The "drawing-to-learn" pedagogical approach has gained considerable attention and evidence supporting its efficacy for facilitating learning. However, little theoretical or empirical research has addressed issues regarding collaborative drawing-to-learn. In this study, we developed collaborative drawing-to-learn activities and three types of scaffolding to facilitate students' carbon cycling modeling, with the aims of investigating the effects of the activities and scaffolding and identifying significant factors contributing to learning outcomes. Participants were 51 high school students. Mixed methods were employed. Student-drawn models of carbon cycling were collected during the pretests, class sessions, and posttests. Students' prior knowledge of carbon cycling was measured in the pretests, and their self-ratings of self-regulation and co-regulation demonstrated during collaborative drawing were collected. Student action and discussion data were also collected and analyzed for result triangulation. The results indicated that the designed activities and scaffolding were effective in terms of facilitating the students' modeling performances, and students reported and demonstrated satisfactory self- and co-regulation. Moreover, correlation and multiple regression results indicated that prior content knowledge and metacognitive co-regulation are significant direct factors affecting students' collaborative drawing products, whereas metacognitive self-regulation may be mediated by co-regulation to have effects. The study contributes by extending theoretical perspectives on the mechanism of how collaborative drawing may facilitate learning, and by providing examples and empirical benefits of effective collaborative drawing as a pedagogical approach.