Economic development and rapid urbanization have caused dramatic changes in regional land use and land cover (LULC), which can directly affect ecosystem function and ecosystem service values. Guangdong Province is located in southern China, with tropical and subtropical monsoon climates. The LULC in this region features a high degree of fragmentation and a rich diversity of land types. Remote sensing has proven to be an effective tool to characterize and quantify LULC information, but the cloud-prone and rainy weather in this region makes it difficult to obtain valid optical remote sensing images. In addition, the complexity of the spectral features of some of the land types also has a negative impact on the accuracy of LULC classification using only optical remote sensing imagery. Synthetic aperture radar (SAR) can transmit energy at microwave frequencies that are unaffected by weather conditions. This advantage gives SAR all-day and all-weather imaging capability. Furthermore, previous research has shown that SAR measurements are sensitive to the biophysical and geophysical characteristics of land targets. In this paper, we propose a preliminary algorithm to improve LULC classification accuracy by combined use of optical remote sensing data from TM and HJ, and microwave remote sensing data from TerraSAR-X collected over Leizhou Peninsula in Guangdong Province. Multitemporal spectral and backscattering features of major land types in the study area are first analyzed using the TM, HJ, and TerraSAR-X images. The analysis shows that it is difficult to discriminate among land types such as banana trees, sugarcane, and forests using TM and HJ data because of the similarity in temporal variation of spectral characteristics of these vegetation types. However, these vegetation types show different backscattering features in the TerraSAR-X images, because of their different structures, sizes, distributions and dielectric properties. Based on this analysis, decision tree rules are then established to detect land types by integrating the reflectance features of land types with their backscattering properties. An object-oriented classifier is then employed to classify the test site using these rules. The classification has been validated using field surveys. The results show that the proposed method can provide higher accuracy of land cover classification compared to using only TM or TerraSAR-X data. The results also indicate that geophysical and biophysical features of crops affect the backscattering characteristics of the crops in X-band SAR data. The scattering mechanisms of different land types should be further explored in future research to better understand the scattering processes of land targets. This knowledge would help in selecting the optimal backscattering features in SAR data to enhance the separability of land types. The imaging date of optical and SAR remote sensing data is also an important factor for LULC classification. Particularly for forests and crops, the selection of optimal acquisition dates of remote sensing data can maximize the differences in reflectance and backscattering features of various land types to improve the efficiency of the decision rules in the classifier. These results further demonstrate that synergetic use of optical and microwave remote sensing has great potential for the application of remote sensing in monitoring changes in land cover/use in southern China.