Módulo computacional para cálculo da necessidade de nitrogênio, fósforo e potássio a partir de suas disponibilidades no solo

Abstract

The spatial variability of soil attributes, deriving from several factors, directly affects crop yields. The main objective of precision agriculture is to provide technologies to enable the correct management of this variability, contributing to the increase in agricultural production. Therefore, it is necessary to use computational tools capable of properly estimating the demand for crops at each point of the productive areas, considering the variables involved, and recommending the appropriate application for each point. The AgDataBox platform is a joint initiative of LAMAP (UNIOESTE, Cascavel Campus) and AgriLab (UTFPR, Medianeira Campus) which aims to develop a complete environment that allows the management of agricultural data. This platform already has several implemented functionalities, but it did not yet have a module for recommending fertilizers. In this sense, the focus of this work was to develop the AgDataBox-Map-AplicNutrient computational module and integrate it with the AgDataBox web platform for integrating data, software, procedures and methodologies for digital agriculture. The dissertation was divided into three articles: (i) ARTICLE 1 - SYSTEMATIC LITERATURE STUDY: identified 12 methods for recommending nitrogen (N), eight for recommending phosphorus (P) and seven for potassium (K), in addition to five computer programs for precision agriculture that carry out the site-specific recommendation of fertilizers; (ii) ARTICLE 2 - AGDATABOX-MAP-APLICNUTRIENT MODULE: presents the application, which allows the user to import data from chemical analysis of the soil, make the recommendation of nutrients N, P and K, generate thematic recommendation maps and export the maps. The results obtained by the app were equivalent to those obtained by calculations performed in an electronic spreadsheet, but in a faster and easier way, especially for its integration with the AgDataBox platform. (iii) ARTICLE 3 - COMPARISON OF FERTILIZER RECOMMENDATION MAPS: this work evaluated and compared four types of drawing of nutrient application maps: traditional thematic maps (TM) with equal interval division (TTM-EI) and quantile (TTM-QT), application zone maps (AZ) and rectified application zones (AZr). AZs were evaluated using the following indices: Variance Reduction (VR%), Fuzziness Performance Index (FPI), Modified Partition Entropy (MPE), Enhanced Cluster Validation Index (ICVI), Improved Cluster Validation Index (ICVI), Smoothness Index (SI%), Fragmentation Index (FI%), and Global Quality Index (GQI); and the comparison of the maps was performed using the Kappa index (Kp). The best design for AZs was obtained with two classes. The results showed that in the creation of TMs, the division of classes by quantiles is more efficient than the division in equal intervals in most cases and can be an alternative to the creation of AZs when the data are normal.