Federally established watersheds, rangelands and forests have produced long-term records of biotic and abiotic measurements that span decades and centuries across the U.S. The goal of this review is to express the value of such long-term data for understanding and predicting ecosystem dynamics and the importance of continued long-term data collection. The basic conclusion is that the current understanding of hydrologic, ecologic, and climatic processes would simply not be possible without these multi-decadal datasets. Similarly, the development of many prediction models has been based largely on parameterization and validation with these data. As a result, long-term data have influenced management strategies for the most important activities and events affecting our natural resources, including livestock grazing, erosion control, logging, urbanization, disease, flood, drought, fire, desertification, and non-native plant invasion. Long-term data collection is just as important now as it was when the federal experimental watersheds, rangelands and forests were established nearly a century ago. There is a movement in the U.S. to develop a network of networks to study spatial patterns as well as temporal trends. Toward this integrated scientific infrastructure, new long-term data collection efforts have been designed to coordinate with existing networks, and existing long-term data collection networks have adapted their measurements to address new science issues. This flexibility and foresight has made, and continues to make, long-term data collection sustainable, relevant and inherently valuable.