<p>Food production is a primary process through which climate change affects human societies, yet the response of agricultural systems to past climate variability and change is poorly understood, largely because detailed data on yields is lacking before the 20^th century. To address this gap, we used crop simulation models adapted for relatively simple, low-yield farming to simulate wheat (<i>Triticum aestivum, Triticum durum)</i> and grape (<i>Vitis vinifera ‘</i>Syrah’) yields under realistic environmental conditions for Italy during Roman times. To capture regional variability, simulations were conducted for one grape-growing and five wheat-growing locations. We used existing paleoclimate proxy records to test the potential sensitivity of agriculture to variations in temperature and precipitation across the Roman Climate Optimum (RCO) (ca. 200 BCE-100 CE) and Late Antique Little Ice age (LALIA) (ca. 530–680 CE). Our simulations indicate that yields for both wheat and grapes declined substantially under the colder conditions associated with the Late Antique Little Ice Age compared with the Roman Climate Optimum. Grapes appear notably sensitive to temperature variation from year to year, while wheat responds most directly to changes in water availability. Because both climatic inputs and crop parameters are reconstructed from proxy and historical data, the precise magnitude of simulated yields depends on modeling assumptions. However, the consistent pattern of substantially declining productivity under sustained cooling is robust across locations and sensitivity tests. Our analysis highlights both the uncertainties and the explanatory potential of model-based reconstructions for understanding the impacts of climate change on ancient agricultural systems.</p>