Poverty can result from complex social-ecological interactions where persistent feedback loops create resistant, unsustainable states. In dryland regions, agricultural innovations intended to break poverty traps can often neglect long-term environmental consequences, leading to a reinforcing cycle of degradation and poverty. This study investigates how cross-level dynamics in agricultural innovation systems generate and sustain poverty traps. We ask: (i) How do poverty traps emerge in agricultural innovation systems? (ii) What characterizes agents experiencing these traps? (iii) How can traps be avoided or overcome? We combine dynamical systems modeling (DSM) and agent-based modeling (ABM) to analyze poverty trap emergence. DSM uses bifurcation analysis to reveal system-level dynamics under two innovation scenarios: low-impact (“gentle”) and high-impact (“strong”). ABM simulates these scenarios, tracking agent attributes across runs and mapping them onto DSM parameter space to identify producers and innovators in poor or non-poor states. Comparing agent outcomes with DSM parameter space identifies characteristics of poor and non-poor states. Together, DSM captures system dynamics while ABM reflects heterogeneity, enabling targeted interventions to escape poverty traps. Under gentle innovations, poverty and well-being depend on thresholds in innovation efficiency, funding, and desire: below thresholds, poverty is inevitable, at intermediate levels, outcomes depend on farmers' initial conditions and above thresholds, all reach well-being. Strong innovations carry higher ecological risks, with traps arising whenever thresholds are unmet. Low efficiency traps all farmers with fragile bistability and oscillating well-being at higher efficiencies. Low innovation funding and desire creates poor equilibria with stable well-being at higher levels. Improving innovation efficiency, through stronger knowledge efficiency (understanding producers' needs), greater innovation demand, and higher capital efficiency (better use of resources), increases the effectiveness of innovations and enables producers to escape poverty traps. Similarly, increasing innovation funding and demand for low-environmental-impact agricultural technologies supports pathways out of poverty by simultaneously improving income, ecological indicators, and crop production. This study highlights the critical role of agricultural innovation in shaping poverty trap dynamics and environmental outcomes. By focusing on cross-level interactions between micro-level producers and meso-level innovators, the study demonstrates how these interactions can create or sustain poverty traps. It emphasizes the importance of ecological feedback for understanding the long-term effects of interventions aimed at reducing poverty. Finally, it identifies pathways for breaking poverty traps that go beyond low-impact innovations, highlighting the need for systemic, coordinated interventions to achieve sustainable and resilient agricultural development. • Poverty traps emerge from social-ecological feedbacks that create resistant, unsustainable states. • We develop a hybrid DSM–ABM framework to model cross-level agricultural innovation dynamics. • Under gentle innovations, low innovation efficiency, funding, and desire lead to globally stable poverty traps. • Under strong innovation, high innovation efficiency traps farmers in fragile or oscillating well-being. • Targeted, scale-sensitive interventions can support sustainable poverty alleviation and system resilience.

Herbicide resistance (HR) has been a major management challenge and ongoing threat in cropping regions over several decades, particularly in the time since the widespread adoption of no-tillage cropping systems. Despite extensive and severe development of resistance in Australian cropping regions, growers have been able to maintain control of new weed populations by introducing new herbicides and non-herbicide practices. These integrated weed management (IWM) practices have helped farming systems maintain productivity and adapt to changing environmental and agronomic conditions. Earlier studies found that the adoption of these alternative practices was associated with a reaction to the initial development of forms of HR on a farm. Practices include those classified as harvest weed seed control (HWSC) that kill potentially herbicide resistant weed seeds and reduce overall weed seedbank populations without increasing HR selection pressure. A largely unexplored aspect of the socioeconomics of HR management is whether the initial ‘reactive’ adoption of IWM has a ‘preventative’ effect on the further spread or proliferation of resistance development to other herbicides and in other weed species. To gain a deeper understanding of the farm-scale factors determining the presence of HR, the adoption of IWM practices, and the relationship between IWM use and HR status, we focus on whether initial reactive adoption of IWM has a preventative effect on further resistance development. Data from 602 broadacre grain growers across 13 agroecological zones of Australia were analyzed using Ordinary Least Squares and Logistic regressions to assess the extent of IWM practice adoption, while Ordered Logit models were used to capture the diversity of IWM practice use, including HWSC strategies. Notably, we found that adoption of IWM (including HWSC) was higher in the Western and Southern regions, most likely where HWSC practices can offer greater control on the most common resistant weed types, such as annual ryegrass. There was no evidence to support the argument that IWM use on farms is associated with reduced herbicide use, but there was considerable evidence to show that early forms of HR development on farms have led to the use of IWM practices, such as HWSC, that have the potential to reduce the seed set of weeds with potential new forms of resistance. The extent of HR was the only consistently highly significant variable to explain IWM use. These results have the capacity to pre-emptively identify sustainable weed management scenarios in various regional farming systems and future contexts, including where earlier ‘reactive’ adoption of resistance management practices may now be offering ‘proactive’ benefits. • Examine farm-scale factors influencing IWM and HWSC in Australian cropping systems. • First nationwide farm-level study links HR presence to proactive adoption of IWM practices. • Analysis of 602 growers' data across 13 AEZ using OLS, logistic, and ordered logit models. • HR presence, farm size, no-till years, and advisory support strongly drive IWM uptake, especially in WA & SA. • Evidence-based guidance for regional policymakers to target HR management and support IWM adoption.