Publications

Zhang, HL; Dong, XQ; Xi, BK; Xin, XZ; Liu, QH; He, HM; Xie, XP; Li, L; Yu, SS (2021). Retrieving high-resolution surface photosynthetically active radiation from the MODIS and GOES-16 ABI data. REMOTE SENSING OF ENVIRONMENT, 260, 112436.

Abstract
Incident photosynthetically active radiation (PAR) is a key parameter in plant physiological, biological and physical process-based terrestrial ecosystem models. Deriving highly accurate and spatially continuous PAR from remote sensing data can fill in the gaps of insufficient ground-based measurements. In this study, we provide an efficient approach for estimating PAR by employing look-up table (LUT) method using observations from both MODIS and GOES-16. The GOES-16 is the first satellite of the new-generation US Geostationary Operational Environmental Satellite R-series (GOES-R), which can capture dynamic changes in the atmosphere every 5 min. First, the diurnal variation of surface bidirectional reflectance is characterized assuming that the surface characteristics and solar zenith angles in tropical and middle latitudes at a given time of day do not change dramatically within a month. Next, the aerosol optical depth (AOD) and cloud optical depth (COD) are retrieved from the GOES-16 visible observations under both clear-sky and cloudy conditions. Finally, PAR is estimated via the retrieved AOD and COD using the LUT method and a MODIS-based surface albedo product. The estimated PAR was compared against ground-based measurements collected from seven SURFace RADiation budget network (SURFRAD) sites and 23 National Ecological Observatory Network (NEON) sites at instantaneous, hourly and daily scales. The instantaneous PAR validation results achieved R-2 values of 0.97 and 0.82, and root mean square errors (RMSEs) of 18.1 and 51.1 W m(-2), for clear and cloudy skies, respectively. The RMSEs of the hourly averaged PAR were reduced to 16.8 W m(-2) and 40.4 W m(-2), respectively, for clear-sky and cloudy conditions. The daily mean PAR evaluation for all sky conditions shows an overall mean bias error (MBE) of 0.97 W m(-2) and an RMSE of 10.5 W m(-2). Compared with other global PAR datasets, the estimated PAR values in this study have lower RMSEs and higher spatial-temporal resolution. Uncertainties in the retrieved PAR caused by topography and solar-cloud-geometry effects (SCGE) can be reduced by spatial and temporal averaging. Based on an evaluation of retrievals at different spatial-temporal scales, we recommend that a 20 x 20 km(2) domainaveraged PAR be utilized for instantaneous PAR validation, whereas hourly and daily averaged PAR values are nearly independent of spatial scale.

DOI:
10.1016/j.rse.2021.112436

ISSN:
0034-4257