Search

Robust estimates of wetland soil organic carbon (SOC) pools are critical to understanding wetland carbon dynamics in the global carbon cycle. However, previous estimates were highly variable and uncertain, due likely to the data sources and method used. Here we used machine learning method to estimate SOC storage and their changes over time in China's wetlands based on wetland SOC density database, associated geospatial environmental data, and recently published wetland maps. We built a database of wetland SOC density in China that contains 809 samples from 181 published studies collected over the last 20 years as presented in the published literature. All samples were extended and standardized to a 1‐m depth, on the basis of the relationship between SOC density data from soil profiles of different depths. We used three different machine learning methods to evaluate their robustness in estimating wetland SOC storage and changes in China. The results indicated that random forest model achieved accurate wetland SOC estimation with R2 being.65. The results showed that average SOC density of top 1 m in China's wetlands was 25.03 ± 3.11 kg C m−2 in 2000 and 26.57 ± 3.73 kg C m−2 in 2020, an increase of 6.15%. SOC storage change from 4.73 ± 0.58 Pg in 2000 to 4.35 ± 0.61 Pg in 2020, a decrease of 8.03%, due to 13.6% decreased in wetland area from 189.12 × 103 to 162.8 × 103 km2 in 2020, despite the increase in SOC density during the same time period. The carbon accumulation rate was 107.5 ± 12.4 g C m−2 year−1 since 2000 in wetlands with no area changes. Climate change caused variations in wetland SOC density, and a future warming and drying climate would lead to decreases in wetland SOC storage. Estimates under Shared Socioeconomic Pathway 1‐2.6 (low‐carbon emissions) suggested that wetland SOC storage in China would not change significantly by 2100, but under Shared Socioeconomic Pathway 5‐8.5 (high‐carbon emissions), it would decrease significantly by approximately 5.77%. In this study, estimates of wetland SOC storage were optimized from three aspects, including sample database, wetland extent, and estimation method. Our study indicates the importance of using consistent SOC density and extent data in estimating and projecting wetland SOC storage. [ABSTRACT FROM AUTHOR]

The accurate characterization of tree species distribution in forest areas can help significantly reduce uncertainties in the estimation of ecosystem parameters and forest resources. Deep learning algorithms have become a hot topic in recent years, but they have so far not been applied to tree species classification. In this study, one-dimensional convolutional neural network (Conv1D), a popular deep learning algorithm, was proposed to automatically identify tree species using OHS-1 hyperspectral images. Additionally, the random forest (RF) classifier was applied to compare to the algorithm of deep learning. Based on our experiments, we drew three main conclusions: First, the OHS-1 hyperspectral images used in this study have high spatial resolution (10 m), which reduces the influence of mixed pixel effect and greatly improves the classification accuracy. Second, limited by the amount of sample data, Conv1D-based classifier does not need too many layers to achieve high classification accuracy. In addition, the size of the convolution kernel has a great influence on the classification accuracy. Finally, the accuracy of Conv1D (85.04%) is higher than that of RF model (80.61%). Especially for broadleaf species with similar spectral characteristics, such as Manchurian walnut and aspen, the accuracy of Conv1D-based classifier is significantly higher than RF classifier (87.15% and 71.77%, respectively). Thus, the Conv1D-based deep learning framework combined with hyperspectral imagery can efficiently improve the accuracy of tree species classification and has great application prospects in the future. [ABSTRACT FROM AUTHOR]