Cellulose delta18O and deltaD can provide insights on climates and hydrological cycling in the distant past and how these factors differ spatially. However, most studies of plant cellulose have used only one isotope, most commonly delta18O, resulting in difficulties partitioning variation in delta18O of precipitation vs. evaporative conditions that affect leaf water isotopic enrichment. Moreover, observations of pronounced diurnal differences from conventional steady-state model predictions of leaf water isotopic fractionation have cast some doubt on single isotope modeling approaches for separating precipitation and evaporation drivers of cellulose delta18O or deltaD. We explore a dual isotope approach akin to the concept of deuterium-excess (d), to establish deuterium deviations from the global meteoric water line in leaf water (deltad(l)) as driven by relative humidity (RH). To demonstrate this concept, we survey studies of leaf water delta18O and deltaD in hardwood vs. conifer trees. We then apply the concept to cellulose delta18O and deltaD using a mechanistic model of cellulose delta18O and deltaD to reconstruct deuterium deviations from the global meteoric water line (deltad(c)) in Quercus macrocarpa, Q. robur, and Pseudotsuga menziesii. For each species, deltad(c) showed strong correlations with RH across sites. deltad(c) agreed well with steady-state predictions for Q. macrocarpa, while for Q. robur, the relationship with RH was steeper than expected. The slope of deltad(c) vs. RH of P. menziesii was also close to steady-state predictions, but deltad(c) were more enriched than predicted. This is in agreement with our leaf water survey showing conifer deltad(l) was more enriched than predicted. Our data reveal that applications of this method should be appropriate for reconstructing RH from cellulose delta18O and deltaD after accounting for differences between hardwoods and conifers. Hence, deltad(c) should be useful for understanding variability in RH associated with past climatic cycles, across regional climates, or across complex terrain where climate modeling is challenging. Furthermore, deltad(c) and inferred RH values should help in constraining variation in source water delta18O.