INTRODUCTION
Watercress (Nasturtium officinale R. Br.) is an economically important vegetable crop in Hawaii. Yield of watercress varies seasonally with the best growth occurring during the cool, wet season from October to April. Seasonal fluctuations in watercress production concern growers because of their need to provide a consistent supply to their markets.
A crop model that estimates yield can help growers determine if they can meet the market demand. A model is an equation or equations that describe how a plant grows. In addition, crop models can help growers make decisions about the management of their crop.
PROCEDURE
Yield
A study to determine the effect of weather on the yield of 'Sylvasprings' watercress was conducted at a commercial farm in Aiea. 'Sylvasprings' is also known locally as 'English'. This strain has characteristic shiny, dark green leaves approximately 20 percent larger than those of local strains, and round to oval leaflets. it does not flower during the summer. Standard Hawaiian production practices were followed.
Yield data were obtained from eight randomly selected beds each 40 feet by 80 feet. The number of harvest per bed ranged from five to eight per year. All data were pooled for analysis, giving a total of 50 observations. The watercress was harvested when it reached a minimum of 15 inches above the water level in the bed. The leaves were fully expanded and dark green with stems approximately ¼ inch in diameter. In this study, the crop cycle, i.e., from planting to harvest, averaged 50 days with a maximum of 76 days.
Model Development
Daily maximum and minimum air temperatures were obtained from the nearby Honolulu International Airport (five miles away). Sunlight readings were taken at the University of Hawaii Manoa campus (10 miles away). Sunlight values at the University were found to be within two percent of those at the airport. The model was developed by regression analysis using the yield and weather data.
RESULTS AND DISCUSSION
Yield
Yield of watercress was related to the amount of sunlight received (cal/cm2) during the crop cycle. The equation that best described the relationship between sunlight and yield was:
YIELD = -0.0000011(CAL²) + 0.05(CAL) - 184.51
where YIELD is the yield of watercress (kg fresh weight/bed) and CAL is the solar radiation received during a crop cycle (cal/cm²). The coefficient of multiple determination (R²) was 0.32, and the probability value (P) was 0.001. The R² value indicates the amount of variation in yield (32 percent) accounted for by the equation.
There was an optimum quantity of approximately 22,700 cal/cm² of sunlight required for maximum yield. When the amount of sunlight received during a crop cycle was less than or greater than this optimum, yield was correspondingly reduced. Table 1, derived from the previous equation, shows the predicted yield of watercress relative to the amount of sunlight received during the crop cycle.
Optimum sunlight during the spring may explain the improved yields observed by farms from late February until early May. During this period, plants exhibit more branching, and the crop is thicker. High calorie/cm² accumulation during the summer appears to reduce yield because the sunlight is not optimum.
Further, during early summer (mid-May to mid-July), yields may be reduced because the lengthening photoperiod (day length) causes the crop to slow its growth and undergo physiological changes for flowering. Watercress is a long-day plant and does not normally flower in Hawaii. However, vegetative growth in 'Sylvasprings is reduced during early summer as the growth habit of the plant changes--increased adventitious rooting at the internodes, less stem branching, and leaves becoming more pointed. Other strains of watercress that flower in Hawaii also show these characteristics.
The greatest variation in yields occurred during the summer months. One reason is that water is very sensitive to the volume or quantity of water in the beds. Water volume decreases during the summer months, leading to uneven distribution of water in the beds and more variable yields. Water also acts as a coolant for watercress, and uneven water flow causes differences in water temperature in the beds which affect yield.
Testing the Accuracy of the Model
The accuracy of the model was tested on data from a different group of 16 randomly selected beds each 40 feet by 80 feet. All but one of the observed yield values were within the 95 percent confidence interval predicted by the model. The average percentage error in predicting yields of the 16 test beds was 19 percent. This indicates that the model can predict yield but only with a ± factor.
The large variation in yield occurring during the summer months was caused primarily by six observations with values of approximately 200 kg/bed or less. The six points occurred at the extremes of the sunlight range in this experiment, and thus greatly influenced the equation describing the relationship between yield and sunlight. Because the equation accounts for the extremes in sunlight and the large variation in yield that occurred during late spring and summer (May to September), growers can expect the model to be more accurate at this time of the year. Similarly, if those six observations taken during the summer were removed, there would be little relationship between yield and solar radiation. Therefore, the model will have little usefulness during October to April.
Discrepancies between the observed and predicted values for yield may be due to a number of reasons. First of all, weather data from sites other than the farm were used in this study. More accurate weather data could be obtainee from weather instruments located at the farm. Secondly, the deviations in the model predictions may be because the model accounted for only a portion of the variation in yield (R² = 0.32). This indicates that other variables need to be investigated to determine their effects on yield. Inclusion of these variable sin the model would likely increase the R² value and improve the accuracy of the model. These variables include water temperature, photoperiod, and relative humidity.
Limitations of the Model
The model was developed for the 'Sylvasprings' or 'English' strain of watercress, which is becoming commonly grown in Hawaii and may not be applicable to other strains. The model is limited to the Pearl Harbor Basin from Aiea to Waipahu, where 95% of the state's watercress is produced. The accuracy of the model for other locations is not known. The model is applicable to wetland watercress areas, but may not be applicable to dryland areas. As described earlier, the model may be useful only during the late spring and summer months.
Temperature does not appear to be a limitation in predicting yield of watercress when grown under overhead sprinkler culture. Daily maximum, minimum, and average air temperatures were not correlated with yield. This model was developed for overhead sprinkler culture watercress, which is the standard for the industry.
Table 1. Relationship of watercress yield to total solar radiation received during a crop cycle. Crop cycle is time from planting until harvest. Harvest date is detemined visually by the grower.
Solar radiation Yield (lb Percent of
received in a fresh weight/ maximum
cycle (cal/cm2) square foot of bed) potential yield
_______________________________________________________
10,000 0.14 54
12,000 0.18 67
14,000 0.21 78
16,000 0.23 87
18,000 0.25 94
20,000 0.26 98
22,000 0.26 100
24,000 0.26 100
26,000 0.26 97
28,000 0.24 92
30,000 0.22 85
32,000 0.20 75
34,000 0.17 64
36,000 0.13 49
38,000 0.09 33
40,000 0.04 14
_______________________________________________________