R code, partial dataset and results for: Modelling the impacts of agricultural landscape changes: A bibliometric review

This dataset include 514 scientific papers, which were used to review the litterature on impact modelling of agricultural landscapes. This study was published in Environmental Modelling and Software [Hossard, L., Chopin, P., in press. Modelling agricultural changes and impacts at landscape scale: A bibliometric review, Environmental Modelling and Software]. The literature search was conducted in April 2017 and involved entering keywords in the Clarivate Analytics’ Web of Science without a time frame limitation. The search was limited to the “Article” document type and to the “English” language. For “Topics”, the following search equation was used: “model* AND (agri* or agro* or crop* or farm*) AND (landscape* OR watershed* OR (water NEAR catchment*)) AND (scenar* OR alternative*)”. This initial search yielded 1,975 hits. We then excluded papers based on article abstracts when they did not match our selection criteria, which were: (1) use of a model, i.e. a simplified representation of the system, as a tool to design or assess future agricultural landscape(s), (2) a focus on agricultural systems (including farming practices and/or agricultural organisation, explaining why we chose not to use “land use*” as a key word), (3) resolution at landscape scale (i.e. beyond the farm level) and (4) with outcomes on alternative agricultural systems (thus excluding papers focusing only on the effects of climate change). We also manually excluded general papers lacking a case study application, e.g. reviews without case study (12 papers) (Figure 1). Our final dataset thus comprised 514 individual papers. The json file (EMS_public.jon) presented here, based on a Zotero extraction and excluding proprietary data (e.g., abstracts), includes: article ID, title, journal, page, volume, issue, country, authors and year of publication. Data can be viewed with Firefox, Zotero, or with the R software, using the package jsonlite (Ooms, 2014). The R code, run with R 3.3.3 (Mac), was used to identify the optimal number of topics for our 514 papers dataset, and perform the LDA analyses (number of groups, top words, distribution of topics for each paper, etc.). Note that the full database (including abstracts) is necessary to run this code. Only a partial database is included here, as the full one includes proprietary data. The excel file (all.art.final.topics.Threshold_0.15.xls) displays the results of our LDA analysis, as provided by the R code on the 514 papers database. It includes: the article number (corresponding to the ID of EMS_public.json), the article title, the number of occurences of words related to Topics 2, 1, 4, and 3, the Best topic (considering a dominancy threshold of 0.15), and the year of publication. Reference Ooms, J., 2014. The jsonlite Package: a practical and consistent mapping between JSON Data and R objects. arXiv:1403.2805 [stat.CO] URL https://arxiv.org/abs/1403.2805 (accessed, December 2017).