The objective of this study was to map agricultural intensity on the scale of France with spatial resolution adequate for policy targeting. Using the French Farm Accountancy Data Network (FADN), we computed an intensity indicator based on input costs per ha ("IC/ha"). Common variables between the FADN and four other datasets were included in a two steps multinomial regression to estimate the IC/ha value of each Small Agricultural Region ("SAR", units with homogeneous agro-ecological characteristics with mean width = 22.4 km). The local indicator of spatial association was used to reveal clusters where SARs with homogeneous intensities were aggregated. We showed that the IC/ha indicator displayed a broad intensity gradient where production types were fairly evenly distributed. Multinomial regression models provided a reliable estimate of the intensity indicator (mean cross-validation error = 23%, mean r2 = 0.7) with SAR resolution. At the scale of France and within the two intensity extremes (<300 euros /ha and >500 euros /ha), SARs were significantly aggregated in several clusters. Most low-input SARs were aggregated into a large cluster ranging across several mountainous regions. Less high-input SARs were significantly aggregated. Our results could be used for infra-regional targeting of conservation policies.