Process-Based Crop Models in Soil Research: A Bibliometric Analysis

: Different types of soil data are used in process-based crop models as input data. Crop models have a diverse range of applications, and soil research is one of them. This bibliographic analysis was conducted to assess the current literature on soil-related applications of crop models using two widely used crop models: Agricultural Production Systems Simulator (APSIM) and Decision Support System for Agrotechnology Transfer (DSSAT). The publications available in the Scopus database during the 2000–2021 period were assessed. Using 523 publications, a database on the application of process-based crop models in soil research was developed and published in an online repository, which is helpful in determining the speciﬁc application in different geographic locations. Soil-related applications on APSIM and DSSAT models were found in 41 and 43 countries, respectively. It was reported that selected crop models were used in soil water, physical properties, greenhouse gas emissions, N leaching, nutrient dynamics, and other physical and chemical properties related to applications. It can be concluded that a crop model is a promising tool for assessing a diverse range of soil-related processes in different geographic regions.


Introduction
With the availability of fine-scale global data, the utilization of digital techniques has increased globally in environmental research. Soil science, which is a part of the Earth system, and environmental research are traditionally based on observations and analysis of observed data. However, there are different tools, approaches, and computer software that are used to assess different aspects of soil science [1][2][3]. Crop simulation models are one such tool to work out.
Understanding of soil processes and conservation of soil are even highlighted in the United Nation's Sustainable Development Goals (SDGs) due to their importance in the overall sustainability of the globe. For example, SDG 1 (no poverty), GDS2 (zero hunger), SDG3 (good health and well-being), SDG6 (clean water and sanitation), SDG13 (climate action), and SDG15 (life of land) can be achieved through proper understanding and management of soil aspects. Therefore, advanced tools such as crop modeling can be used to understand the soil process that will eventually contribute to the UN SDGs.
Crop models or crop simulation models are tools that use mathematical algorithms to quantify crop growth, development, and yield in interaction with the environment [4].

Data Source and Data
Extensive datasets with an adequate number of records for a certain period are needed for the bibliometric analyses. In order to maintain the quality of data, publications available in the Scopus database (www.scopus.com; accessed on 28 December 2022) were used, and the publication during the 2000-2021 period was considered in this study. The title, abstract, and keywords were evaluated in the "Scopus" search. Accordingly, the following search criteria were used to obtain the publication details.
TITLE-ABS-KEY ((APSIM OR (Agricultural Production Systems Simulator)) AND soil) AND PUBYEAR > 1999 AND PUBYEAR < 2022 TITLE-ABS-KEY ((DSSAT OR (Decision Support System for Agrotechnology Transfer)) AND soil) AND PUBYEAR > 1999 AND PUBYEAR < 2022 TITLE-ABS-KEY ((Aquacrop AND soil) AND PUBYEAR > 1999 AND PUBYEAR < 2022 Then, the relevance of the publication to the topic of interest was manually evaluated, and irrelevant publications were removed. The removed publications include sources that mention soil data as input files. The screened publications were used for further analysis.

Analysis
Several tools and methods are used to perform the bibliometric analysis [30]. Different types of analysis methods were used in this study that covers major categories of bibliometric analysis; performance analysis, science mapping, and network mapping as follows.

Citation-Related Metrics
Normalized citations-Since the publications were distributed throughout the period, the generalization of citations in terms of average and total citations can lead to misleading values. To avoid this issue and to remove the impact of the number of years from the date of publication, normalized citations were calculated as follows [33].
Nomalized citations = Total no. of citations No. of years from the published year (1)

CiteScore
The recent CiteScore available in the Scopus database (the year 2021) was used in this study. The CiteScore of 2021 counts the total citations received for the publications during the 2018-2021 period, which was divided by the number of published documents [34].
Other types of bibliometric analysis performed in this study include bibliographic coupling, network analysis, clustering, co-work analysis, and keyword analysis. Bibliographic coupling occurs when two research documents cite the same reference (already published paper). This was introduced by Kessler [35]. More information on bibliographic coupling can be found in Jarneving et al. [36]. Citation network analysis maps related research as citation practice. It is a review method to quantitatively present the research work. It also provides the citation timeline of the research work too. More information on citation network analysis can be found in McLaren and Bruner [37]. The principle and the analysis procedure of these methods are described in detail by Donthu et al. [30].

Visualization
The network mapping, bibliographic coupling, keyword analysis, and visualizations were performed using VOSviewer version 1.6.18 [32,38,39]. VOSviewer is a software that can be used in visualizing bibliometric networks. It can create bibliometric networks and has advanced layout and clustering techniques and natural language processing techniques. The text mining option is used in co-occurrence networks and keyword analysis [39]. In the visualizations, different clusters were marked in different colors.

Development of a Database
In the selected papers used in the analysis (Section 2.2), the soil-related application, geographic location, and major type of application mentioned in the publications were evaluated as the next step and documented. The dataset was used to develop a database on crop model applications in soil research.

Performance Analysis
The Scopus search found 1446 publications, 660 on APSIM, 299 on AquaCrop, and 487 on the DSSAT model. After the manual screening of publications on applications of soil science, a total of 523 articles were remained (336 APSIM and 187 DSSAT) and used in the analysis. The selection of APSIM or DSSAT or other crop models depends on the preference and the expertise of the researcher, along with the input data and intended use/application of the model.
Being a water-driven model, AquaCrop applications are mainly on soil water balance, evapotranspiration, and water use efficiency; therefore, the AquaCrop model was not used in the analysis described in this paper. The number of publications on both models (APSIM and DSSAT) showed an increasing trend, while the highest number of publications was from the most recent year (2021). In the year 2021, 39 and 21 publications were published on APSIM and DSSAT models, respectively. The highest number of publications belongs to the last decade (2012-2021), where 75.6% of APSIM and 69% of DSSAT-related documents were published. The variation of publications throughout time is shown in Figure 1. Around 91% of soil-related crop modeling papers in the Scopus database were published in peer-reviewed journals ( Figure 2). were performed using VOSviewer version 1.6.18 [32,38,39]. VOSviewer is a software that can be used in visualizing bibliometric networks. It can create bibliometric networks and has advanced layout and clustering techniques and natural language processing techniques. The text mining option is used in co-occurrence networks and keyword analysis [39]. In the visualizations, different clusters were marked in different colors.

Development of a Database
In the selected papers used in the analysis (Section 2.2), the soil-related application, geographic location, and major type of application mentioned in the publications were evaluated as the next step and documented. The dataset was used to develop a database on crop model applications in soil research.

Performance Analysis
The Scopus search found 1446 publications, 660 on APSIM, 299 on AquaCrop, and 487 on the DSSAT model. After the manual screening of publications on applications of soil science, a total of 523 articles were remained (336 APSIM and 187 DSSAT) and used in the analysis. The selection of APSIM or DSSAT or other crop models depends on the preference and the expertise of the researcher, along with the input data and intended use/application of the model.
Being a water-driven model, AquaCrop applications are mainly on soil water balance, evapotranspiration, and water use efficiency; therefore, the AquaCrop model was not used in the analysis described in this paper. The number of publications on both models (APSIM and DSSAT) showed an increasing trend, while the highest number of publications was from the most recent year (2021). In the year 2021, 39 and 21 publications were published on APSIM and DSSAT models, respectively. The highest number of publications belongs to the last decade (2012-2021), where 75.6% of APSIM and 69% of DSSATrelated documents were published. The variation of publications throughout time is shown in Figure 1. Around 91% of soil-related crop modeling papers in the Scopus database were published in peer-reviewed journals ( Figure 2).

Geographic Distribution of Applications
Soil-related applications using the APSIM and DSSAT models were conducted in 58 countries. Out of all, APSIM-related applications are found in 41 countries, whereas DSSAT applications are available in 43 countries ( Figure 3). The highest number of APSIM publications were reported from Australia, followed by China and New Zealand, while DSSAT publications were reported from the USA (Figure 3).

Geographic Distribution of Applications
Soil-related applications using the APSIM and DSSAT models were conducted in 58 countries. Out of all, APSIM-related applications are found in 41 countries, whereas DSSAT applications are available in 43 countries ( Figure 3). The highest number of APSIM publications were reported from Australia, followed by China and New Zealand, while DSSAT publications were reported from the USA ( Figure 3).

Geographic Distribution of Applications
Soil-related applications using the APSIM and DSSAT models were conducted in 58 countries. Out of all, APSIM-related applications are found in 41 countries, whereas DSSAT applications are available in 43 countries ( Figure 3). The highest number of APSIM publications were reported from Australia, followed by China and New Zealand, while DSSAT publications were reported from the USA (Figure 3).

Citation Analysis
Summary statistics of the citations and normalized citations are mentioned in Table 1. The DSSAT model has relatively higher mean citations (30.8) and mean normalized citations (4.7) than the APSIM model. Even though the number of publications is lower for DSSAT than the APSIM, the median and 75th percentiles of both citations and normalized citations are higher in DSSAT than in the other model (refer to Table 1). On average, journal publications received a comparatively higher number of citations and normalized citations than other types of publications.  The collaborative network of countries in soil-related crop modeling applications is shown in Figure 4a (for APSIM) and Figure 4b (for DSSAT), where a higher collaboration strength can be observed between the countries in the same cluster. The collaborative country network revealed that there are 53 and 52 countries for APSIM and DSSAT models, respectively. In the APSIM model, there were 25 countries with more than 3 publications, whereas 22 countries for the DSSAT model were under the same criteria.   There are five clusters for APSIM collaborative network, and Germany, China, Australia, Netherlands, and India ranked top in each cluster. Australia has had the highest total link strength among all the countries since APSIM was developed there. As shown in Figure 4, there are five clusters for DSSAT collaborative network, and the United States has the highest total link strength. Germany (cluster 1), Egypt and Spain (cluster 2), the United

Co-Occurrence of Author Keywords
In the APSIM, out of the 875 author keywords, 49 of them appeared a minimum of 5 times. Out of the 49 selected keywords with the highest appearance, nitrogen, evapotranspiration, water use efficiency, carbon sequestration, and nitrate leaching ranked 6, 7, 10, 11, and 12, respectively (refer to Appendix A Table A2). In the DSSAT, out of 516 author keywords, 34 appeared a minimum of 5 times, where soil moisture, water use efficiency, and evapotranspiration ranked 4, 6, and 8, respectively (Appendix A Table A2). The link between author keywords is shown in Figure 5. Generally, it is evident that the authors used some soil-related keywords as author keywords.

Keywords from Abstracts
Out of the 10,195 terms in the abstracts, 54 appeared more than 50 times in APSIMrelated papers, while 17 terms appeared more than 100 times. In DSSAT, out of 5906 terms, 21 terms appeared more than 50 times, while there are 4 terms that appeared more than 100 times. Soil-related terms such as PAWC (plant available water content), SOC (soil organic carbon), N 2 O emission, deep drainage, temperature, soil moisture, soil, and soil water content terms showed higher relevance scores compared to other terms. Interestingly, the top two terms in APSIM were plant available water content and soil organic carbon. The top 20 terms based on the relevance score, which were extracted from abstracts, are shown in Table 4.   When considering both models together, out of 14,352 terms, 96 appeared more than 50 times, while 29 terms appeared more than 100 times (refer to Figure 6). Out of them, soil-related terms such as soil organic carbon, soil, water, soil water, and nitrogen ranked 2, 9, 11, 19, and 27, respectively, which suggests the use of soil terms in the abstracts. 100 times. Soil-related terms such as PAWC (plant available water content), SOC (so ganic carbon), N2O emission, deep drainage, temperature, soil moisture, soil, and so ter content terms showed higher relevance scores compared to other terms. Interest the top two terms in APSIM were plant available water content and soil organic ca The top 20 terms based on the relevance score, which were extracted from abstract shown in Table 4.
When considering both models together, out of 14,352 terms, 96 appeared more 50 times, while 29 terms appeared more than 100 times (refer to Figure 6). Out of soil-related terms such as soil organic carbon, soil, water, soil water, and nitrogen ra 2, 9, 11, 19, and 27, respectively, which suggests the use of soil terms in the abstracts Figure 6. Author keywords extracted from abstracts-both models.

Applications of Crop Models in Soil Research
Soil-related applications of APSIM and DSSAT crop models are shown in Figure 7. Similar types of applications were categorized into clusters to identify the major applications. For example, soil water balance, drainage, and water use efficiency were clustered under soil water, while soil compaction, permeability, and other physical properties that do not belong to major classes were clustered into physical properties. In addition, greenhouse gas emission (GHG) includes N 2 O, CH 4 , and CO 2 emissions. It is evident that the APSIM model (15 types) has a relatively higher number of application types than the DSSAT model (11 types). In both models, the highest number of applications were reported on soil water-related publications (215 and 127 in APSIM and DSSAT models, respectively), followed by N dynamics (129 and 45) and organic carbon (87 and 31). Other than that, both models were used to study N leaching, evapotranspiration, nutrient dynamics, P dynamics, physical properties, soil productivity, and temperature. APSIM model also was used to simulate GHG emission from soils and K dynamics (Figure 7).

Database on Soil-Related Applications of APSIM and DSSAT Models
A database on the application of process-based crop models in soil research was developed and published in an online repository (doi: 10.17632/8yh57yxcrw.1, accessed on 23 February 2023). This database contains details on the title, authors, source of publication and the published year, crop model and associated module in each model (if any), geographic location of application and the major application type. The initial version of the database contains details on publications on APSIM and DSSAT models during the 2000-2021 period only. However, it is expected to improve the database with other crop models, and the database will be updated with recent data. The database can be used to obtain an idea of the specific application in different geographic locations.

Discussion
The data from the Scopus analysis illustrates that for 21 years (until the year 2021), several applications of soil-related processes were implemented using process-based crop models. However, soil-related applications of crop models have a limited number of publications when compared with the total number of publications of the selected models (data not presented here). This review reveals that soil-related applications in crop models have significantly increased in the last decade of study (2012-2021), with over 70% of publications appearing for that period. Therefore, this positive trend suggests the use of crop models in soil research for future studies. This increased number of publications can be accelerated after the development of soil-associated modules [40] and/or modifications in the traditional/regular crop models.
Out of the two models studied in detail, Agricultural Production Systems Simulator (APSIM) has a more diverse range of applications than the Decision Support System for Agrotechnology Transfer (DSSAT) model. This is proven by the fact that the number of publications is also higher for APSIM than for DSSAT. However, other than the crop models reviewed here, there are other crop models with specific soil-related capabilities and applications. For example, the ARMOSA crop model was used to simulate NO3-N leaching from agricultural systems [40]. AquaCrop [41,42], the model omitted in the analysis, is a water-driven model that is mainly used to simulate soil water-related conditions. Therefore, two widely used crop models were used in this paper to assess the applications

Database on Soil-Related Applications of APSIM and DSSAT Models
A database on the application of process-based crop models in soil research was developed and published in an online repository (doi: 10.17632/8yh57yxcrw.1, accessed on 23 February 2023). This database contains details on the title, authors, source of publication and the published year, crop model and associated module in each model (if any), geographic location of application and the major application type. The initial version of the database contains details on publications on APSIM and DSSAT models during the 2000-2021 period only. However, it is expected to improve the database with other crop models, and the database will be updated with recent data. The database can be used to obtain an idea of the specific application in different geographic locations.

Discussion
The data from the Scopus analysis illustrates that for 21 years (until the year 2021), several applications of soil-related processes were implemented using process-based crop models. However, soil-related applications of crop models have a limited number of publications when compared with the total number of publications of the selected models (data not presented here). This review reveals that soil-related applications in crop models have significantly increased in the last decade of study (2012-2021), with over 70% of publications appearing for that period. Therefore, this positive trend suggests the use of crop models in soil research for future studies. This increased number of publications can be accelerated after the development of soil-associated modules [40] and/or modifications in the traditional/regular crop models.
Out of the two models studied in detail, Agricultural Production Systems Simulator (APSIM) has a more diverse range of applications than the Decision Support System for Agrotechnology Transfer (DSSAT) model. This is proven by the fact that the number of publications is also higher for APSIM than for DSSAT. However, other than the crop models reviewed here, there are other crop models with specific soil-related capabilities and applications. For example, the ARMOSA crop model was used to simulate NO3-N leaching from agricultural systems [40]. AquaCrop [41,42], the model omitted in the analysis, is a water-driven model that is mainly used to simulate soil water-related conditions. Therefore, two widely used crop models were used in this paper to assess the applications of crop models in soil research as a case study. Even though there are several crop models with different capabilities and applications, only APSIM and DSSAT models were assessed here, which is a limitation of the current study. It is important to assess the soil-related applications of other crop models and strengthen the database.
Also, there are stand-alone soil models such as QUantifying Interactions between terrestrial Nutrient CYcles and the climate system (QUINCY) [43] and Jena Soil Model (JSM) [44] with high capacities to simulate complex soil properties. The possibility of combining these specific soil models with crop models is important, which creates new research avenues. Other than the crop-and soil-related process, these crop models can be used in other agri-environmental-related studies. For example, the inclusion of soil water components in crop models is beneficial in climate change studies [45]. As it was already stated, food production in the world to date is in scarcity due to ongoing climate change. Therefore, the possibility of combining process-based crop models in a wider range of environmental research should be assessed in detail.
The collaborative country network was developed in this study (refer to Figure 4). It is important to promote the publications and collaborations between researchers from developing and developed countries in crop model applications in soil research. Since the applications are available for limited countries, it is also important to popularize crop model applications in other countries. Dissemination of knowledge in the related studies would enhance food sustainability in developing countries. The published database can be used to determine the suitable model for a specific region and discipline of interest.
According to this review, researchers have explored various aspects of soil-related processes in crop models that ranged from nutrient dynamics, water, greenhouse gas emission, and other physical and chemical properties (Figure 7). Not only the plant processes but also performance in livestock farms were also assessed. The main purpose of some of the publications was to study the soil-related processes [14,[46][47][48][49][50], and in most of the publications, it was either documented as supportive output to evaluate any other plantrelated processes. The developed database can be used in selecting crop models/supporting modules for various types of applications, such as soil nutrient dynamics, GHG emissions, and other soil physicochemical applications in different geographic regions.
Another limitation of the study is the less perfect bibliometric analysis results due to the use of one database [51]. There are few other literature databases other than Scopus. Therefore, it is expected to improve the current dataset using more data obtained from the other literature databases, such as Web of Science, PubMed, Directory of Open Access Journals (DOAJ), and Chinese National Knowledge Infrastructure (CKNI), etc., for the future study. The information from other databases can then be used to perform a comprehensive review and strengthen the developed database. Further, limitations with the accessibility to some of the databases hinder such applications.
Bibliometric analysis is an important tool that was used to assess the global scientific production on soil-related applications of two widely used crop models; APSIM and DSSAT. It can be used to assess the intellectual structure of a specific field of interest and detect patterns of publications and different types of networking [52]. The indicators used in this study are an indication of the impact and the importance of the work tested here, where a wider research community can get benefits. The findings of collaboration and country networks can be used in designing collaborative research and funding applications. This will also create several research avenues, such as detailed soil-related applications, agri-environmental modeling, multi-model comparisons, and model strengthening and development.

Conclusions
The bibliometric analysis conducted in this study determined the applications and trends of process-based crop models in soil research. This research was carried out to fill the research gap in providing a bibliometric analysis on collaboration and applications of crop models in different geographic regions. The generated collaborative country network and keyword maps revealed the collaboration between countries and different terms used in soil-related research in crop modeling. A database on the application of process-based crop models in soil research was developed and published in an online repository that can be used to determine the specific application of crop models in different geographic locations. The knowledge gaps in soil-related applications of crop models were identified through the bibliometric analysis conducted in this study, which can be addressed through further research. The publications reviewed here have indicated that the crop models could be successfully used as a decision support tool in soil-related research.

Data Availability Statement:
The dataset is available for research work from the corresponding author.

Conflicts of Interest:
The authors declare no conflict of interest.