Otimização de modelos para a estimativa do armazenamento de água no solo em cultivo de feijão irrigado

Abstract

Common bean productivity in the Northeast region of Brazil (NEB) is considered low compared to other regions. One of the limiting factors for agricultural production, especially in the agroecological regions included in the NEB, is the irregular seasonal distribution of rainfall. To reduce the effects of water deficit, irrigation is an essential practice for a good agronomic performance of the crop, in addition to maintaining competitive bean productivity in the agricultural sector, as irrigated agriculture reduces risks and provides an increase in productivity. Thus, it is necessary to know the storage of water in the soil, responsible for supplying the demand of the crop. Models for estimating soil water storage, based on meteorological data and soil and crop parameters, stand out in relation to direct measurements, due to the possibility of extrapolation and greater spatial representation. Thus, the objective of this research was to optimize and evaluate the water balance determined by the Thornthwaite and Mather (1955) method adapted for agricultural crops and the FAO-56 method with the KcDual approach, in which it is possible to estimate soil moisture. Agrometeorological and crop data were obtained in a field experiment conducted in the municipality of Rio Largo, Alagoas (09°28'02" S; 35°49'43" W; 127 m) from 17th November 2015 to 1st February 2016, with the bean crop, rosinha variety. The experiment had a sprinkler irrigation system in which 25%, 50%, 75%, 100%, 125% and 150% of crop evapotranspiration (CET) were applied. The models were implemented in a software modeling environment, OpenModel©, where the statistical analysis of the performance of the models and the optimization of parameters of the culture (Kc, Kcb and f) and of the soil were also carried out (PMP) by the Levenberg-Marquardt method (LM). Statistical analysis was based on the coefficient of determination (r2) between estimated and observed water storage, the Willmott concordance index (d) modified by the NashSutcliffe index (NI) and the root mean square error (RMSE). It was concluded that the methods can be used to estimate the ARM, as they presented good precision for the estimates made in bean cultivation in Rio Largo, AL. However, adjustments are necessary considering the characteristics of the culture, soil and climate of the region. LM inverse modeling optimization returned average Kc (0,95), initial (0,20) and average (1,08) Kcb values, f (0,23) and PMP (0,06 – 0,16) within the from literature found, however initial Kc showed high values (1,65). Therefore, it is recommended to test other inverse modeling methods. After a period of water stress and with the return of water infiltration into the soil, both methods present underestimated ARM estimates and when the system presented a new dry period, the methods returned overestimated values. It is considered to evaluate restrictions present in the model used, which does not consider some processes involved in the system. The methods presented greater errors in the estimation of ARM in the treatments in water deficit and excess of water by irrigation. The BH by the method of Thornthwaite and Mather (1955) presented smaller estimation errors and, therefore, it can be indicated for analyzes of the water storage for the bean crop in similar soil and climatic conditions. flow of water for agricultural cultures.

Key words: Modeling, Phaseolus vulgaris L., irrigated agriculture, water balance.