Here we present the development of a new adjusted dataset composed of the 22 longest and most reliable Spanish daily temperature records (maximum and minimum temperatures and derived daily mean temperature) covering the period 1850–2003. The paper describes the approaches followed for compiling, controlling the quality and homogenising these 22 daily Spanish records, leading to the creation of a dataset called 'Spanish Daily Adjusted Temperature Series' (SDATS). An assessment of the sources of data and metadata used is followed by a reliability assessment of the selected network. Data quality control (QC) procedures applied to raw daily maximum (T max) and minimum (T min) temperatures and their results are then considered. For the very first time, an empirical minimisation of the bias related to the impact of changing exposure of thermometers on the records has been undertaken. The application of the Standard Normal Homogeneity Test (SNHT) to check homogeneity of raw T max and T min data on a monthly basis is presented, together with a discussion of the causes, magnitudes and timings of the various inhomogeneities. All 22 records contained a number of inhomogeneities (2.6 on average), mainly associated with documented station relocations confirmed by the metadata available. Monthly adjustments calculated for both screen developments and from the SNHT were linearly interpolated to a daily basis following the Vincent et al. (2002) scheme. Finally, the procedures adopted for creating the regional average, the Spanish Temperature Series (STS), together with an exploratory analysis of long-term trends of each T max and T min records, are provided. The final analysis shows that over mainland Spain highly significant rates of temperature increases have occurred for T max and T min (0.12 ° C/decade and 0.10 ° C/decade, respectively) over 1850–2003.