Open Access is an initiative that aims to make scientific research freely available to all. To date our community has made over 100 million downloads. It’s based on principles of collaboration, unobstructed discovery, and, most importantly, scientific progression. As PhD students, we found it difficult to access the research we needed, so we decided to create a new Open Access publisher that levels the playing field for scientists across the world. How? By making research easy to access, and puts the academic needs of the researchers before the business interests of publishers.
We are a community of more than 103,000 authors and editors from 3,291 institutions spanning 160 countries, including Nobel Prize winners and some of the world’s most-cited researchers. Publishing on IntechOpen allows authors to earn citations and find new collaborators, meaning more people see your work not only from your own field of study, but from other related fields too.
This chapter introduces an advanced decision-support software aimed at enhancing mechanized agricultural tillage practices. It emphasizes the necessity of detailed planning before sowing to efficiently utilize resources and prevent soil deterioration. The developed algorithm harnesses the power of compatibility matrices to analyze the complex interrelationships among various factors such as soil types, crop types, and machinery options. The study collected exhaustive data on tillage practices and uses this information to create compatibility matrices, enabling an intelligent algorithm to guide decision-making processes. The central feature of this software is its ability to generate a comprehensive tillage plan in natural language, serving as a detailed guide for farmers and other stakeholders. This algorithm is incorporated into a user-friendly web application, offering stakeholders an interactive platform for decision-making. The software is thoroughly validated by domain experts to ensure its reliability and accuracy.
Faculty of Engineering and Architecture, Catholic University of El Salvador, El Salvador, Central America
*Address all correspondence to: carlos.martinez@catolica.edu.sv
1. Introduction
In order to satisfy the current needs of agriculture, tillage nowadays must be mechanized, for which reason it is necessary to have knowledge about the different implements that must be used and their functions, as well as the type of machine that must be coupled to carry out the cultivation—traction. An efficient tillage strategy would preserve the soil, save financial outlays, and contribute to the minimum-necessary use of diesel fuel for the general operation. Not knowing how to till effectively prior to sowing can contribute to its deterioration. Improper tillage practices can lead to the disturbance and erosion of topsoil [1]. For instance, this phenomenon can occur through the mixing of surface layers with deeper horizons or through erosion caused by wind or rain. Developing a decision-making software for tillage is important due to the complexity of the problem. The software can analyze and optimize the interrelations involved, providing accurate recommendations. It reduces the burden on farmers, ensures consistency, and maximizes productivity and sustainability.
When developing the algorithm for decision support in tillage planning, a deliberate choice was made to utilize a compatibility matrices model instead of machine learning techniques. While machine learning is a powerful tool commonly employed for pattern recognition and prediction, it may not be the optimal approach for addressing the intricacies and interrelationships among multiple variables [2, 3, 4], typical of the work with natural resources. The complexity of these relationships necessitated a tailored solution that could effectively analyze and process the dependencies between soil types, crop types, machinery options, and other relevant factors. By leveraging compatibility matrices, a structured framework that captured and represented these interrelationships was developed, enabling the algorithm to make informed decisions and provide recommendations based on the compatibility of various options. This approach not only ensured accuracy and reliability in tillage planning but also provided a transparent and interpretable model that could be easily understood and validated by domain experts. Thus, the use of compatibility matrices proved to be a more suitable and effective solution for this specific software application. It should be noted that, similar to machine learning models, the capabilities and accuracy of this model can be further improved if the categories and compatibility matrices are supplemented with more information.
The process for modeling mechanized agricultural tillage involved several starting diagnoses, such as an in-depth analysis of cultural framing practices, review of technical guides, and research papers, to gather a wide range of information on the most appropriate and effective tillage practices for each specific crop. In addition, interviews were conducted with experienced professionals and farmers in the field of agricultural tillage to gain valuable insights and practical knowledge regarding the most successful techniques.
The methodology involved the development of compatibility matrices to process complex relationships between variables in the tillage decision-making process. These matrices facilitated storage and analysis, enabling the algorithm to select suitable machinery and tillage practices. Parameters, such as soil type, crop type, and equipment availability, were considered. A theoretical model using tuple relational calculus was created to efficiently store and organize the data [5, 6]. Matrix arrays were developed and populated with collected variables and information, along with rules governing their interrelationships. This comprehensive dataset formed the foundation for subsequent stages of the project. Then, the algorithm was developed taking into account factors such as crop requirements, soil characteristics, machinery capabilities, and environmental considerations. To ensure the accessibility and usability of the algorithm, a user-friendly web interface was developed. This interface served as a platform for users, including farmers, agronomists, and other stakeholders, to interact with the algorithm and utilize its decision-making capabilities. To ensure reliability and accuracy, the model underwent validation by a panel of domain experts including agronomists, farmers, and machine operators specialized in tillage practices. Through rigorous evaluation, they scrutinized the model’s outputs and assessed its performance in real-world scenarios. Valuable feedback was provided, potential limitations were identified, and the precision and effectiveness of the algorithm in generating suitable tillage strategies for diverse crops were validated.
In the context of developing software for decision support in environmental sciences and farming, data categorization played a crucial role for several reasons. First, it facilitated a better understanding of the complex relationships between various environmental factors and farming practices. By identifying distinct categories, such as soil types, weather conditions, and crop growth stages, it became possible to analyze the interactions between these factors, thereby enabling more informed decision-making. Furthermore, data categorization led to the creation of more accurate decision-support processes. By grouping data based on specific characteristics, the computer algorithm could identify patterns and trends within each category, resulting in improved predictions and tailored recommendations for individual farming situations. The following variables were identified and categorized:
Soil compaction: Soil compaction is a process that occurs when external pressure, such as from heavy machinery or foot traffic, causes the soil particles to become more densely packed together. This reduces the pore spaces between the particles, leading to decreased air and water infiltration, as well as diminished root penetration [7]. Soil compaction can have negative consequences on plant growth and agricultural productivity, as it restricts the movement of nutrients, water, and oxygen, which are essential for healthy plant development (Table 1).
Crop type: This factor refers to the classification of various cultivated plants based on their specific characteristics, uses, and growth habits (Table 2). Different crop types have unique requirements for optimal growth, including specific soil conditions, water needs, and management practices.
Land extension: This factor categorizes the possible extensions of land to be cultivated in terms of whether having their own or rented machinery (Table 3). This decision depends on the economic market opportunities specific to each growing region, which may determine the feasibility of acquiring machinery [8].
Soil humidity: Also referred to as soil moisture, soil humidity is a crucial parameter for agronomic crops [9]. It indicates the quantity of water present in the soil, which can potentially restrict the plant’s uptake of vital nutrients and water (Table 4).
Organic matter (OM): Organic matter is composed of decomposed plant and animal residues, microorganisms, and humus. Soil organic matter plays a crucial role in soil fertility by supplying vital nutrients for plant growth, enhancing soil structure, and improving water retention and infiltration [10] (Table 5).
Type of tillage: Tillage is a crucial practice for agronomic crops, with various types available that depend on factors such as the specific crop, soil texture, and other phenomena. Table 6 shows the principal tillage operations that are used for preparing the soil by breaking up and turning over the soil layers.
Agricultural soil depth: This factor refers to the thickness of the uppermost layer of the ground surface that is suitable for crop production. This layer, commonly known as the root zone, is where the majority of biological activity takes place, and it is where crops obtain their nutrients and water. The depth of this layer can vary significantly depending on factors such as soil type, landscape characteristics, and the specific crop being grown (Table 7). It also determines the appropriate machinery required for proper tillage. For instance, if plowing is done too deeply into a shallow agricultural layer, it can result in the mixing of soil horizons and subsequently lead to a loss in fertility.
Soil rockiness: This factor refers to the approximate depth at which rock or stone content is present, which can potentially obstruct the operation of farm machinery during tillage (Table 8).
Soil hardness: This factor affects the ease and effectiveness of tillage operations and greatly depends upon the type of soil (Table 9). Heavier soils may require more powerful machinery or specific tillage methods, while softer soils are generally easier to be tilled [11].
Tractor horsepower: This variable determines the ability to work with different soils and operate specific types of farm implements. Higher horsepower is necessary for effectively tilling harder and more compacted soils, while less compacted soils require lower horsepower (Table 10). It is also essential to consider that farm implements are designed to be operated by tractors with specific horsepower capacities.
Machine implement: This machine implement refers to a device or tool that can be attached to a tractor or any other farm vehicle to perform specific agricultural tasks (Table 11).
Value
Category
Description
1
low
Soft texture
2
medium
Somewhat plowed
3
high
Heavy soil
4
very high
Heavy soil and no plowed
5
extreme
Very compacted with plow pan
Table 1.
Soil compaction categories.
Value
Category
Description
1
maize
Rainfed agronomic crop
2
sorghum
Rainfed agronomic crop
3
rice
Rainfed agronomic crop
4
rice
Planted in dry soil and then flooded
5
sugar cane
Planted residual soil humidity
6
sugar cane
Maintenance to established crop
7
vegetables
Rainfed short-cycle crop vegetables
8
vegetables
Dry planting beds for irrigation
9
vegetables
Use of residual soil humidity
Table 2.
Crop categories.
Value
Category
Description
1
area < 5 ha
Pay for service or rent machinery
2
area < 15 ha
Rent machinery or buy own if possible
3
area < 40 ha
Own machinery recommended
4
area < 90 ha
Must own machinery
5
area > = 90 ha
Must own machinery
Table 3.
Land extension categories.
Value
Category
Description
1
permanent wilt
Soil is almost as dry as possible
2
wilt point
Soil is dry but plant could recover
3
field capacity
Best amount of humidity
4
saturated
Ground is not firm due to moisture
5
oversaturated
Soft mud and waterlogged areas
Table 4.
Soil humidity categories.
Value
Category
Description
1
0% < OM < = 1%
Very scarce
2
1% < OM < = 3%
Scarce
3
3% < OM < = 4%
Medium
4
4% < OM < = 7%
Abundant
5
7% < OM
Overabundant
Table 5.
Categories of presence of organic matter (OM).
Value
Category
1
Pulverize
2
Harrow tillage
3
Disk plow
4
Subsoil plow
5
Fine seed planting
6
Regular seed planting
7
Compaction
8
Seedbed preparation
9
Crop spraying
10
Strip plow
11
Narrow furrow preparation
12
Wide furrow preparation
13
Wide furrow leveling
Table 6.
Categories of tillage.
Value
Category
Description
1
0 cm < depth < = 15 cm
Very shallow soil
2
15 cm < depth < = 30 cm
Shallow soil
3
30 cm < depth < = 50 cm
Regular soil
4
50 < depth
Deep soil
Table 7.
Categories of agricultural soil depth.
Value
Category
Description
1
0 cm < depth < = 15 cm
Very shallow rockiness
2
15 cm < depth < = 30 cm
Shallow rockiness
3
30 cm < depth < = 50 cm
Intermediate rockiness
4
50 < depth
Deep rockiness
Table 8.
Categories of soil rockiness.
Value
Category
Empirical hardness index
1
Sand
1
2
Loamy sand
3
3
Sandy loam
2
4
Loam
4
5
Silt loam
5
6
Silt
3
7
Sandy clay loam
4
8
Clay loam
5
9
Silty clay loam
5
10
Sandy clay
5
11
Silty clay
6
12
Clay
7
Table 9.
Categories of types of soil and their respective given hardness indexes.
Nine matrices were created to store the relationships between the variables defined earlier. These matrices included rows representing the values of each category and a field indicating the boolean acceptability of their respective combinations [12, 13], similarly serving as relationship tables [14, 15]. The compatibility matrices were filled with information that trained the model, serving as the basis for subsequent computations. The following is a description of each of these matrices:
Compatibility of crop and humidity: Each crop type thrives better under specific soil humidity conditions. For instance, most crops exhibit promising yield potential when the soil is at field capacity (as indicated by a value of 3 in Table 4). However, the productivity of rice improves when it is flooded (value of 4 in Table 2), often requiring high levels of humidity found in saturated and oversaturated soils (values of 4 and 5 in Table 4). Consequently, an example of the compatibility rows representing these relationships is shown in Table 12.
Given that memory arrays or database tables are used to store such information, relational theory models can be used to depict the theoretical design. Therefore, the relational calculus expression illustrating the compatibility matrix between each crop and soil humidity value is presented as follows (1).
Compatibility of crop and tillage: Certain categories of tillage are suitable for specific crops but must be performed in a strict order; otherwise, they may be ineffective or even prove counterproductive. For instance, furrow preparation cannot be done if compacted soil has not been softened beforehand. The appropriate tillage techniques were determined for each crop and categorized based on their respective steps (2).
Crop value
Soil humidity value
Accepted
4
1
False
4
2
False
4
3
False
4
4
True
4
5
True
Table 12.
Example of compatibility of a flooded rice crop and its required humidity.
Where the steps are represented by ordinals, as follows:
preparation prior to decompaction,
decompaction,
soil smoothing,
furrow or bed preparation,
sowing, and
soil finishing.
Compatibility of crop and organic matter: Excess of organic matter can obstruct tillage, seed birth, and water use efficiency. Therefore, the software model considers a maximum recommended percentage of organic matter for each crop to ensure acceptability (3).
Compatibility of crop and soil depth: Crops like sugar cane require deep planting, necessitating deep agronomic soils. In contrast, rice and red beans can grow in shallow agronomic soils. As demonstrated by these examples, it is essential for each crop to have a minimum agronomic soil depth for proper development. The compatibility tuple between these two variables is as follows:
Compatibility of crop and soil rockiness: The presence of rocks in soil can limit the depth of tillage. Therefore, if some categories of tillage are not allowed in a soil, the crop--type options are also reduced, the reason for which those two variables are closely related (5).
Compatibility of crop and soil compaction: For proper development, each crop can tolerate determined type of soil compaction. The following compatibility tuple relates both variables and their respective determinants of acceptance (6).
Compatibility of implement, tillage, depth, and rockiness: Each piece of machinery is engineered to execute a specific type of tillage, operating to a maximum soil depth and under particular conditions of soil rockiness. The subsequent tuple links all these variables and their respective concordance acceptance values (7).
Compatibility of land extension and tractor: The profitability of farmland may hinge on its size. Depending on the extent of land, it might be more cost-effective to pay for tillage services, lease equipment, or invest in one’s own machinery. Hence, the size of the land can guide certain recommendations concerning economic investments that can be fed into the algorithm as String attributes (8).
Compatibility of tractor and implement: Each implement requires a tractor with a minimum horsepower specification for operation. However, using an oversized tractor with a small implement can lead to increased soil compaction and unnecessary fuel consumption, and can also shorten the lifespan of the machinery. For these reasons, a compatibility matrix has been established to align each implement with its ideal tractor category (9).
The decision-making process has resulted in a comprehensive guide providing systematic instructions on the practice of tillage. These instructions include specific recommendations on the type of farm machinery that is optimal for improving compacted or poorly tilled soils (refer to Figure 1A and B). The aim was to provide necessary information to support the creation of a more pliable soil texture (refer to Figure 1C). Images in Figure 1 not only illustrate the outcomes of different types of tillage, but they can also serve as an analogy for the tillage process itself, which begins with breaking up rough soil and continues until a soft and aerated soil structure is achieved.
Figure 1.
Different soil textures of the same farmland.
A web-based application1 was developed to offer a graphical user interface (GUI) and to process the information in accordance with the algorithm depicted in Figure 2. Upon accessing the site (Figure 3), users are presented with a series of options corresponding to the categories of the primary variables (refer to Section 3 of this chapter). These variables encompass crop type, land size, presence of organic matter, tractor horsepower, type of farm implement, tillage, and crucial soil attributes such as humidity, depth, rockiness, compaction, and inherent hardness.
Figure 2.
Flowchart of the decision-making algorithm.
Figure 3.
Graphical user interface (GUI) for selecting options (A) and displaying the algorithm results (B).
Users are expected to select one category for each variable and submit these variables to the server. However, given the possibility of some selections being inconsistent, the data need to be validated before proceeding with the software algorithm. For instance, rice, as a crop, cannot thrive in sandy soils due to their inadequate water retention capacity. Likewise, tillage for planting is not feasible when the soil is wet. Thus, the consistency of the variables is evaluated against the compatibility matrices and further corroborated through the following cross-checking of information (10).
The matrix that encapsulates the compatibilities among implements, depth, tillage, and rockiness was conjoined with the crop matrix through a comparison of tillage categories. In the same manner, matrices containing data on organic matter, soil compaction, and soil humidity were also merged to verify if all “accepted” parameters were indeed True. Given that each row in every compatibility matrix contains a unique descriptive String, it was possible to compile the reasons for the acceptance or rejection of each option selected via the GUI. This mechanism serves to provide valuable feedback to the user.
During the “model setup” phase, the need for tillage was evaluated using a scoring system that escalated in response to the soil texture category, its level of compaction, and the quantity of organic matter needing integration into the soil. This requirement score (11) was also diminished if a moderate level of humidity was inputted into the options.
The selection of the tractor type was made based on the size of the land and the Validation matrix, in which each implement was paired with the necessary tillage operations. Considering that larger tasks usually require more powerful tractors (with greater horsepower), the heaviest implement in the matrix influenced the tractor recommendation for the entire model (12).
Once a tractor was selected, the Validation matrix was filtered to only include those implements compatible with the chosen tractor. The remaining rows represented the exclusive options for farm machinery available for formulating the tillage plan.
Deep tillage requires heavy implements, the penetration depth of which is limited by soil rockiness, agronomic characteristics, and excessive soil humidity (Table 4). The algorithm assesses this process first by filtering for the dryness condition of the soil and for those implements deemed suitable for deep tillage (2). It then applies the following inequality to eliminate unsuitable implements (13).
Each plowing implement was previously assigned a numerical decompaction score. After selecting the largest viable implement, the number of tillage repetitions required to satisfy the “Requirement” (10) was determined. This requirement was accomplished by performing a simple modulus operation between these two values. This process was similarly applied to a lighter category of implements with the condition of Compcxg.step = 3, which includes harrows and rotatills (2). Following the same rules, the operational depth and the number of repetitions necessary to adequately soften the soil were also evaluated.
The finishing tillage was considered to take place when the soil is completely smoothed (14). This process makes use of implements numbered 15 to 22 as listed in Table 11, the primary purpose of which is to plant the seeds or condition the seedbed or furrows. The step condition of the model (13) was changed to: 4 ≤ Compcxg.step≤6. The usage of implements depends solely on the requirements of the crop, and they were preselected in the Validation matrix. The final selection was based on the tractor’s capacity (Table 10).
The final algorithm has been developed as part of a decision-support tool to assist with optimal tillage practices. The algorithm incorporated each of the important agricultural input variables as such and calculated the requirement for tillage based on various soil parameters, and subsequently recommended suitable machinery and tillage practices. Furthermore, a functionality that suggests whether to buy, rent, or pay for machinery services based on the land extension was included. The corresponding code for this algorithm is presented below:
Begin
Inputs:= (c, l, o, w, i, g, h, d, r, a, m)
For every u in Inputs, if u is not in Validation then,
return “Invalid input “ + Validation.description
For every T in Compwxi & Compwxl then,
if T is compatible with TractorHP then,
SelectedTractor = T
Requirement:= m + o + a - h
Filter Implements to Validation compatible with SelectedTractor
The graphical user interface (GUI) that incorporates the algorithm is depicted in Figure 3. Part B exhibits the list of implements and the recommended tillage plan, explained in natural language using the descriptions derived from the variable and compatibility matrices. This interface also proved valuable in validating the algorithm, as it was presented to a focus group of 45 farmers to gather their feedback. Each participant agreed that the software considered all necessary values to represent the model. After using the software for an hour, they reported an accuracy level of 95% or higher.
The software model was initially created by identifying the variables that influence the process of tillage planning, as well as their complex interrelationships. Each variable was numerically categorized and represented as rows in a matrix, with additional matrices created to store the interrelationships between these categories. The model can be calibrated and enhanced for specific situations simply by supplementing these matrices with more information. In operation, the algorithm takes in the user-configured variables, validates them against the restrictions in the categories matrices, and then begins filtering machine and tillage options by assessing their compatibilities. Subsequently, a tillage effort score is estimated and gradually reduced with each feasible tillage application that can be implemented. The final product is an accurate, mechanized tillage plan, expressed in natural language. This development can be immensely useful for inexperienced farmers, agronomy students, and other stakeholders, providing them with a tillage plan that not only aims to improve crop productivity but also works to reduce machine use waste and the risk of soil degradation.
1.Azab YF, Abbas HH, Jalhoum ME, Farid IM, Abdelhameed AE, Mohamed ES. Soil erosion assessment in arid region: A case study in Wadi Naghamish, northwest coast, Egypt. The Egyptian Journal of Remote Sensing and Space Science. 2021;24(3):1111-1118
2.Vowels MJ. Trying to outrun causality with machine learning: Limitations of model explainability techniques for identifying predictive variables. arXiv preprint arXiv:2202.09875. 2022. pp. 2-3
3.Pearl J. The limitations of opaque learning machines. Possible Minds. 2019;25:13-19
4.Gajendran MK, Kabir IF, Purohit S, Ng EY. On the limitations of machine learning (ML) methodologies in predicting the wake characteristics of wind turbines. In: Renewable Energy Systems in Smart Grid: Select Proceedings of International Conference on Renewable and Clean Energy (ICRCE) 2022. Singapore: Springer Nature Singapore; 2022. pp. 15-23
5.Gatterbauer W, Dunne C, Riedewald M. Relational diagrams: A pattern-preserving diagrammatic representation of non-disjunctive relational queries. arXiv preprint arXiv:2203.07284. 2022. pp. 8-11
6.Bossé É, Barès M. Preliminaries on crisp and fuzzy relational calculus. In: Relational Calculus for Actionable Knowledge. Cham: Springer International Publishing; 2022. pp. 165-216
7.Idowu J, Angadi S. Understanding and Managing Soil Compaction in Agricultural Fields. USA: New Mexico State University; 2013
8.Kirui O. The agricultural mechanization in Africa: Micro-level analysis of state drivers and effects. ZEF-Discussion Papers on Development Policy. 2019. pp. 18
9.Hohenegger C, Stevens B. The role of the permanent wilting point in controlling the spatial distribution of precipitation. In: Proceedings of the National Academy of Sciences. 2018;115(22):5692-5697. DOI: 10.1073/pnas.1718842115
10.Sullivan DM, Moore AD, Brewer LJ. Soil Organic Matter as a Soil Health Indicator: Sampling, Testing, and Interpretation. Oregon, USA: Oregon State University Extension Service; 2019
11.Moreno-Maroto JM, Alonso-Azcarate J. Evaluation of the USDA soil texture triangle through Atterberg limits and an alternative classification system. Applied Clay Science. 2022;229:106689
12.Lee SY, Liou RL. A multi-granularity locking model for concurrency control in object-oriented database systems. IEEE Transactions on Knowledge and Data Engineering. 1996;8(1):144-156
13.Muth P, Rakow TC, Weikum G, Brossler P, Hasse C. Semantic concurrency control in object-oriented database systems. In: Proceedings of IEEE 9th International Conference on Data Engineering. Vienna, Austria: IEEE; 1993. pp. 233-242
14.Yeh D, Li Y, Chu W. Extracting entity-relationship diagram from a table-based legacy database. Journal of Systems and Software. 2008;81(5):764-771
15.Bagui S, Earp R. Database Design Using Entity-Relationship Diagrams. Florida, USA: CRC Press; 2011
Open access peer-reviewed chapter - ONLINE FIRST
