An innovative trend analysis (ITA) with significant test was proposed for detecting the annual and seasonal variation trends of solar radiation at 48 stations in five different climatic zones across China during 1962-2015. The solar radiation generally showed a significant decreasing trend (p<.05 or p<.01) at most stations, however, some stations exhibited significant increasing trends (p<.05 or p<.01) in eastern part of temperate monsoon climatic zone, western part of subtropical monsoon climatic zone and tropical monsoon climatic zone using ITA. The ITA method was compared with two traditional trend analysis methods, i.e., linear regression analysis (LRA) and Mann-Kendall (M-K) test. The results indicated almost all significant trends (P<.05 or P<.01) that can be detected by LRA or M-K test (117 time series) can be effectively identified using ITA (116 time series). Meanwhile, many significant trends (93 time series) that cannot be effectively detected by LRA or M-K test can be identified using ITA. So ITA could detect hidden-trends that cannot be observed using traditional LRA and M-K test. The possible causes for decreasing trends at most stations in China were investigated by discussing the annual and seasonal variations of anthropogenic aerosol loadings and sunshine duration. Moreover, the differences, similarities and advantages of ITA, LRA and M-K test were compared and evaluated. (C) 2017 Elsevier Ltd. All rights reserved.