کد مقاله | کد نشریه | سال انتشار | مقاله انگلیسی | نسخه تمام متن |
---|---|---|---|---|
84129 | 158861 | 2015 | 11 صفحه PDF | دانلود رایگان |
• Soft computing was applied to agro-meteorological and soil moisture data.
• Soil moisture at 0.3 m, day of year and air temperature were the most relevant inputs.
• Fuzzy rules-based model with only 5 rules estimated ΨstΨst with 86% variance explained.
• Supplied agro-linguistic terms improved the interpretability of the fuzzy rules.
• The model was almost as accurate as artificial neural networks, while being simpler.
Measuring the stem water potential (ΨstΨst), which is an essential parameter for assessing plant water status, is a tedious and labor-consuming task. In this work, hybrid soft computing techniques were applied to design a model able to estimate ΨstΨst based on agro-meteorological and soil water content data. A Takagi–Sugeno–Kang fuzzy inference system (TSK-FIS) was obtained. This kind of model approximates non-linear systems by combining a set of functions local to fuzzy regions described by fuzzy rules. Such models have approximative power and are sufficiently descriptive. Starting from a set of input–output data, inputs relevant to ΨstΨst were automatically selected and fuzzy rules were identified based on the fuzzy clusters found in the data. The rule parameters were optimized by means of a neuro-fuzzy approach. The result was an accurate (86% variance explained) and simple model with five rules that considered soil water content at 0.3 m depth, the day of the year and mean daily air temperature as input variables, confirming the suitability of such approach. In addition, a rule simplification method allowed a consistent agro-linguistic interpretation of the fuzzy sets of the rules: DRY, MOIST and WET for the soil water content, BLOOM, FRUIT GROWTH, EARLY POSTHARVEST and LATE POSTHARVEST for the day of the year, and COLD, MILD and WARM for mean daily air temperature.
Journal: Computers and Electronics in Agriculture - Volume 115, July 2015, Pages 150–160