| Almonds and walnuts are important nut crops in California and were worth about ;A proximal sensor suite consisting of an infrared thermometer, an air temperature sensor, a humidity sensor, a photosynthetically active radiation (PAR) sensor, and an anemometer was developed to measured leaf temperature and other relevant microclimatic information to determine plant water status. A series of experiments were conducted in almond and walnut orchards to study the relationship between data obtained using the sensor suite and SWP measured using a standard pressure chamber. The sensor suite worked well under field conditions in both almond and walnut orchards.;Energy balance based nonlinear regression analysis was used to relate leaf temperature of almond and walnut trees to their respective leaf resistance values for sunlit and shaded leaves. The validation of the calibration equation for almonds resulted in coefficients of determination (R2 values) of 0.79 (shaded) and 0.76 (sunlit). The corresponding R2 values for walnuts were 0.66 (shaded) and 0.55 (sunlit). However, a stepwise selection based multiple linear regression (MLR) analysis that used leaf temperature as the dependent variable yielded empirical models that had better coefficients of multiple determination (R2 values of 0.80 or higher) compared to the energy balance based the nonlinear regression analysis. The leaf conductance was always a significant variable in these empirical models. In addition, the leaf conductance for almond and walnut trees was found to depend on SWP under both sunlit and shaded conditions.;Stepwise selection based MLR models of leaf temperature as functions of SWP, air temperature, relative humidity, PAR, and wind speed were developed and validated for almond and walnut crops under sunlit and shaded conditions with high R2 values. Since, the resultant empirical models for shaded leaves had better R2 values than for sunlit leaves in almond and walnut, the final prediction models based on shaded leaf temperature were used for water stress determination in these two crops. The final prediction models are: Almonds: TL=2.540+0.781 Ta-2.480 SWP-0.012 PAR+0.667 u (R2 = 0.88), Walnuts: T L=-18.730-7.670 SWP+0.981 Ta+0.249 RH+1.396 u (R2 = 0.75).;Three different methods-- multiple linear regression (MLR), stepwise discriminant analysis (SDA), and canonical discriminant analysis (CDA) --- were used for plant water stress classification (as stressed and unstressed). In almonds, critically wrong errors (i.e., stress trees being classified as unstressed trees) were 8.1, 8.8, and 9.0% respectively for the MLR, SDA, and CDA methods. Over-irrigation errors (i.e., unstressed trees being classified as stressed trees) were 10.6, 5.5, and 8.0% respectively for MLR, SDA, and CDA methods. In walnut, critically wrong errors were 12.9, 10.0, and 10.1% respectively for the MLR, SDA, and CDA methods. Over-irrigation errors were 7.7, 12.7, and 9.8 % respectively for the MLR, SDA, and CDA methods.;SDA method performed slightly better in almonds compared to MLR and CDA methods primarily due to lower over-irrigation error. For walnuts, the CDA method performed slightly better than the MLR and SDA methods. However, all three methods resulted in both critically wrong and over-irrigation errors of about 10% or less.;Finally, all three classification techniques were used to create spatial variability maps of plant water stress in almond and walnut. The spatial variability maps developed from all three methods were similar to each other as well as to the map created by actual measured plant water stress using a pressure chamber. |
