A handheld-based agricultural decision support system (HADSS) for advising the efficient and economical management of N fertilizer application for wheat cropping in Australia was developed. With this HADSS, farmers are capable to electronically record soil and crop data, to retrieve in-situ meteorological data through wireless internet connection, to run and calibrate a series of widely-recognized regression-derived empirical and process-oriented biophysical models for agroecosystems, and to make practice decisions in the field, in pursuing the site-specific best management practices. It largely extends farmers' capability to adopt or to try the latest and innovative agricultural research outcomes to help improve cropping management and to economically and environmentally benefit from doing so. This HADSS includes a handheld computer with wireless internet and a number of handheld-synchronized software. Software are a HADSS manager, which manages soil and crop database, site meteorological dataset through wireless networking, model simulation, post-processing of simulation (graphing and reporting), and model upgrading through wireless networking, and a series of empirical and biophysical models for advising agricultural practices of irrigation and N fertilizer application. At the first stage, an empirical model using wheat shoot density to estimate soil N availability was incorporated into this HADSS to advise farmers' decision on N fertilization. In this model, the field crop images are closely monitored at different wheat early growing stages by the built-in digital camera of the handheld computer to analyze the ground vegetation fraction (VF) by the handheld-installed software. With this VF, crop leaf area index and crop shoot density are estimated, then crop shoot density is related to soil N availability represented by crop N uptake. Given the optimal crop N uptake at different growing stages, the decision on necessary N fertilization is made in-situ in the cropping field, and the N fertilizer use efficiency is largely improved.