Reducing use and dependency on pesticides in Denmark has been driven by political action plans since the 1980ies, and a series of nationally funded accompanying R&D programs were completed in the period 1989-2006.

One result of these programs was a decision support system (DSS) for integrated weed management. The 4^th generation (2016) of the agro-biological models and IT-tools in this DSS, named IPMwise. The concept of IPMwise is to systematically exploit that:

• occurrence of weeds differs in time and space
• weeds species differ in terms of need for control and susceptibility to herbicides

This knowledge combined with optimization algorithms enables reductions of herbicide use, without compromising agronomic requirements for weed control. With field scouting of weed species, sizes, and numbers, IPMwise can:

• Quantify needs for control
• Identify accompanying single herbicide products and calculate dose rates using a dose-response function parameterized using publicly available efficacy data while respecting herbicide legal restrictions.
• Compose and optimize (for cost, TFI or similar) 2-4 way tank mixtures and suggest adjuvants.
• Recommend specific mechanical control options in cases where herbicides are less competitive
• Assist herbicide resistance management

Developments began with the construction and field testing of many prototypes in Danish cereal crops. By 2021, 30 crops, 120 weed species and 108 herbicides have been integrated. Field validation experiments show no yield loss, satisfactorily low residual weed infestation and a 40% reduction in herbicide input. In crops grown with greater row spacing e.g., sugar beet and maize, a 20% reduction in herbicide input was found.

In the 2000s, initiatives were taken to implement the DSS concept in other countries. While retaining original mathematical functions and decision-making algorithms, complete adaptations of all agronomic, legal, and linguistic content were made to construct national versions. In all countries, a complete range of weeds and herbicide products was included. Starting with major crops, the number of included crops gradually increased, with the current figures being: Denmark 30, Norway 4, Germany 3 and Spain 19. Across all these countries and crops, the potential for herbicide reduction, documented by field validation trials, was 20- 40% of the cost, as compared to 'local best practice' treatments. Recognition in all 4 countries of this DSS as a professional reference point shows that this DSS has the potential for scaling.

However, the need for manual weed scouting has been identified as a major obstacle. This problem was addressed in 2017-2020 in the RoboWeedMaps project, which aims to replace manual weed scouting with systematic photographing and automatic determination of weed infestations with a specially trained deep learning framework. When this system becomes operational, additional potent opportunities for both field- and site-specific weed control arise.