Urban street-side greenery, as an indispensable element of urban green spaces, is beneficial to residents’ physical and mental health. As readily available internet data, street view images have been widely used in urban green spaces research. While the relevant research using multiple images from different directions at a sampling point, researchers need to calculate the index of visible vegetation cover for many times. However, one Baidu panoramic street view image can cover the 360° view similar to that of a pedestrian. In this study, we selected 9644 points at 50-m intervals along the street lines in the central district of Sanya city, China, and acquired panoramic images via the Baidu application programming interface (API). The sky pixels were detected within the Baidu panoramic street view images using a proposed reflectance indicator. The green vegetation was extracted according to the Back Propagation (BP) neural-network method. Our proposed method was validated by comparing the results of the manual recognition and PSPNet method, and the accuracy met the requirements of the study. The Panoramic Green View Index (PGVI) was proposed to quantitatively evaluate greenery around streets. The authors found that the highest frequency value in the distribution was 0.075, which accounted for 32% of the total sample points, and the average PGVI value in this study area was low; the PGVI values between different roads varied greatly, and primary roads tended to have higher PGVI values than other roads. This case study proved that the PGVI is well suited for evaluating greenery around streets. We suggest that the PGVI derived from Baidu panoramic street view images may be a useful tool for city managers to support urban green spaces planning and management.

Soil dielectric model is an essential part of the microwave soil moisture retrieval process. This study compared the performance of four widely used dielectric models, the Wang-Schmugge model (WS), Dobson model, generalized refractive mixing dielectric model (GRMDM), and multi-relaxation generalized refractive mixing dielectric model (MRGRMDM), and investigated the effects of the uncertainties of each model on soil moisture retrievals. Furthermore, the simulated soil dielectric constants were evaluated by measured dielectric data at the P/L/C/X bands. The results showed that the uncertainties induced in soil moisture retrievals by an alternative dielectric model exceeded 0.09 m³ m⁻³ in the worst case. The Dobson model is sensitive to the sand content. WS, GRMDM, and MRGRMDM model are sensitive to the clay content. The measured dielectric data further verified that the applicability of each dielectric model depends on the soil texture type and soil moisture condition. Compared with Dobson model, WS showed better performance at dry soil. GRMDM and MRGRMDM provided better results under lower clay content soil. Especially, MRGRMDM has better simulation accuracy than GRMDM in the low-frequency range (< 1 GHz).