A 30 m terrace mapping in China using Landsat 8 imagery and digital elevation model based on the Google Earth Engine

Abstract. The construction of terraces is a key soil conservation practice on agricultural land in China providing multiple valuable ecosystem services. Accurate spatial information needed for both management and research. In this study, the first 30 m resolution terracing map entire territory produced by supervised pixel-based classification using multisource multi-temporal data based Google Earth Engine (GEE) platform. We extracted time-series spectral features topographic from Landsat 8 images Shuttle Radar Topography Mission digital elevation model (SRTM DEM) data, classifying cropland area (cultivated Globeland30) into terraced non-terraced types through random forest classifier. overall accuracy kappa coefficient were evaluated 10 875 test samples achieved values 94 % 0.72, respectively. For terrace class, producer's (PA) was 79.945 %, user's (UA) 71.149 %. performed best Loess Plateau southwestern China, where are most numerous. Some northeastern, eastern-central, southern areas had relatively high uncertainty. Typical errors mapping results sloping (non-terrace with slope ? 5?), low-slope terraces, non-crop vegetation. Terraces widely distributed total estimated to be 53.55 Mha (i.e., 26.43 China's area) pixel counting (PC) method 58.46 ± 2.99 28.85 1.48 error-matrix-based model-assisted estimation (EM) method. Elevation identified as main terrace/non-terrace classification, (such percentiles NDVI, TIRS2, BSI) also essential. more challenging identify than other use because intra-class feature heterogeneity, interclass similarity, fragmented patches, which should focus future Our algorithm can used large-scale regions globally, our will serve landmark studies service assessments including erosion control, carbon sequestration, biodiversity conservation. available public at https://doi.org/10.5281/zenodo.3895585 (Cao et al., 2020).