The temporal-spatial interaction of land cover and non-point source (NPS) nutrient pollution were analyzed with the Soil and Water Assessment Tool (SWAT) to simulate the temporal-spatial features of NPS nutrient loading in the upper stream of the Yellow River catchment. The corresponding land cover data variance was expressed by the normalized difference vegetation index (NDVI) that was calculated from MODIS images. It was noted that the temporal variation of land cover NDVI was significantly correlated with NPS nutrient loading. The regression analysis indicated that vegetation not only detained NPS nutrient pollution transportation, but also contributed to sustainable loading. The temporal analysis also confirmed that regional NDVI was an effective index for monthly assessment of NPS nitrogen and phosphorus loading. The spatial variations of NPS nutrient loading can be classified with land cover status. The high loadings of NPS nitrogen in high NDVI subbasins indicated that forestry and farmland are the main critical loss areas. Farmland contributed sustainable soluble N, but the loading of soluble and organic N from grassland subbasins was much lower. Most P loading came from the areas covered with dense grassland and forestry, which cannot directly discharge to local water bodies. However, some NPS phosphorus from suburban farmland can directly discharge into adjacent water bodies. The interactions among nutrient loading, NDVI, and slope were also analyzed. This study confirmed that the integration of NPS modeling, geographic information systems, and remote sensing is needed to understand the interactive dynamics of NPS nutrient loading. Understanding the temporal-spatial variation of NPS nutrients and their correlations with land cover will help NPS pollution prevention and water quality management efforts. Therefore, the proposed method for evaluating NPS nutrient loading by land cover NDVI can be an effective tool for pollution evaluation and watersheds planning.