The threat imposed by increased sediment loading rates ranks among the most important stressors affecting coral reef ecosystems worldwide. This study represents an effort to quantify the effects of unpaved roads on erosion rates in a dry sub-tropical area of Puerto Rico and is intended to aid in developing scientifically-based erosion mitigation strategies. Hence, the specific objectives of this study were to: (1) measure sediment production rates from unpaved roads; (2) evaluate the effect of precipitation, rainfall erosivity, slope, plot length, and vegetation cover on sediment production rates; and (3) compare measured sediment production rates to published surface erosion data from roaded and natural sites in the Eastern Caribbean. Sediment production from nine abandoned road segments with varying slopes and plot lengths were measured with sediment traps in southwestern Puerto Rico from August 2003 to September 2005. The overall average sediment production rate was 0.84 Mg ha−1 yr−1, and the range of observed values was 15–50 times higher than locally-measured natural erosion rates. Only four of the nine study sites had a statistically significant correlation between sediment production and total rainfall and this is attributed to progressive changes in some of the conditions controlling erosion rates. Sediment production rates were dependent on slope raised to the 1.6th power, as well as to the product of plot length times slope1.6. Average erosion rates were inversely but poorly related to vegetation cover. An observed decline in sediment production rates was observed for all nine study segments, and this amounted to a statistically significant difference between observations made during the early stages of monitoring (Period 1: August 2003–April 2004) relative to those during the latter parts of the study (Period 2: May 2004– September 2005). Annual erosion rates during Period 1 amounted to 0.18 to 4.0 Mg ha−1 yr−1 for road segments with 1% and 22% slopes, respectively; rates during Period 2 were between 0.024 and 0.52Mg ha−1 yr−1, or only 13% of those during Period 1. Differences in sediment production rates between the two periods are attributed to more intense rainfall during Period 1 and to a notably higher vegetation cover during Period 2. Rainfall appears to play a paradoxical role in controlling surface erosion rates on abandoned road surfaces in a sub-tropical dry region.While ample rainfall is needed to generate erosion by rainsplash and overland flow, once rainfall satisfies soil moisture requirements for sustaining vegetation colonization it may also contribute to declining sediment production rates. Therefore, any model that attempts to properly address the temporal variation in erosion rates occurring on abandoned roads in a climatic setting where moisture availability is a limiting factor must not only follow the more traditional surface armoring-based approach but must also integrate the effects of re-vegetation. Such types of modelswill eventually become useful tools to properly assess the effects of past, current, and future land use practices on erosion rates, and to improve mitigation and land development strategies to lessen the impact on vital marine habitats