The interaction between agriculture and technology has been improved. The search for more efficiency in the field combined with the reduction of environmental impacts meets the need to develop more advanced methods for processing agricultural data, which will help rural producers to make better decisions.
One of the subprojects developed by the Rede de AgroPesquisa e Formação Aplicada Paraná (Rede AgroParaná), which has financial support from SENAR-PR and the State government, seeks to determine how erosive processes occur based on a computational model called Artificial Neural Network ( RNA). Through the massive analysis of data on the chemical, physical and biological properties of the soil, the method compiles results from other subprojects of the AgroParaná Network conducted in the Southwest of the State.
ANN is a computational methodology inspired by the learning process of the human brain, with the ability to model, predict and classify data faster and simultaneously. According to the associate professor at the Federal Technological University of Paraná (UTFPR) and coordinator of the subproject, Eder da Costa dos Santos, artificial intelligence is similar to the concept of machine learning. In other words, based on defined rules, the ANN algorithms will make the most appropriate decision for the context, based on the recognition of patterns within the analyzed data, to understand the erosion processes.
The coordinator explains that the chemical, physical and biological parameters of the soil interact dynamically and, therefore, an erosion process is conditioned by the unpredictability of variables, which occur in different combinations. Using ANN, it will be possible to model and predict how this process will behave, depending on the interactions of the analyzed characteristics.
“In other traditional models [called deterministic], the algorithm makes some assignments, for example, the higher the humidity, the greater the probability that soil compaction will occur. But, in this correlation, all other measurable biotic and abiotic factors that may be interacting are discarded. The neural network seeks a simultaneous analysis of all these dynamics”, points out Santos.
From this subproject it will be possible to advance the understanding of how an erosion process occurs in a given agricultural area, and estimate the chances of occurrence in other parts of the property and the relationships with aspects of that soil. “When compiling the data, we look for factors in the convergence of interactions and how this can accelerate or slow down the erosion process”, he explains.
As the research continues, the idea is to use aerial images of crops and water collected during precipitation events, aiming to correlate soil cover and surface runoff with the other data used. The soil productivity index will also be calculated, correlated to the erosion prediction.
In the future, the expectation is that this combined model can be applied to the use of images in a smartphone application. “We envision that the farmer can
send photos of a runoff event to understand the area's susceptibility to the erosion process. This must also be associated with a questionnaire, which the producer will feed with soil analysis data. This way, he can have a property scenario”, proposes Santos.
In this sense, the researcher emphasizes that it is essential that rural producers do not stop doing their part. “Science indicates possibilities to alleviate erosion, increase productivity and preserve the production system, but it is the producer who must manage this”, he concludes.
Receive the latest agriculture news by email
Receive the latest agriculture news by email
Data analyzed by FAESP reveal that the volume is 5% lower than the last harvest; The reasons are the lack of rain and reduction of the area under production
