Comparative analysis of MODIS-based daily cloud free snow cover extent products over the Qinghai-Tibet Plateau

Rapid snow dynamics and persistent cloud cover over the Qinghai–Tibet Plateau (QTP) hinder the acquisition of spatially and temporally continuous daily snow cover observations. This study evaluates five publicly available MODIS-based daily cloud-free snow cover extent (SCE) products: Dysno_Cloudfree, HMRF-MODIS, I-MODIS, F-MODIS, and NIEER-MODIS.These products employ different reconstruction approaches, including multistep filtering, Hidden Markov Random Field modeling, MODIS–IMS fusion, temporal interpolation, and land-cover-adaptive snow detection combined with multisource gap filling, respectively. Using high-resolution snow-cover maps as reference snow maps, we assess snow detection under cloudy conditions with confusion matrix-based metrics (OA, precision, recall, F1 score, and bias) and evaluate snow-edge representation in rugged terrain using fractal dimension (FD) RMSE and mean structural similarity index (SSIM). NIEER-MODIS achieves the highest recall (0.879) and F1 score (0.741), with an OA of 0.792, but substantially overestimates snow extent (bias = 1.373). Dysno_Cloudfree achieves a comparable OA (0.791) and the best FD RMSE (0.131), while maintaining a high SSIM (0.973). I-MODIS records the highest SSIM (0.974). These findings highlight the strengths and limitations of existing MODIS cloud-free SCE products over the QTP and provide guidance for improving snow mapping under persistent cloud cover in complex terrain.