Since I started writing for our blog, I always try to emphasize the large amount of data that can be generated and collected in the all of the different activities of the agricultural environment, both outside and inside farms. This data is an essential part of the agricultural digital transformation, which relies on data generated by field activities to direct efforts, solve problems, and identify opportunities for agribusinesses.
In farms, there is a large amount of data that can be collected — either by sensors, manually or through other techniques. With this mass of information collected, the possibilities of analysis grow very quickly. If this data is correctly analyzed, it is possible to generate very important information that can increase productivity and quality of the agricultural production.
In a recent blog post, we commented on the need for field data collection mechanisms, in order to extract standardized data according to the type of survey desired and collecting the data as reliably as possible. In this context, we have introduced VNT.Field.Collector — a Venturus product that allows producers to generate forms suitable for the collection of information in their farms — precisely because of the need to standardize the data collected on the fields.
In the current article, we will focus on discussing what types of information can be extracted from the analysis of data collected persistently during many production cycles — persistent because it takes work to extract and maintain history data —regarding to the agricultural environment. Data analysis is essential, since maintaining data on the agricultural production and development process can be a costly task and does not bring effective results in isolation.
Imagine the following scenario: a coffee producer would like to assess the impact of spraying against Leucoptera coffeella (know in Brazil as “bicho mineiro” or mining bug) — to this day, one of the main pests of coffee crops— on their coffee farm. It would be very good to have a lot of data about the pest within its culture, that is, a true history of the coffee crops.
Examples of information that can be collected are: weather data (temperature, humidity, wind, amount and time of rainfall), data of soil fertility levels year-by-year, added to sample values of the amount of pests detected per sample, dates of sprays and which products were used and in what dosage.
With reliable data, farmers can analyze the information and know if the effect of a defensive against the miner bug is working as expected. The analysis can identify how long the effect lasted and, additionally, it can assess the effect of different variables — such as humidity, temperature, soil fertility etc. — and their correlations with each activity performed in the field.
That means that farmers can identify very important information through the analysis of the different data collected. The correct interpretation of the data collected and its history can help farmers plan and analyze their cultural treatments in coffee crops with greater clarity. Thus, they become able to perform their work more effectively and use their resources to increase production.
The purpose of Data Analytics is the extraction of significant data to search for information that is useful in research or business, using tools and methods to collect, organize and evaluate this data. This analytics system can transform, organize, and model data so that it is possible to find valuable insights about trends, behavior, and process improvements.
With this information, all types of companies can make more accurate decisions. In Agribusinesses, data analysis can also bring valuable insights in the way of reducing costs and increasing the efficiency of farming processes. Data analysis requires, however, some previous steps, which involve, respectively, the collection and treatment of the data to be analyzed:
Therefore, the first phase of the process is the data collection itself — it has already been covered in a previous article , in which we talked about the need for reliable data that conforms to the standards necessary to perform its analysis. That means that, in order to have a good data analysis, it is necessary that the extraction of data that will be used is as reliable as possible.
For instance, if data collection is done through sensors, it is necessary that there always be power at the time of data collection or that there are no data transmission problems. In case of manual data collection, it is necessary that operators do not forget to enter all the necessary data (fields in spreadsheets and forms must be filled correctly and with appropriate numerical values). Poor data collection quality will inevitably lead to incorrect or poor quality in the data analysis. The quality of the data collected is just as important as the data analysis itself.
In the next phase, in which the data is organized, it is necessary to verify the quality and nature of the data that will be used for the analysis itself. This phase would be data processing. When analyzing the information collected, it is important to verify that the existing data in the database is consistent. There are several factors that can cause problems in data analysis — such as missing or duplicated data, information in incorrect units of measurement or simply unrealistic, for example — and, therefore, the information needs to be investigated and dealt with before analysis.
After the data collection and processing steps have been performed, we enter the data analysis phase itself. In it, we find four main types of data analysis:
Prescriptive analysis examines existing data, trying to discover future possibilities. Mapping past patterns in the database, we seek to find future definitions in each field. Thus, prescriptive analysis uses data mining, statistics, and history data techniques to identify future trends.
A use of prescriptive analysis in agribusiness would be the monitoring of diseases of a poultry farm. Generating data to track the entire history of the chickens and determine the data in cases where there have occurred problems, such as certain diseases. With this data in hand, in a next cycle, if the factors are repeated, the farmer can act to prevent that same sequence from bringing damage or losses to their production.
This type of analysis evaluates interesting patterns from the past in order to try to predict the future. By identifying past patterns in the database, you can map future definitions in your business. For example, some businesses use predictive analytics for the entire selling process, analyzing the main sources, the number of contacts, type of communication, documents. That means that this data, when used correctly, can support sales, marketing and other types of forecasts.
Diagnostic analysis is used more in the sense of understanding or finding out why something happened. In the case of agricultural scenarios, one use would be to analyze the impacts of different cultural tracts on the crop (for example, a specific fungicide spraying) and, subsequently, to verify the impact caused by this action. Checking whether the action performed had the expected effect and that the crop managed to avoid a spread of plant disease and, if not, try to understand what led to such a result.
Descriptive analysis analyzes current data, based on data input. This analysis enables the understanding of data and events, preferably in real time. It’s a way to visualize the data, understand how the data provided is organized and what it means, without the need to relate it to past and future patterns. An example in the agricultural field would be the data of sensors in the soil that identify the amount of water that the soil has managed to absorb at the different depths of soil layer. With this information, when performing an irrigation, it is possible to verify how efficient and at what soil levels the irrigation performed managed to achieve the objectives set.
An example of data analysis
During Unicamp’s continuation education course in Data Science, we carried out some practical work, in which we analyzed the data of a weather station in Campinas (Brazil). The sample used had temperature, wind speed and humidity data, which are collected at a weather station every ten minutes, from 2015 to 2019.
Therefore, by working with the data extracted from the table with meteorological data of several years in the Campinas area and analyzing the data of each time of day, we sought to understand what was the behavior of the winds in this area throughout the 4 seasons of the year.
The objective of this data analysis was to understand which would be the best periods of the day to perform tasks that could be influenced by the wind in an agricultural property. Some examples of tasks that could be affected by the winds are: agricultural spraying, spreading limestone or powdered fertilizers and air-borne irrigations (such as sprinkling and central pivot irrigation, among others).
An important analysis regarding winds is that, usually, early in the morning, they tend to have a lower average speed and, from eight o’clock in the morning onwards, their speed tends to increase gradually. In all seasons of the year, we find that at, when night falls, there is a tendency for wind speed to increase, reaching its peaks around 11 p.m. or midnight, and gradually decreasing until dawn.
To agricultural operations, in cases where it is desirable to carry out some agricultural operation that can be negatively influenced by the speed of the winds, the conclusion is that the night period does not seem to be the most favorable. During the day, although there isn’t a big difference in the wind speed, there is a tendency for less wind action in the morning, until noon.
Another interesting fact to note with regards to wind speed is that, on average, it is usually lower during the summer. Therefore, there may be days when we have stronger winds in the summer, but, on average, the speed of the winds is usually lower.
Furthermore, we can notice that, in the spring, night irrigations and sprays are probably not the best idea, since there is a greater chance for higher wind speed. In these conditions, irrigation or spraying may float to the nearby crops (not the intended target), since the applied products can reach improper places, due to the wind.
Currently, we are hearing a lot about the use of drones in several agricultural activities. However, based on the wind charts, we can note that opting for the use of drones at night seems not to be a good alternative (especially during autumn, winter and spring), since the winds can cause the spray/application of product to drift, and, in more extreme cases, can affect the path of the drones.
This example was made based on simple Meteorology data. In terms of agriculture, if we had more information, we would be able to do a number of other analyses. With pest infestation data, for example, we could analyze possible effects of temperature or humidity in regards to the presence of pests. Every farmer knows what kind of information would be useful and can suggest the different analyses needed, according to the objective of each study and business needs.
Another point to be checked with the collected data refers to using machine learning algorithms for the analysis. With increasingly better algorithms, it is possible to analyze associations, correlations or even groupings of data that make sense in machine learning and can provide increasingly accurate data regarding crop conditions and processes. In this article, we have focused on showing that the simple analysis of crop data can bring us great and important information about the conduct of agricultural activity.
In every agricultural activity, there is a lot of data that can be extracted, from every step of the production process, whether in agriculture, livestock, poultry, seed storage, genetic research and in various other areas. A correct data extraction strategy, be it with sensors, location information or manual information can bring us information that, by itself, does not bring much benefit, but, when analyzed along with other data, can bring us information and knowledge that streamline the decision-making process and the adoption of more economically viable techniques.
Data science is a technique that is being used more and more often, and, in order to benefit from the results it can bring to farms, it is important that the data from the crops is stored and collected as soon as possible, since, in many cases, history is one piece of information that brings greater knowledge of the whole process.
Venturus is prepared to provide data collection tools, either by sensors or tools — such as VNT.App.Colletor. But not only in data extraction, also in the most modern techniques of data analysis and also with machine learning. Technical knowledge combined with knowledge of agricultural processes can bring very important results in productivity and efficiency gain in farms.
Will we analyze new field data and discover new relationships between existing information in the crops? People who experience the day-to-day of the farm are able to identify which problems need to be analyzed and can collect the data necessary for us to do the analysis. Venturus, with its experience in technology and the knowledge it has been generating in agribusiness, can help you in the most different analyzes.