BACKGROUND Enzymatic hydrolysis is a crucial step of biomass conversion into biofuels and different pretreatments have been proposed to improve the process efficiency. Amongst the various factors affecting hydrolysis yields of biomass samples, porosity and water accessibility stand out due to their intimate relation with enzymes accessibility to the cellulose and hemicellulose fractions of the biomass. In this work, sugarcane bagasse was subjected to acid and alkali pretreatments. The changes in the total surface area, hydrophilicity, porosity and water accessibility of cellulose were investigated by scanning electron microscopy (SEM) and nuclear magnetic resonance (NMR). RESULTS Changes in chemical and physical properties of the samples, caused by the partial removal of hemicellulose and lignin, led to the increase in porosity of the cell walls and unwinding of the cellulose bundles, as observed by SEM. (1)H NMR relaxation data revealed the existence of water molecules occupying the cores of wide and narrow vessels as well as the cell wall internal structure. Upon drying, the water molecules associated with the structure of the cell wall did not undergo significant dynamical and partial moisture changes, while those located in the cores of wide and narrow vessels kept continuously evaporating until reaching approximately 20% of relative humidity. This indicates that water is first removed from the cores of lumens and, in the dry sample, the only remaining water molecules are those bound to the cell walls. The stronger interaction of water with pretreated bagasse is consistent with better enzymes accessibility to cellulose and higher efficiency of the enzymatic hydrolysis. CONCLUSIONS We were able to identify that sugarcane bagasse modification under acid and basic pretreatments change the water accessibility to different sites of the sample, associated with both bagasse structure (lumens and cell walls) and hydrophilicity (lignin removal). Furthermore, we show that the substrates with increased water accessibility correspond to those with higher hydrolysis yields and that there is a correlation between experimentally NMR-measured transverse relaxation times and the efficiency of enzymatic hydrolysis. This might allow for semiquantitative estimates of the enzymatic hydrolysis efficiency of biomass samples using inexpensive and non-destructive low-field (1)H NMR relaxometry methods.