부산대학교 도서관

Shortwave net radiation (SWNR) plays an important role in the surface radiation balance and serves as the primary driving force for the exchange of surface and atmospheric materials. Although numerous algorithms exist for estimating SWNR, most of them tend to ignore the influence of aerosols and digital elevation model (DEM) on SWNR. Specifically, the impact of different aerosol types on SWNR can vary significantly, and the SWNR also exhibits considerable variations at different altitudes. It is true that many algorithms demonstrate higher accuracy in low-altitude regions with less polluted rural aerosol (nonabsorbent aerosol) areas. However, their accuracy tends to decrease when applied to high-altitude areas and heavily polluted urban aerosol (absorbent aerosol) regions. In this study, an improved all-sky parameterized algorithm is proposed to estimate SWNR by fully considering solar zenith angle (SZA), DEM, and different aerosol types, and rural and urban aerosol types are distinguished by a random forest (RF) method. The new algorithm is verified versus surface radiation budget network (SURFRAD) and baseline surface radiation network (BSRN) observations and compared with the traditional algorithms (Tang-2006 and Li-1993) and Clouds and the Earth’s Radiant Energy System (CERES) products. The results reveal that the new algorithm exhibits excellent accuracy at both instantaneous and hourly scales. For rural and urban aerosol types under all-sky conditions, the bias and root mean square error (RMSE) of the new algorithm are both less than 3.5 and 106.5 W/m on the instantaneous scale and less than 12 and 77 W/ $\text{m}^{2}$ on hourly scale, respectively. However, the existing algorithms show a significant overestimation (bias > 50 W/ $\text{m}^{2}$ ) for the urban aerosol type under various atmospheres conditions. For the CERES single scanner footprint (SSF) (instantaneous) and CERES SYNergy (CERES SYN, 1-hourly) products, the overestimation phenomenon is also detected under urban aerosol type, with bias greater than 40 and 15 W/ $\text{m}^{2}$ , respectively. Compared with the existing algorithms, the new algorithm demonstrates superior applicability under larger SZA conditions. When the SZA exceeds 70°, the rate of estimated effective value can be increased by up to 14%. In addition, the new algorithm can effectively solve the problem of underestimation in high-altitude areas, which frequently occurs in most existing algorithms (bias $< -16$ W/ $\text{m}^{2}$ ). The improved accuracy and applicability of the new algorithm, along with its strategy of distinguishing aerosol types, can provide valuable insights for the SWNR estimation from space.