Since the commercialization and widespread adoption of dicamba-tolerant (DT) soybean cultivars across the United States, numerous cases of off-target damage to non-DT soybean have been reported. Previous studies have focused on understanding the impact of growth stage, dosage, frequency, and duration of dicamba exposure on the severity of symptomology and yield loss. To date, little research has investigated the effect of genetic components on the observed responses. Therefore, this research included three major components including i) the estimation of yield losses caused by prolonged off-target dicamba exposure, ii) the development of a method to differentiate soybean response to dicamba using unmanned-aerial-vehicle-based imagery and machine learning models, and iii) identification of genomic regions associated with soybean response to off-target dicamba exposure. Across 553 soybean genotypes derived from 239 unique bi-parental populations, a yield penalty of 8.8 percent was observed for every increment in damage score on a 1-4 scale with losses as high as 40 percent. Although the interaction between damage and maturity group (MG) significantly affected yield, genotypes showing the most tolerance had similar yields independent of their MG. This indicated that natural tolerance to off-target dicamba may be conferred by physiological mechanisms other than the length of the recovery window. A quadcopter with a built-in RGB camera was used to collect images of field plots at a height of 20 m above ground level. Seven image features were extracted for each plot, including canopy coverage, contrast, entropy, green leaf index, hue, saturation, and triangular greenness index. Classification models based on artificial neural network (ANN) and random forest (RF) algorithms were developed to differentiate the three classes of response to dicamba. Significant differences for each feature were observed among classes and no significant differences across fields were observed. The ANN and RF models were able to precisely distinguish tolerant and susceptible lines with an overall accuracy of 0.74 and 0.75, respectively. Lastly, two models were implemented to detect significant marker-trait associations: the Bayesian-information and Linkage-disequilibrium Iteratively Nested Keyway (BLINK) and a model that allows the inclusion of population structure in interaction with the environment (GxE) to account for variable patterns of genotype responses in different environments. Most accessions (84 percent) showed a moderate response, either moderately tolerant or moderately susceptible, with approximately 8 percent showing tolerance and susceptibility. No differences in off-target dicamba damage were observed across maturity groups and centers of origin. Both models identified significant associations in regions of chromosomes 10 and 19. The BLINK model identified additional significant marker-trait associations on chromosomes 11, 14, and 18, while the GxE model identified another significant marker-trait association on chromosome 15. The significant SNPs identified by both models are located within candidate genes possessing annotated functions involving different phases of herbicide detoxification in plants. Given the widespread adoption of DT systems and potential yield losses in non-DT soybean genotypes, identification of non-DT soybean genotypes with higher tolerance to off-target dicamba may sustain and improve the production of other non-DT herbicide soybean production systems, including the niche markets of organic and conventional soybean.