Search

Spatial quantification of blue carbon ecosystem stocks is crucial for developing policies to mitigate climate change, especially in regions experiencing ongoing wetland disturbance from biological invasions. We integrated multiple machine learning models with the space-for-time substitution method to quantify the spatiotemporal impact of Spartina alterniflora invasion on tidal marsh sediment blue carbon (soil organic carbon – 'SOC') stocks at 100 cm depth in the Yangtze Estuary. Our results show that the invasive S. alterniflora contributed more than half of the total SOC stocks (2,056 ± 379 Gg C, 1 Gg = 106 kg) in the 27,600 ha tidal marshes of the Yangtze Estuary, which were estimated to be 1,107 ± 176 Gg C. S. alterniflora increased the SOC stocks in the Yangtze Estuary within the first 15 years, but this gain was not sustained in the long term, with a gradual decline (by 13.14 Mg C/ha) observed after 15 years of S. alterniflora growth. We found that sediment salinity, tidal range, and human accessibility were strong indicators for modeling and predicting SOC stocks, with Random Forest providing the best simulation of tidal marsh SOC stocks (R2 = 0.894, RMSE=7.646 Mg C/ha, and MAPE=9.469 %). Our study provides much needed information on blue carbon stocks in the Yangtze Estuary under biological invasion stress, and offers guidance for targeted S. alterniflora management actions in the future.

Abstract: Tidal salt marshes in the San Francisco Estuary region display heterogeneous vegetation patterns that influence wetland function and provide adequate habitat for native or endangered wildlife. In addition to analyzing the extent of vegetation, monitoring the dynamics of vegetation pattern within restoring wetlands can offer valuable information about the restoration process. Pattern metrics, derived from classified remotely sensed imagery, have been used to measure composition and configuration of patches and landscapes, but they can be unpredictable across scales, and inconsistent across time. We sought to identify pattern metrics that are consistent across spatial scale and time – and thus robust measures of vegetation and habitat configuration – for a restored tidal marsh in the San Francisco Bay, CA, USA. We used high-resolution (20cm) remotely sensed color infrared imagery to map vegetation pattern over 2 years, and performed a multi-scale analysis of derived vegetation pattern metrics. We looked at the influence on metrics of changes in grain size through resampling and changes in minimum mapping unit (MMU) through smoothing. We examined composition, complexity, connectivity and heterogeneity metrics, focusing on perennial pickleweed (Sarcocornia pacifica), a dominant marsh plant. At our site, pickleweed patches grew larger, more irregularly shaped, and closely spaced over time, while the overall landscape became more diverse. Of the two scale factors examined, grain size was more consistent than MMU in terms of identifying relative change in composition and configuration of wetland marsh vegetation over time. Most metrics exhibited unstable behavior with larger MMUs. With small MMUs, most metrics were consistent across grain sizes, from fine (e.g. 0.16m2) to relatively large (e.g. 16m2) pixel sizes. Scale relationships were more variable at the landcover class level than at the landscape level (across all classes). This information may be useful to applied restoration practitioners, and adds to our general understanding of vegetation change in a restoring marsh. [Copyright &y& Elsevier]