Studying carbon dynamics is important to evaluate greenhouse gas mitigation options, especially the carbon sequestration potential in agricultural soils. The carbon inputs from crop residue to soils can be estimated from crop aboveground biomass, which in turn can be estimated from crop yields. We derived relationships between remotely sensed information and aboveground biomass of crops in Iowa in order to estimate carbon inputs.. We used multi-temporal biweekly AVHRR NDVI, and county level crop yield data from NASS, for the years 1992 and 1997, to evaluate relationships between remotely sensed information and crop aboveground biomass. NDVI pixel values extracted for different biweekly periods were analyzed with aboveground dry biomass of the major crops in Iowa (i.e. corn, soy, and oats), in simple best subset multiple linear regressions and canonical correlation analyses. Both simple best subset multiple linear regression models and models incorporating canonical variables gave good relationships between NDVI pixel values and crop biomass values. But using canonical correlation yielded better and more reliable and comprehensible models, as it removed any autocorrelation among the NDVI pixel values from different biweekly periods. Since NDVI and biomass values of the major crops for Iowa were well correlated under canonical correlation analyses, agreeing well with phenological changes shown by the field crops, the methodology used was found to be a good approach in predicting biomass (and any subsequent estimation of crop residue carbon inputs) from remotely sensed data.