MEDIUM- AND EXTENDED-RANGE WEATHER AND CLIMATE FORECASTS SCALED AND TESTED FOR IMPROVED IPM DECISION SUPPORT IN US STATES

• Sponsoring Institution: National Institute of Food and Agriculture
• Project Status: COMPLETE
• Funding Source: OTHER GRANTS
• Reporting Frequency: Annual
• Accession No.: 1004996
• Grant No.: 2014-70006-22631
• Cumulative Award Amt.: $240,845.00
• Proposal No.: 2014-07773
• Project Start Date: Sep 1, 2014
• Project End Date: Aug 31, 2017
• Grant Year: 2014
• Program Code: [ARDP]- Applied Research and Development Program

Recipient Organization
OREGON STATE UNIVERSITY
(N/A)
CORVALLIS,OR 97331

Performing Department
Integrated Plant Protection

Non Technical Summary
Weather-driven pest and crop models are used increasingly for whole-farm planning as well as for specific crop and pest management decision making needs. Temperature and rainfall can explain much of the year-to-year variability in the the timing of pest occurrence and outbreaks, and models have become essential tools to make predictions in order to follow through with informed management actions. Currently most pest modeling systems make use of up to a 7-day weather forecast, if any at all, to make future predictions. Beyond that, the models have typically used 30-year average weather, which in the face of rapidly changing climate, often provides a less useful prediction than other options. In fact the National Weather Service (NWS) has longer term weather and climate models that currently are relied upon for mostly qualitative guidance in many realms including agriculture. These NWS and similar forecasts have increasingly shown more skill (i. e. better performance vs. 30-year normals), as weather models and expertise in their application have improved in recent decades. We propose a process that will convert the NWS forecasts into quantitative (and thus usable for pest and crop models) data forecasts and link them with over 100 online models already in service for pest and crop management needs for the U.S. This will provide much more useful and skillful predictions for up to 90 days into the future. We will adapt the NWS forecasts into two versions: one version will be designed to help with temperature driven (degree-day) insect and crop development models. A second version will make use of rainfall patterns needed for moisture and temperature-driven plant disease models. We will test this new adaptive technology for several significant pest forecasting problems including the spotted wing Drosophila, which is a new invasive pest of many fruit crops throughout the US, for potato late blight disease forecasting in the Columbia basin of Washington state, and for U.S. mapping of multiple invasive species pest events, used for both integrated pest management and APHIS plant protection and quarantine (PPQ) pest survey programs. Together these applications will be used to test the potential for longer term forecasts and improving precision of models, and for reducing the uncertainty of agricultural decision making. This should lead to more informed, more profitable, and environmentally sound pest control and other farm planning decisions across a broad array of regions and cropping systems, including organic, sustainable and conventional production approaches.

Animal Health Component

45%

Research Effort Categories

Basic

5%

Applied

45%

Developmental

50%

Classification

Knowledge Area (KA)	Subject of Investigation (SOI)	Field of Science (FOS)	Percent
216	5220	2070	40%
132	2410	1130	30%
132	2410	1170	15%
205	2410	3030	15%

Knowledge Area
205 - Plant Management Systems; 132 - Weather and Climate; 216 - Integrated Pest Management Systems;

Subject Of Investigation
2410 - Cross-commodity research--multiple crops; 5220 - Pesticides;

Field Of Science
2070 - Meteorology and climatology; 1130 - Entomology and acarology; 1170 - Epidemiology; 3030 - Information and communication;

Keywords

decision support systems

invasive pest management

weather forecasting

Goals / Objectives
The goal of this project is to develop new weather forecasting decision support tools that can extend the forecast horizon used by agricultural producers in planning multiple management activities, especially those involving crops and pests, that are affected by the weather. Our specific objectives are to:1. Develop, calibrate, and verify downscaled NOAA NWS 15, 30, and 90 day forecast anomaly grids as daily temperature and precipitation forecasts for multiple IPM and crop management modeling products. 2. Implement, extend, support, and validate the new medium- and extended-term forecasts to serve a variety of IPM challenges including for insects, plant diseases, and for invasive species pest event maps. 3. Document and evaluate adoption and usage of the forecast systems and provide opportunities for further integration into US IPM decision support modeling infrastructures.

Project Methods
Much of the effort for this project involves data processing to convert NWS 8-14, 15-45, and 46-105 day forecast "outlook anomalies" (deviations from 30-year normals) into formats that can be used in a wide array of pest and crop models. First, we will perform statistical conversion from NWS outlook anomalies that are expressed as probabilities above or below normal into actual values of temperature and rainfall in degrees F. and inches rainfall. Second, we will blend, smooth and interpolate the forecasts for the three intervals into daily values. Our team of crop modeling, weather forecasting, mathematical, and statistical experts will devise and evaluate the procedures needed for the conversions. Two types are needed, "smoothed temperature" forecasts that look similar to adjusted 30-year normal data, and "realistic rainfall pattern" forecasts that are generally produced by weather generator programs used to drive crop growth models. We will perform calibration and rescaling exercise to produce the proposed data types, and verify that these two approaches work as intended using archival forecast and weather data, and will document error rates as forecast skill, difference from forecast vs. observed degree-day (temperature heat unit) accumulations, and days prediction error for selected agricultural pest events such as first adult emergence. We will use geographic information systems (GIS) analysis and PRISM climate maps to spatialize the forecasts to a high spatial resolution of ca. 800 meters for the full continental US. These extended forecasts will then be matched to all available public weather stations (at least 16,000 stations in U.S. at last count), for virtual weather station locations (which use interpolated data to create virtual weather stations) and be available as GIS data grids for degree-day mapping needs. The data will be used at OSU and WSU pest management support websites, and, as public data, made available without charge to all other regional and state run agricultural decision support websites wishing to incorporate the forecasts into pest management and related models.Initially the forecasts will be offerred as "experimental" options in all applications at OSU and WSU IPM (integrated pest management) websites. We will focus on three areas to verify and test that the forecasts work as intended and to validate the skill for each pest management problem. First, WSU will be using the "patterned rainfall" version of the forecast to predict late-blight infection risk. Medium term precipitation forecasts have been evaluated in the past for late-blight management planning needs, and this new system is expected to provide an improvement that can both decrease fungicide use and losses from the disease in a broad region of the Columbia basin of SE Washington. Second, new models for spotted wing Drosophila (SWD) development and population build-up can greatly benefit from the extended forecasts by offering an informed-risk tool at critical times in the season, including late spring/early summer ripening crops such as cherries and early blueberries, which, depending on weather patterns, can either escape risk of SWD attack, or be completely rejected by fruit packers and processors if infestation levels exceed certain thresholds. In addition to using the forecasts to drive the risk models to provide longer and more accurate warning of pest pressure, this project will support a part time researcher and technical help to better pinpoint female ovary maturation times needed to improve the models. This again will help with spring infestation timing prediction, and with fall overwintering (reproductive diapause, a form of insect hibernation) behavior, which is not yet fully documented. The third area of focus for development and testing of skill levels of the new forecasts is for Pest Event Maps (PEMs) which use degree-day (heat unit) models to map the timing (dates) of predicted events described by the degree-day models. These PEMs are a new product that will eventually be made available for public use, but currently are being supported and tested under contract with USDA APHIS PPQ (plant protection and quarantine) for certain invasive pests, for example the brown marmorated stink bug and European and Asian gypsy moth. This type of degree-day map is inherently a forecast in that it is often a prediction of a future date at which a pest event will occur, and thus can greatly benefit from a more informed and skillful forecast than just using 30-year averages. Thus these new maps and forecasts will be directly used to help monitor and protect American agriculture from numerous alien and invasive species.The forecasts will also be available for the entire suite of 80+ insect, weed, plant disease, and crop models and a general-use degree-day calculator, and have potential to later be used to drive 21+ hourly-weather based models at the http://uspest.org/wea website. Similarly the forecasts will be available for integration with over 15 models at the WSU DAS pest management website at http://das.wsu.edu. A new crop modeling project (USDA SARE funded CROPTIME project) is planning to add up to 100 new vegetable crop models to the uspest.org system, which will greatly benefit from the longer term forecasts for producers to improve their planting and harvest scheduling.In all of the above described projects, we will document relative error rates of the new forecasts compared to 30-year average data. Forecast skill scores, mean absolute error rates, and days of over- and under-prediction of selected pest events will document how well the forecasts perform. This information will be published and made available to model users to better inform them as to reliability and accuracy of the forecasts for their particular needs.

Progress 09/01/14 to 08/31/17

Outputs
Target Audience:Farmers and other agricultural decision makers who plan farming activities for the season. Researchers and extension personnel interested in climate-science based weather forecasts that extend from 1 to 7 months into the future. Students looking for examples where crop and pest models can use monthly-updated extended weather forecasts. Changes/Problems:The issues encountered during this project include 1) Late receipt of funds, causing delays in project initiation, especially regarding funding received by collaborators. 2) Proposed usage of the NWS standard extended forecast system turned out to be qualitative and unusable; but this led to discovery of the NMME system, which is used to inform the standard NWS forecast, and is quantitative, stable, and amply usable for the goals and objectives of this project. 3) Change in personnel (Co-PI at WSU from Hoogenboom to Grove), which led to delays but have been resolved. 4) Outages in our long-term forecast data archive system, plus the nature of long-term forecasts, has made forecast skill assessment difficult, and delayed by nearly a year. We now are archiving sufficient data to fully evaluate the forecasts. Publication of results, however, remains in the works. What opportunities for training and professional development has the project provided? Nothing Reported How have the results been disseminated to communities of interest?Via workshops, seminars, talks at grower events, and through decision support websites. What do you plan to do during the next reporting period to accomplish the goals? Nothing Reported

Impacts
What was accomplished under these goals? Overview Agricultural crop and pest models lack science-based forecasts extending past short time-frame, numerical weather forecasts. We investigated several options that can help to meet this need, including a NOAA supported effort to combine multiple extended forecasts in the NMME system (North American Multi-Model Ensemble). We focused on translating this system to usable forecast data, which is available out 7 months into the future. We linked these forecasts to the growing suite of agricultural models online at OSU's website < http://uspest.org > and at WSU's website < https://weather.wsu.edu > , and made the data available for others to adopt as well. The extended forecasts were developed to support several individual models, with positive on-going developments. An evaluation of forecast performance is also on-going, but it appears that NMME does have useful skill versus other options that are available to use as extended forecasts such as 10 year average data (described below). Specific Objectives and Methodologies Our specific objectives were to first, translate NOAA NWS forecasts out to 90 days. We used the NMME model, which is an ensemble (average) of multiple models from 7 forecast science groups, and all of these extend not just 90 days but for 7 months. As this system is quantitative and can be translated to meet our needs, we began using this product. One NMME model component, the CFS (vers. 2) model, is an important forecast system that our collaborator, Fox Weather, LLC, elected to use and to include realistic rainfall pattern trends in conjunction with temperatures, but to extend this system for 90 days as the original proposal specified. Our second objective, to develop, calibrate, and verify these forecasts for multiple uses, was completed during 2016-2017 at OSU and the forecasts have been in use since that time. Both WSU and OSU are developing protocols to validate the forecasts in various contexts (discussed below). Our third objective was to make use of the forecasts for several example model needs, including 1) Use in forecasting late blight of potato in the Columbia Basin of WA, 2) Use in the OSU and WSU decision support websites, 3) Use for forecasting spotted wing Drosophila reproductive maturation, and 4) Use in degree-day mapping of invasive species to support APHIS PPQ CAPS (Cooperative Agricultural Pest Surveys). WSU Late Blight Forecasts WSU made use of the Fox-CFSv2 version of the forecast (particularly daily rain forecasts) to help predict the severity and need for protective fungicide treatment of late blight in the Columbia Basin of WA, where over 180,000 acres of potato are produced. During the 2015 season, the late blight information call line was accessed 589 times, in addition to numerous visits to the late blight web site. Late blight was correctly forecasted to be a minor problem during the 2015 season and the number of fungicides applied in commercial fields was reduced by at least six. Extension programs resulted in change of practice by 80% of potato growers. In 2016, late blight forecasting was used to schedule fungicide applications and cultural practices. The information line was called 600 times, plus numerous visits to the web site. A potentially severe outbreak was averted because of the forecast system, as growers followed recommendations and the disease was abated. WSU AgWeatherNet A WSU AgWeatherNet degree-day dashboard with follow-on mobile app versions are in development and scheduled for public release in 2019. The dashboard, once fully implemented, will be used to aid in decision support for late blight of potato, downy mildew of hop, powdery mildew of grape, fire blight of apple and pear, favorability of spray conditions, fruit frost conditions, and for insect and crop phenology modeling. OSU Uspest.org This modeling platform was used at OSU to drive over 69,000 degree-day crop and pest models to forecast crop and pest development during 2017, up from 45,000 model runs back in 2013. Forecast Evaluations Both OSU and WSU are performing statistical evaluations of the forecasts. WSU focused on both evaluation of the extended term and short term Fox-MtnRT 7-day forecast system (in comparison to their own WSU-WRF model forecast system) based on 22 AgWeatherNet stations in WA for the period of Oct 1-31 2015. The Fox-MtnRT was found to generally well but had more temperature overestimation bias than the WSU-WRF model, which was also more expensive to run and maintain. OSU analysis of extended forecast performance compared errors in days versus observed events using degree-days (DDs) for hypothetical 250 and 500 DD events, using 82 different weather station locations throughout OR, WA, and ID, and 8 different forecast issue dates, ranging from July 2016 through Feb. 2018. Results are preliminary, but it appears that for 500 DD events, for the first 2 months, the 10 year average data tended to outperform the NMME, CFSv2, and 30 year normals used as a forecast. For forecasts out 2.6, 3.5, 4.2, and 5.0 months into the future, NMME tended to outperform, with less bias and less absolute error, the other forecasts including the 10 year average. For now NMME along with the other options can provide a range of forecast outcomes that should provide a better planning tool, in serving up a "risk envelop" than any single forecast. Degree-Day Model Forecasts As mentioned above, both WSU and OSU have incorporated the extended forecast system into their weather station-based decision support crop and pest models. OSU has two major interfaces to degree-day models. The newer one, at < http://uspest.org/dd/model >, has been set to use the NMME forecast as the default. This interface also includes, as a default, a newer charting system with all forecasts displayed after the current date, as mentioned above, allowing visualization of a range of outcomes. Modeling of Spotted Wing Drosophila (SWD) In order to determine seasonality of SWD reproduction and egg laying potential, thousands of females were dissected and egg maturity rated over the course of two years through 2015. Of particular interest, mature eggs, although in a very small percentage of females, were found as early as late Feb. and early Mar., whereas no known crop hosts are expected until late April or May in W. OR and WA. Because results indicate wide population variability in reproductive maturation, there are no exact events that have yet been "thresholded" to improve model predictions. A comprehensive update of the Uspest.org SWD phenology model is pending other research. APHIS PPQ - Mapping Technology Support Project The NMME forecast consists of monthly gridded data out to 7 months into the future, and must be converted to daily data for use in degree-day models, including the spatial modeling infrastructure written to support APHIS PPQ CAPS invasive species monitoring program, known as DDRP. The programming to convert these forecasts to formats was completed during May 2016 and have been in continuous production since then.

Publications

• Type: Other Status: Published Year Published: 2015 Citation: Andrews, N., L. Coop, and H. Noordijk. 2015. Scheduling vegetables using degree-days. New crop planning, planting model from Oregon State University. Tilth Producers Quarterly 25:4:1-6. Access online via:
• Type: Conference Papers and Presentations Status: Published Year Published: 2016 Citation: Coop, L. A. Fox, G. Grove, and G. Cook. 2016. Medium and Extended Range Weather and Climate Forecasts Scaled and Teested for IPM Decision Support in US States. Poster presented at NW Climate Conference, Nov. 15, 2016, Stephenson, WA. Online at:
• Type: Conference Papers and Presentations Status: Published Year Published: 2015 Citation: Andrews, N., L. B. Coop, H. E. Noordijk, and J. R. Myers. 2015. Crop Time: Degree-day Models and an Online Decision Tool for the Vegetable Industry. HortScience Supplement. 50:S138. Not avail. Online.
• Type: Conference Papers and Presentations Status: Published Year Published: 2016 Citation: Coop. L., A. Fox, G. Grove, A. Dreves. 2016. Extended forecasts for IPM Decision Making. NIFA-CPPM-ARDP Grant Report at WERA-1017: Western Region IPM Coordinators Meeting. July 8, 2016. Boise, ID.
• Type: Conference Papers and Presentations Status: Published Year Published: 2018 Citation: Coop, L. A. Fox, and P. Jepson. 2018. Weather and Climate driven models for IPM and invasive species management. Poster presented at 9th International IPM Symposium, Mar. 21, 2018, Baltimore, MD. Online at:
• Type: Conference Papers and Presentations Status: Published Year Published: 2017 Citation: Coop, L. 2017. Weather Models and Predictive Tools for IPM. Pesticide Stewardship Conference and Recertification Course, Univ. Idaho Extension. Nov. 30, 2017. Boise, ID. 1 hr invited talk.
• Type: Websites Status: Published Year Published: 2015 Citation: Coop, L. B., D. Upper, and N. Andrews. 2016. CROPTIME: phenology models to schedule vegetable plantings and harvests. Version 1.01. Oregon State University Integrated Plant Protection Center Web Site: [first version online 2015]
• Type: Websites Status: Published Year Published: 2016 Citation: Coop, L., D. Debrito, D. Upper. 2016. MyPest Page: Hourly Weather, Plant Disease Risk, and Degree-day/Phenology Models. Integrated Plant Protection Center Web Site Publication E. 16-01-1: [first version online 2010]
• Type: Websites Status: Published Year Published: 2016 Citation: Coop, L. B. 2016. Online phenology degree-day models. 2016 version. Oregon State University Integrated Plant Protection Center Web Site: [first version online 2013]
• Type: Conference Papers and Presentations Status: Published Year Published: 2017 Citation: Coop, L. 2017. Web based decision tools for pest management: New and Used. Pesticide Recertification Course. Jan 24, 2017. Central Point, OR. 1 hr invited talk.
• Type: Conference Papers and Presentations Status: Published Year Published: 2016 Citation: Coop, L. and N. Andrews. 2016. Introducing and Using CROPTIME: Forecast Options for DD Models. Hands-on computer workshop. Mar. 14, 2016. Aurora, OR.
• Type: Conference Papers and Presentations Status: Published Year Published: 2016 Citation: Coop, L. and N. Andrews. 2016. Weather forecasting (long-term forecasts) and future capacity for the modeling system and user interface. In: Introducing and Using CROPTIME: Vegetable Crop Schedule with Degree-Days. 2.5 hr lecture and hands-on computer workshop. 2016 Small Farms Conference. Feb. 20, 2016. Corvallis, OR.
• Type: Conference Papers and Presentations Status: Published Year Published: 2016 Citation: Coop, L. 2016. Integrated Pest Management as it Relates to Climate. Blue Mountain Horticulture Society Annual Meeting. Feb. 10, 2016. Milton Freewater, OR.
• Type: Journal Articles Status: Published Year Published: 2015 Citation: Johnson, D. A., Cummings, T. F., and Fox, A. D. 2015. Accuracy of rain forecasts for use in scheduling late blight management tactics in the Columbia Basin of Washington and Oregon. Plant Dis. 99:683-690