Articles
NEURO-FUZZY MODELING OF TRANSPIRATION RATES OF GREENHOUSE TOMATOES UNDER TEMPERATE WEATHER CONDITIONS OF CENTRAL MEXICO
Article number
1037_42
Pages
345 – 352
Language
English
Abstract
Cultivation of greenhouse tomatoes (Solanum lycopersicum L.) has been increasing smoothly during the last 20 years in México.
However, only under high-tech conditions crop irrigation is based on information and measurements of the climate inside and outside the greenhouse.
To optimize water and nutrients supply is necessary not only the use and development of crop transpiration models but also the collection of detailed and accurate measurements from the environment inside the greenhouse and also the actual crop status.
An experiment was carried out, during the summer days of 2011 in order to measure tomato crop transpiration rates inside a plastic covered greenhouse, ventilated naturally and under outside temperate weather conditions, located at central Mexico.
Climatic variables global solar radiation, air temperature, wind speed, and relative humidity inside the greenhouse were measured.
Measurements recorded each minute were used to generate neuro-fuzzy models, to predict the crop transpiration rates considering indoor climatic variables as inputs.
The ANFIS (Adaptive Neuro-fuzzy training of Sugeno-type Inference System) was used.
Also several membership functions on the inputs and the outputs were tested to generate the fuzzy inference system.
Both grid partitioning and subtractive clustering were used to generate the initial Sugeno type fuzzy inference system.
A total of 25 (2,400 data) and 24 (2,304 data) days of measurements were used for models calibration and validation, respectively.
Results showed better quality of fitting between predicted and measured tomato crop transpiration rates in case of the neuro-fuzzy model using subtractive clustering.
Main statistics for subtractive clustering and grid portioning were: RMSE on training was 17.85 against 18.80 and R2 was 0.979 against 0.976. In case of model validation RMSE was 28.56 against 31.5, and R2 was 0.96 against 0.94. This work showed that the neuro-fuzzy modeling is a promising approach to predict greenhouse tomato crop transpiration rates.
However, only under high-tech conditions crop irrigation is based on information and measurements of the climate inside and outside the greenhouse.
To optimize water and nutrients supply is necessary not only the use and development of crop transpiration models but also the collection of detailed and accurate measurements from the environment inside the greenhouse and also the actual crop status.
An experiment was carried out, during the summer days of 2011 in order to measure tomato crop transpiration rates inside a plastic covered greenhouse, ventilated naturally and under outside temperate weather conditions, located at central Mexico.
Climatic variables global solar radiation, air temperature, wind speed, and relative humidity inside the greenhouse were measured.
Measurements recorded each minute were used to generate neuro-fuzzy models, to predict the crop transpiration rates considering indoor climatic variables as inputs.
The ANFIS (Adaptive Neuro-fuzzy training of Sugeno-type Inference System) was used.
Also several membership functions on the inputs and the outputs were tested to generate the fuzzy inference system.
Both grid partitioning and subtractive clustering were used to generate the initial Sugeno type fuzzy inference system.
A total of 25 (2,400 data) and 24 (2,304 data) days of measurements were used for models calibration and validation, respectively.
Results showed better quality of fitting between predicted and measured tomato crop transpiration rates in case of the neuro-fuzzy model using subtractive clustering.
Main statistics for subtractive clustering and grid portioning were: RMSE on training was 17.85 against 18.80 and R2 was 0.979 against 0.976. In case of model validation RMSE was 28.56 against 31.5, and R2 was 0.96 against 0.94. This work showed that the neuro-fuzzy modeling is a promising approach to predict greenhouse tomato crop transpiration rates.
Authors
I.L. López-Cruz, A. Ruíz-García, L. Hernández-Larragoiti
Keywords
black-box model, weighting lysimeter, sensors
Online Articles (152)