Brazil is the fourth largest producer of cassava in the world, with climate conditions being the main factor regulating its production. This study aimed to develop agrometeorological models to estimate the sweet cassava yield for the São Paulo state, as well as to identify which climatic variables have more influence on yield. The models were built with multiple linear regression and classified by the following statistical indexes: lower mean absolute percentage error, higher adjusted determination coefficient and significance (p-value < 0.05). It was observed that the mean air temperature has a great influence on the sweet cassava yield during the whole cycle for all regions in the state. Water deficit and soil water storage were the most influential variables at the beginning and final stages. The models accuracy ranged in 3.11 %, 6.40 %, 6.77 % and 7.15 %, respectively for Registro, Mogi Mirim, Assis and Jaboticabal.

The State of São Paulo, Brazil is the major orange producer in the world. The "Valência" orange is one of the most important cultivars for industry with a high efficiency of processed juice. Climate is the main factor of influence for citrus yield and quality and its study is fundamental for understanding the climatic requirements of the crop. The estimation of yield and quality by agrometeorological models helps to understand the effects of climate on crop cycle, besides being an option for orchards planning activities. Understanding the relationships between water deficiencies (DEF), phenological phases, yield (fruits per box (FRBOX) and quality of "Valência" orange grafted on rangpur lime (VACR) are important to improve the water use in the production areas and to provide information about water stress for plants during its cycle. The present study aimed to investigate the influence of monthly DEF on yield and quality parameters of VACR, in order to develop agrometeorological models for the main four producers regions of State of São Paulo, Brazil. Data of 13 years were used for analysis, being the period from 2001 to 2009 used for calibration and from 2010 to 2012 for validation. Multiple linear regression for model construction was used. All the developed agrometeorological models were accurate, ranging the values of mean absolute percentage error (MAPE) of 5.25 to 9.27% for mean annual yield (FRBOX) and 2.74 to 14.14% for quality (RATIO) among all regions. The angular coefficients indicate which phenological phase of VACR is more sensitive to DEF. Bauru and Limeira FRBOX was related to DEF during bud formation and vegetative dormancy. Yield at Bebedouro were related to DEF between vegetative dormancy and flowering and at Matão bud formation. Fruit quality was more sensitive to DEF during maturity at all regions.

ABSTRACT. Forecast is the act of estimating a future event based on current data. Ten-day period (TDP) meteorological data were used for modeling: mean air temperature, precipitation and water balance components (water deficit (DEF) and surplus (EXC) and soil water storage (SWS)). Meteorological and yield data from 1990-2004 were used for calibration, and 2005-2010 were used for testing. First step was the selection of variables via correlation analysis to determine which TDP and climatic variables have more influence on the crop yield. The selected variables were used to construct models by multiple linear regression, using a stepwise backwards process. Among all analyzed models, the following was notable: Palavras-chaves: modelo de cultivo, balanço hídrico, predição, produção.

Forecasting is the act of predicting unknown future events using available data. Estimating, in contrast, uses data to simulate an actual condition. Brazil is the world's largest producer of oranges, and the state of São Paulo is the largest producer in Brazil. The BValência^orange is among the most common cultivars in the state. We analyzed the influence of monthly meteorological variables during the growth cycle of Valência oranges grafted onto BRangpur^lime rootstocks (VACR) for São Paulo, and developed monthly agrometeorological models for forecasting the qualitative attributes of VACR in mature orchard. For fruits per box for all months, the best accuracy was of 0.84 % and the minimum forecast range of 4 months. For the relation between°brix and juice acidity (RATIO) the best accuracy was of 0.69 % and the minimum forecast range of 5 months. Minimum, mean and maximum air temperatures, and relative evapotranspiration were the most important variables in the models.