Forecasting Site-Specific Leaf Wetness Duration for Input to Disease-Warning Systems

Weather Based Strawberry Yield Forecasts at Field Scale Using Statistical and Machine Learning Models

Strawberry is a high value and labor-intensive specialty crop in California. The three major fruit production areas on the Central Coast complement each other in producing fruits almost throughout the year. Forecasting strawberry yield with some lead time can help growers plan for required and often limited human resources and aid in making strategic business decisions. The objectives of this paper were to investigate the correlation among various weather parameters related with strawberry yield at the field level and to evaluate yield forecasts using the predictive principal component regression (PPCR) and two machine-learning techniques: (a) a single layer neural network (NN) and (b) generic random forest (RF). The meteorological parameters were a combination of the sensor data measured in the strawberry field, meteorological data obtained from the nearest weather station, and calculated agroclimatic indices such as chill hours. The correlation analysis showed that all of the parameters were significantly correlated with strawberry yield and provided the potential to develop weekly yield forecast models. In general, the machine learning technique showed better skills in predicting strawberry yields when compared to the principal component regression. More specifically, the NN provided the most skills in forecasting strawberry yield. While observations of one growing season are capable of forecasting crop yield with reasonable skills, more efforts are needed to validate this approach in various fields in the region.

Effects of leaf wetness duration and temperature on infection of Prunus by Xanthomonas arboricola pv. pruni

Xanthomonas arboricola pv. pruni is the causal agent of bacterial spot disease of stone fruits and almond. The bacterium is distributed throughout the major stone-fruit-producing areas of the World and is considered a quarantine organism in the European Union according to the Council Directive 2000/29/EC, and by the European and Mediterranean Plant Protection Organization. The effect of leaf wetness duration and temperature on infection of Prunus by X. arboricola pv. pruni was determined in controlled environment experiments. Potted plants of the peach-almond hybrid GF-677 were inoculated with bacterial suspensions and exposed to combinations of six leaf wetness durations (from 0 to 24 h) and seven fixed temperatures (from 5 to 35°C) during the infection period. Then, plants were transferred to a biosafety greenhouse, removed from bags, and incubated at optimal conditions for disease development. Although leaf wetness was required for infection of Prunus by X. arboricola pv. pruni, temperature had a greater effect than leaf wetness duration on disease severity. The combined effect of wetness duration and temperature on disease severity was quantified using a modification of the Weibull equation proposed by Duthie. The reduced-form of Duthie’s model obtained by nonlinear regression analysis fitted well to data (R = 0.87 and R²_adj = 0.85), and all parameters were significantly different from 0. The estimated optimal temperature for infection by X. arboricola pv. pruni was 28.9°C. Wetness periods longer than 10 h at temperatures close to 20°C, or 5 h at temperatures between 25 and 35°C were necessary to cause high disease severity. The predictive capacity of the model was evaluated using an additional set of data obtained from new wetness duration-temperature combinations. In 92% of the events the observed severity agreed with the predicted level of infection risk. The risk chart derived from the reduced form of Duthie’s model can be used to estimate the potential risk for infection of Prunus by X. arboricola pv. pruni based on observed or forecasted temperature and wetness duration.

Use of an Empirical Model to Estimate Leaf Wetness Duration for Operation of a Disease Warning System Under a Shade in a Ginseng Field

Ginseng foliar diseases are typically controlled by spray application using periodic schedules. Few disease warning systems have been used for effective control of ginseng foliar diseases because ginseng is grown under shade nettings, which makes it difficult to obtain weather data for operation of the disease warning system. Using weather data measured outside the shade as inputs to an empirical leaf wetness duration (LWD) model, LWD was estimated to examine if operation of a disease warning system would be feasible for control of ginseng foliar diseases. An empirical model based on a fuzzy logic system (fuzzy model) was used to estimate LWD at two commercial ginseng fields located in Gochang-gun and Jeongeup-si, Korea, in 2011 and 2012. Accuracy of LWD estimates was assessed in terms of mean error (ME) and mean absolute error (MAE). The fuzzy model tended to overestimate LWD during dew eligible days whereas it tended to underestimate LWD during rainfall eligible days. Still, daily disease risk ratings of the TOM-CAST disease warning system, which are derived from estimates of wetness duration and temperature, had a tendency to coincide with that derived from measurements of weather variables. As a result, spray advisory dates for the TOM-CAST disease warning system were predicted within ±3 days for about 78% of time windows during which the action threshold for spray application was reached. This result suggested that estimates of LWD using an empirical model would be helpful in control of a foliar disease in a ginseng field. It was also found that a spray application time model using meteorological observations may prove successful without the requirement of leaf wetness sensors within the field. Development of empirical correction schemes to the fuzzy model and a physical model for LWD estimation in a ginseng field could improve accuracy of LWD estimates and, as a result, spray advisory date prediction, which merits further studies.

Assessment of mycotoxin risk on corn in the Philippines under current and future climate change conditions

This study attempts to assess the risk of mycotoxins (aflatoxins and fumonisins) contamination on corn in the Philippines under current and projected climate change conditions using fuzzy logic methodology based on the published range of temperature and rainfall conditions that favor mycotoxin development. Based on the analysis, projected climatic change will reduce the risk of aflatoxin contamination in the country due to increased rainfall. In the case of fumonisin contamination, most parts of the country are at a very high risk both under current conditions and the projected climate change conditions.

Variability Among Forecast Models for the Apple Sooty Blotch/Flyspeck Disease Complex

Several disease forecast models have been developed to guide treatment of the sooty blotch and flyspeck (SBFS) disease complex of apple. Generally, these empirical models are based on the accumulation of hours of leaf wetness (leaf wetness duration [LWD]) from a biofix at or near the phenological growth stage petal fall, when apple flower petals senesce and drop. The models recommend timing of the initial fungicide application targeting SBFS. However, there are significant differences among SBFS forecast models in terms of biofix and the length of LWD thresholds. A comparison of models using a single input data set generated recommendations for the first SBFS fungicide application that differed by up to 5 weeks. In an attempt to improve consistency among models, potential sources for differences were examined. Leaf wetness (LW) is a particularly variable parameter among models, depending on whether on-site or remote weather data were used, the types of sensors and their placement for on-site monitors, and the models used to estimate LW remotely. When SBFS models are applied in the field, recommended treatment thresholds do not always match the method of data acquisition, leading to potential failures. Horticultural factors, such as tree size, canopy density, and cultivar, and orchard site factors such as the distance to potential inoculum sources can impact risk of SBFS and should also be considered in forecast models. The number of fungal species identified as contributors to the SBFS disease complex has expanded tremendously in recent years. A lack of understanding of key epidemiological factors for different fungi in the complex, and which fungi represent the most challenging management problems, are obvious issues in the development of improved SBFS models. If SBFS forecast models are to be adopted, researchers will need to address these issues.

Evaluation of a Warning System for Controlling Primary Infections of Grapevine Downy Mildew

A warning system based on (i) a model that simulates the development of all cohorts of Plasmopara viticola oospores, from oospore germination to infection; (ii) short-term weather forecasts; and (iii) a mobile phone short message system was tested in Northern Italy, from 2006 to 2008. An unsprayed control was compared with a "Warning A" treatment (WA, fungicides were applied whenever the warning system predicted an infection period), a "Warning B" treatment (WB, fungicides were applied as in the WA treatment but only when the relative dimension of any oospore cohort predicted by the model exceeded a threshold), and a "grower" treatment (fungicides were applied according to a conventional schedule). Average disease incidence on leaves was reduced by up to 90% in sprayed plots compared with unsprayed plots. On bunches, efficacy was always >90% at fruit set; when most berries were touching, efficacy was higher for the WA (96%) than for grower (89%) and WB (85%) treatments. On average, 6.8 fungicide sprays were applied following the grower's schedule; use of the warning system reduced applications by about one-half (WA treatment) or two-thirds (WB treatment). The grower's schedule had an average cost of 337 €/ha; the average saving with the WA and the WB treatments was 174 and 224 €/ha, respectively.

A Site-Specific, Weather-Based Disease Regression Model for Sclerotinia Blight of Peanut

In North Carolina, losses due to Sclerotinia blight of peanut, caused by the fungus Sclerotinia minor, are an estimated 1 to 4 million dollars annually. In general, peanut (Arachis hypogaea) is very susceptible to Sclerotinia blight, but some partially resistant virginia-type cultivars are available. Up to three fungicide applications per season are necessary to maintain a healthy crop in years highly favorable for disease development. Improved prediction of epidemic initiation and identification of periods when fungicides are not required would increase fungicide efficiency and reduce production costs on resistant and susceptible cultivars. A Sclerotinia blight disease model was developed using regression strategies in an effort to describe the relationships between modeled environmental variables and disease increase. Changes in incremental disease incidence (% of newly infected plants of the total plant population per plot) for the 2002-2005 growing seasons were statistically transformed and described using 5-day moving averages of modeled site-specific weather variables (localized, mathematical estimations of weather data derived at a remote location) obtained from SkyBit (ZedX, Inc.). Variables in the regression to describe the Sclerotinia blight disease index included: mean relative humidity (linear and quadratic), mean soil temperature (quadratic), maximum air temperature (linear and quadratic), maximum relative humidity (linear and quadratic), minimum air temperature (linear and quadratic), minimum relative humidity (linear and quadratic), and minimum soil temperature (linear and quadratic). The model explained approximately 50% of the variability in Sclerotinia blight index over 4 years of field research in eight environments. The relationships between weather variables and Sclerotinia blight index were independent of host partial resistance. Linear regression models were used to describe progress of Sclerotinia blight on cultivars and breeding lines with varying levels of partial resistance. Resistance affected the rate of disease progress, but not disease onset. The results of this study will be used to develop site- and cultivar-specific spray advisories for Sclerotinia blight.

Integrating Disease Thresholds with TOM-CAST for Carrot Foliar Blight Management

Cercospora carotae and Alternaria dauci cause foliar blight on carrot (Daucus carota subsp. sativus) and are managed with fungicides to prevent yield loss. Sprays are initiated prior to disease symptoms and reapplied frequently, but some of these applications may not be necessary when the threat of disease is low. Delaying the initial fungicide spray until disease symptoms are observed and applying subsequent sprays according to the TOM-CAST disease forecaster may reduce fungicide inputs. The objective of this 2-year field study was to compare preventive, calendar-based application schedules with an integrated management approach that incorporates disease scouting to initiate fungicide application and the TOM-CAST system for timing subsequent fungicide sprays to manage foliar blight on processing, fresh market, and cut-and-peel carrot cultivars in Michigan. Applications of the fungicides chlorothalonil alternated with azoxystrobin were made prior to disease symptoms (0% blight) or when the foliage became blighted at a trace, 5%, or 10% severity level. Fungicides were reapplied every 7 or 10 days or according to TOM-CAST using disease severity value (DSV) thresholds of 15, 20, or 25. Initiating fungicide treatment at a trace level of disease and timing subsequent sprays according to the TOM-CAST 15-DSV forecaster was comparable to the preventive, calendar-based fungicide regime. One to five fewer applications were needed, while fungicide costs were reduced by $21 to $141 per hectare, when spraying at the trace disease threshold and reapplying according to the TOM-CAST 15-DSV program compared with the 7- or 10-day intervals initiated at 0% blight. Fungicide programs initiated at 5 or 10% leaf blight often provided less control than programs initiated at 0% and trace disease. This study highlights the importance of initiating a fungicide program prior to advanced foliar blight and validates the TOM-CAST 15-DSV forecaster for managing Cercospora leaf spot and Alternaria leaf blight in three carrot production systems.