The aim of this paper is to assess the latest version of the MODIS LAI/FPAR product (MOD15A2H), namely Collection 6 (C6). We comprehensively evaluate this product through three approaches: validation with field measurements, intercomparison with other LAI/FPAR products and comparison with climate variables. Comparisons between ground measurements and C6, as well as C5 LAI/FPAR indicate: (1) MODIS LAI is closer to true LAI than effective LAI; (2) the C6 product is considerably better than C5 with RMSE decreasing from 0.80 down to 0.66; (3) both C5 and C6 products overestimate FPAR over sparsely-vegetated areas. Intercomparisons with three existing global LAI/FPAR products (GLASS, CYCLOPES and GEOV1) are carried out at site, continental and global scales. MODIS and GLASS (CYCLOPES and GEOV1) agree better with each other. This is expected because the surface reflectances, from which these products were derived, were obtained from the same instrument. Considering all biome types, the RMSE of LAI (FPAR) derived from any two products ranges between 0.36 (0.05) and 0.56 (0.09). Temporal comparisons over seven sites for the 2001–2004 period indicate that all products properly capture the seasonality in different biomes, except evergreen broadleaf forests, where infrequent observations due to cloud contamination induce unrealistic variations. Thirteen years of C6 LAI, temperature and precipitation time series data are used to assess the degree of correspondence between their variations. The statistically-significant associations between C6 LAI and climate variables indicate that C6 LAI has the potential to provide reliable biophysical information about the land surface when diagnosing climate-driven vegetation responses.

As the latest version of Moderate Resolution Imaging Spectroradiometer (MODIS) Leaf Area Index (LAI) and Fraction of Photosynthetically Active Radiation (FPAR) products, Collection 6 (C6) has been distributed since August 2015. This collection is evaluated in this two-part series with the goal of assessing product accuracy, uncertainty and consistency with the previous version. In this first paper, we compare C6 (MOD15A2H) with Collection 5 (C5) to check for consistency and discuss the scale effects associated with changing spatial resolution between the two collections and benefits from improvements to algorithm inputs. Compared with C5, C6 benefits from two improved inputs: (1) L2G–lite surface reflectance at 500 m resolution in place of reflectance at 1 km resolution; and (2) new multi-year land-cover product at 500 m resolution in place of the 1 km static land-cover product. Global and seasonal comparison between C5 and C6 indicates good continuity and consistency for all biome types. Moreover, inter-annual LAI anomalies at the regional scale from C5 and C6 agree well. The proportion of main radiative transfer algorithm retrievals in C6 increased slightly in most biome types, notably including a 17% improvement in evergreen broadleaf forests. With same biome input, the mean RMSE of LAI and FPAR between C5 and C6 at global scale are 0.29 and 0.091, respectively, but biome type disagreement worsens the consistency (LAI: 0.39, FPAR: 0.102). By quantifying the impact of input changes, we find that the improvements of both land-cover and reflectance products improve LAI/FPAR products. Moreover, we find that spatial scale effects due to a resolution change from 1 km to 500 m do not cause any significant differences.

Deep learning has been successfully applied to infrared and visible image fusion due to its powerful ability of feature representation. Existing most deep learning based infrared and visible image fusion methods mainly utilize pure convolution model or pure transformer model, which leads to that the fused image cannot preserve long-range dependencies (global context) and local features simultaneously. To this end, we propose a convolution-guided transformer framework for infrared and visible image fusion (CGTF), which aims to combine the local features of convolutional network and the long-range dependency features of transformer to produce satisfactory fused image. In CGTF, the local features are calculated by convolution feature extraction module, and then the local features are used to guide the transformer feature extraction module to capture the long-range dependencies of the image, which can overcome not only the lack of long-range dependencies that exists in convolutional fusion methods, but also the deficiency of local feature that exists in transformer models. Moreover, the convolution-guided transformer fusion framework can consider the inherent relationship of local feature and long-range dependencies due to the alternate use of convolution feature extraction module and transformer module. In addition, to strengthen local feature propagation, we employ dense connections among convolution feature extraction modules. Ablation experiments demonstrate the effectiveness of convolution-guided transformer fusion framework and loss function. We employ two datasets to compare our method with other nine methods, which includes three traditional methods, five deep learning based methods and one transformer based method. Qualitative and quantitative experiments demonstrate the advantages of our method.

Change detection in time series is among the most critical problems in Earth monitoring and attracts extensive attention in the remote sensing community. The task is, however, nontrivial because available images are irregularly collected due to interference by clouds and shadows. Traditional recurrent neural networks neglect such information and, thus, degrade the possibility of distinguishing pseudochanges (caused by intra-annual and interannual dynamics) and real changes. To this end, we proposed an unsupervised time-distance-guided convolutional recurrent network (UTRNet) for change detection in irregularly collected images. UTRNet is distinctive because the influence of pseudochanges can be suppressed by adopting a novel time-distance-guided long short-term memory (TLSTM) unit, in which input and forget gates are modified to adapt to irregular time distances. To the best of our knowledge, this is the first time that the influence of pseudochanges can be suppressed using irregular time distances. Moreover, to make UTRNet more applicable, a weighted prechange detection model is proposed to extract the most reliable training samples automatically. In the training process, unlike existing approaches that only care about changed and unchanged sample imbalance, our UTRNet also pays attention to imbalance between hard and easy samples, and proposes a new focal weighted cross-entropy loss, which helps to make the training process focus on hard changed samples. The proposed UTRNet is validated on Landsat 8 time-series data collected over nine typical scenes in the 2013–2021 period. Qualitative and quantitative comparisons with several state-of-the-art methods suggest the superior performance of UTRNet. Our dataset and codes are available at https://github.com/thebinyang/UTRNet .