguish the different entries.
+ code = the abbreviation, variable / data / column name used.
+ dataset = in which of the six datasets the code is used
+ used = location where the code is used. (filename, foldername, columnname, datapoint, protocol, etc.).
+ meaning = the literal meaning of the code. (e.g., fully written out abbreviation)
+ represents = what the code represents in terms of data or usage. (e.g., units of measurements, 
coding used,more in depth explanation)

Note that this file is ';' delimited. To avoid possible confusion and inconsistencies, sentences within 
cells do not contain reading symbols as comma's or semicolons. When required, separation within sections of
a sentence is made possible using the hashtag symbol (#). Example: The sex of an animal is described as
"m = male pig (boar) # f = female pig (sow)" where the hashtag separates the element in a sentence.


============================================================================================================
## Other
============================================================================================================

[describe any other attention points that will help understandability of your data package; delete this 
explanation]


============================================================================================================
END readme
============================================================================================================
