Analysis of soil properties by laboratory methods is expensive and time consuming, spectroscopy has been considered as an alternative method to reduce the need for laboratory analysis.  Spectral studies using the mid infrared (MIR) region of the spectrum have often resulted in models of high predictive capabilities for multiple soil properties over a broad range of soils, however few studies have investigated its potential for determining soil carbon fractions.  In an effort to understand climate change processes, there has been a recent focus on the assessment of soil carbon stocks and sequestration rates.  Soil carbon stocks are estimated to contain up to 80% of the earth’s terrestrial carbon.  Due to climate change and human activities soil carbon shows spatial and temporal variations. To capture soil carbon evolution through space and time spectral based methods are ideally suited. Our objective was to analyze soil carbon pools (organic, recalcitrant, and hydrolysable) using chemometric analysis of MIR spectra from 1014 soil samples collected across Florida.  Florida (~150,000 km²) represents a landscape which covers carbon-rich and carbon-poor soils which occur along diverse trajectories of hydrologic, vegetation, land use, and climatic gradients. We used various chemometric methods to assess the prediction performance of MIR spectra and identify the best performing models. Overall, MIR prediction models performed as well as combustion methods to infer on soil organic carbon, and have much promise to predict accurately carbon fractions.  Thus in addition to considering multiple carbon pools, this study provides an opportunity to expand the calibration range of total carbon models that have previously been studied and thereby determine if models are stable across large carbon ranges.