The climate is changing yet many rely on long term average temperature and precipitation data to get an idea of what is “normal” for a field. However, weather is complex and relying on station data for long term averages isn’t the best method. First, station data is valid for that point and that point only. Single-site station data does not represent the spatial coverage needed to understand historic yields in the context of weather. In addition, the maintenance of that station and its reliability is suspect at best. This means you could be dealing with a dataset that is flawed and may be missing data due to poor data quality. 

But there is another way! Weather Decision Technologies has created the ability to interrogate weather data in a unique way that combines the very best of observation platforms from station data, to satellite, to radar for the best possible result. By applying advanced data and meteorological techniques, WDT has generated a massive historical database of daily weather variables around the globe at extremely high resolution. Utilizing past weather information helps seed companies and agriculture retailers understand weather’s role in yield and ultimately return on investment over an entire field vs over a point location or coarse grid.

Use cases for WDT's high resolution gridded dataset includes feed info into disease models, whereas retailers have the opportunity to look back and understand the weather conditions that led to a specific pest or disease outbreak. Also, feeding this weather information allows for seed companies to improve yield by more precisely placing seed in favorable fields with suggested planting dates.

Looking into the past is only a small part of the equation when it comes to understanding how to optimize weather data; whether it’s a seed, chemical, ag retailer, precision ag platform, OEM, agronomist, or grower. The same high resolution datasets that generate the past weather information feed all of WDT’s numerical weather models for forecast information which in turn can feed predictive analytics in agriculture.

Efficient water usage is one of the greatest challenge of 21^st century especially in agriculture that consumes more than 70% of fresh water.  Irrigation methods, which are based on scientific models (such as  Penman-Monteith, Sebal, and Metric models) have the potential to improve on current irrigation practices. Generally, such approaches rely on combining two data sources; satellite data that provide information about the vegetation/biomass and weather that can be used to derive the  evapo-transpiration. The challenge to run satellite and weather based models operationally on continental/global scale is due to enormous volume of data that needs to be processed in real time for daily updates for irrigation recommendations. Towards that end, we developed a big data platform called Physical Analytics Integrated Repository and Services (PAIRS) that curates satellite, soil, topography, land use, and high resolution weather model forecast and combines them to enable real time irrigation forecast to run on continentals scale. The platform automatically ingest new datasets as they become available like the Landsat tiles and the models are seamlessly updated with new data. Here we demonstrate a scalable modeling technique that provides 10 days ahead irrigation forecast at 30 m spatial resolution. We discuss also the integration of the irrigation forecast with variable rate irrigation control system. Variable rate irrigation can reduce up to 20% the water usage compared with uniform irrigation.   The accuracy of the irrigation forecast was tested and demonstrated on various crops situated in Israel, India and USA that proves the applicability of the variable rate irrigation.

Crop models are often fueled by (historical) yield measurements to help better predict the yields at a farm level. However to extend yield predictions to a regional or even global scale, there is a gap in the information required to calibrate such crop models. In addition, the heterogeneity of data obtained from various sources requires extensive pre-processing to provide the necessary inputs for running these crop models. Physical Analytics Integrated Data Repository and Services (PAIRS) is a geo-spatial big data platform that was used in this study to source weather, soil, and satellite data necessary for running crop models. Data curation and sensor error corrections are automatically handled by PAIRS and all data is indexed and stored in a distributed storage system for quick and parallelized access and processing. In recent years, satellite based vegetation indices have been used to derive the key phenological stages of crops and improve the accuracy of yield forecasts. In this work we have studied vegetation index changes for specific crops across multiple years to recognize crops and track the bio mass changes. Specifically winter wheat (Triticum aestivum L.) yield was simulated by coupling PAIRS with the Decision Support System for Agrotechnology Transfer (DSSAT). The model was calibrated by obtaining several phenological data derived from PAIRS. This study successfully demonstrates how to leverage big data technologies for providing the necessary data to improve the accuracy of crop yield forecasts.

Can’t farming be simpler?  Yes…with an Agriculture Operations Center -- we call it the AGOC, and it’s the next big step for precision agriculture.  Leveraging decades of lessons from the US Air Force, the AGOC provides the ability to schedule, execute, collect, consolidate, and distribute all the support a farmer needs from satellites, piloted aircraft, unmanned aircraft, sensing, modeling, and analysis…all packaged into “one stop shopping.”  The result is a plan that makes sense, fused data and information creating a higher level of understanding for the grower, and the kind of user-friendly products a grower wants.  That’s the best of Precision Agriculture.