Estudo da dinâmica de fendas no solo: abordagem morfológica e modelos de representações de aprendizagem

Abstract

The non-linear dynamics of crack in expansive soils makes their typification a challenge, especially in field conditions. To fill this gap, the present study was based on the following objectives: (i) quantify and analyze the cracking process in expansive soil in a semi-arid region under field conditions during the dry and rainy season; (ii) quantify the degree of development of cracks in the soil with different granulometries and different vegetation cover in a hydrographic basin; (iii) estimate the morphometric characteristics of cracks: area, depth and volume and (iv) evaluate the prediction of cracks from environmental data. Initially, the effect of vegetation on the formation of cracks in the soil was monitored in 1m² plots with and without vegetation. Subsequently, in six plots without vegetation cover, the morphometric characteristics of soil cracks (area, depth and volume) and environmental factors were monitored during the period from July 2019 to June 2020, in a watershed of 2.8 ha. (soil moisture, rainfall, potential evapotranspiration and water balance). Precipitation presented a unimodal distribution from December to May, after five months without precipitation. The vegetation showed influence on the connections between cracks when compared to the bare soil. Soil cracks sealed superficially first in plots with predominance of sand under two conditions: (a) soil moisture was equal to or greater than 24%; (b) an accumulated annual precipitation of 109 mm. The plots with the highest content of fine particles (clay and silt) had the cracks sealed under the following conditions: (a) soil moisture equal to or greater than 31%; (b) accumulated precipitation in the wet season above 222 mm. Once sealed, the cracks in the soil resurface, for different soil textures after a 4% reduction in soil moisture. Plots with soils with a higher plasticity index developed more cracks with greater depth and volume. The environmental factors (soil moisture and accumulated potential evapotranspiration from 1 to 10 days) that present high correlation with the morphometric characteristics of the cracks in the soil, showed good performance in modeling the morphometric characteristics of cracks in the soil for the plots above 30% of clay after undergoing training and applied in neural networks. The greatest dispersion of modeled values over predicted values for all morphometric characteristics was found in soils with sand content greater than 40%.