Pesticides are increasingly used in modern agriculture and are expected to play a crucial role in protecting crops from pests to meet the growing demand for food and fiber for the foreseeable future. Pesticides, despite their beneficial effects, can have adverse economic and environmental consequences when spray drift occurs. The unused and lost portion of pesticides, which can account for 50% to 60% of sprayed volume, not only pollutes the air, water, and soil, but also results in poor pest control and significant economic losses. As a result, appropriate use of pesticides is a critical aspect for their environmentally and economically sound use. Therefore, the search for a better and more innovative spraying technique as well as developing a mathematical model for pesticide risk assessment and identification of the most important drift-related variables is a timely and important topic for pesticide optimization. Through three topics, this thesis investigates how spray application methods and understanding the spraying process through spray modeling might help reduce spray losses and improve biological efficacy in the context of pesticide optimization. The first theme looks at how magnetic spraying affects spray drift and on-target deposition under field conditions. Spray drift and deposition experiments were conducted to investigate the effects of magnetic spraying on spray drift and deposition in sugarcane plantations. The result showed that magnetic spraying was found to have a considerable influence on drift reduction when compared to traditional knapsack and backpack boom sprayers. The lowest drift values were achieved with the magnetic sprayer with a TeeJetXR110015 nozzle; it was significantly lower than the backpack boom sprayer with both TeeJetXR110015 and TeeJetXR11001 nozzles and the knapsack sprayer. Significant differences between treatments were also observed for on-target deposition at both the top and middle canopies. The highest deposition was obtained by the magnetic sprayer with a TeeJetXR110015 nozzle at both the upper and middle canopies. However, the deposition for the magnetic sprayer coupled with the TeeJetXR11001 nozzle was statistically at par with the knapsack at the iii upper canopy and with both knapsack and backpack boom sprayers with the TeeJetXR110015 nozzle in the middle canopy. None of the application methods, except magnetic sprayer with TeeJetXR110015, gave acceptable spray deposition uniformity. In conclusion, the results clearly showed the potential of magnetic spraying in reducing pesticide drift and improving on-target deposition. The second theme investigates the tradeoffs between spray drift reduction and biological efficacy through a field experiment on sugarcane and wheat crops. In sugar cane crop, for each “experimental site-weed” combination, dose-response curves were applied to estimate the dose required to obtain 70% and 90% weed control (ED70 and ED90) using a magnetic sprayer. The results showed that the weed density was significantly reduced due to different treatments at all assessment periods in comparison to that of the weedy check. The current herbicide dose rate could at least be reduced by 30% without sacrificing total weed control efficacy using magnetic spraying, which effectively controlled broadleaved and grass weeds. The control of sedge weeds was poor even at the full recommended dose. While in the wheat crop trial, a post-emergence application of herbicide (Pallas*45 OD) at 70-85% of the recommended dose with a magnetic sprayer can be used to completely control broad leaf weeds and grasses while enhancing wheat yield. A successful shift from conventional spraying methods to more optimal pesticide usage would necessitate not only the use of innovative application technology but also a thorough grasp of the physical fundamentals of spray drift and drift-causing factors. As a result, the third theme deals with the development of an empirical model for predicting spray drift and understanding the various factors influencing spray drift from handheld sprayers. Two advanced machine learning models, ANN and SVR, were developed for ground drift prediction and compared to three conventional regression models: MLR, GLM, and GNLS. The models were evaluated with five-fold cross validation and external validation. Based on the results, from regression-based models, GLM and GNLS models performed very well when evaluated by cross validation (R 2 = 0.96 and 0.95 and RMSE= 0.70 and 0.82 respectively), while MLR performed less with R 2 of 0.65 and RMSE of 2.25. Simultaneously, ANN and SVR models performed very well, with a R 2 = 0.98 and 0.97 and iv RMSE= 0.58 and 0.71 respectively. Overall, the ANN model performed best compared to the other four, followed by SVR. Therefore, the ANN model is a potentially promising new method for predicting pesticide drift. In conclusion, the new approaches, ANN and SVR based techniques, for drift modelling have showed better predictive power than conventional regression models. Their ability to model complex relationships among the data is a clear benefit in drift modelling where the variability in drift is often affected by several variables and the relationships between drift and predictors are very complicated. These insights will pave better way forward for the application of machine learning toward spray drift modelling.