Accurate and continuous measurement of the subsoil CO2 is critical to better understand the terrestrial and atmosphere gas transfer process. This work aims to develop and field test a specific flow system to continuously measure the soil gas concentration (χc) and understand its main physical drivers. Hourly data measured in situ were collected through two dedicated wells at 1 m and 6 m depth coupled with micrometeorological measurement. Our study shows that χc at -1 m was at the lowest in winter and highest in summer. Meanwhile, the seasonal variation of χc at -6m is somewhat unclear. While it is inevitable that temperature plays a significant role, this factor related to biological activity cannot fully explain the variation. The decrease in χc at both depths in summer coincides with an increase of friction velocity, especially during dry periods with R2 of 0.68, which shows strong empirical evidence that wind turbulence plays a significant role in driving the deep soil CO2. A monitoring strategy for gas measurement combining borehole and micrometeorological measurement offers excellent long-term monitoring possibilities to derive the vertical distribution of CO2 and better understand the main physical drivers of gas exchange.
21 June 2018 in ILMU KELAUTAN: Indonesian Journal of Marine Sciences
A precise quantification of the sea surface partial pressure of carbon dioxide (pCO2(water)) at the water surface is required in order to define the role of the sea in air-sea exchange of CO2. Even though the pCO2(water) can be measured directly, the semi-empirical model has seen numerous application in determining the pCO2 (water) due to a time-and cost-efficient. This study aims to compare the pCO2 and FCO2 (Flux of CO2) calculated using Zhai and Zhu algorithm with the underway datasets of pCO2 obtained during the scientific cruise of CISKA-SPICE III in April 2013. The partial pressure of CO2 (pCO2) was measured using a high-accuracy electrochemical instrumentation underway HydroC/CO2 FT (flow through) with an error ±1 μ atm. Furthermore, in order to calculate the pCO2 and the FCO2 employing widely used algorithms, some data were needed including wind speed, sea surface temperature and chlorophyll-a extracted from MODIS (Moderate Resolution Imaging Spectroradiometer). According to the results obtained, the difference between the pCO2 and FCO2 derived from those two algorithms are significant. The underway datasets of pCO2 are ranging from 409.52-544.01 µatm. Meanwhile, the pCO2 derived using the Zhai algorithm and Zhu algorithm are between 405.003–422.79 µatm and 398.94-752.06 µatm respectively. The FCO2 are varied between 0.02–0.06 molC.m-2.day-1 (Zhai algorithm), 0.02-0.57 molC.m-2.day-1 (Zhu algorithm) dan 0.04-0.23 molC.m-2.day-1 (the underway datasets). A comparison of the two results reveals that pCO2 derived using Zhai algorithm is closer with the underway datasets compared with the result of pCO2 calculated using Zhu algorithm with the MRE (Mean Relative Estimation Error) as large as 19.4% and 39% respectively. Taken together, these results suggest that the Zhai algorithm is more appropriate to determine algorithms for estimating the air-sea exchange of carbon dioxide in the Malacca Straits.Keywords: carbon dioxide, Malacca Straits, pCO2, FCO2, Zhai and Zhu algorithm
In this study, we utilize a self-supervised ResNet50 model for classify land-cover imagery combining learning from both Sentinel-1 and Sentinel-2 images using the SEN12MS dataset. The model is further fine-tuned with the DFC2020 dataset, enriched by our addition of 336 new rubber data patches covering $75 \times 90$ km 2 area. The model achieved 98.1% accuracy in classifying rubber using only a 25% training data split, and 74.9% accuracy in classifying forest using only 5% training data split. This study demonstrates that accurate classification outcomes are achievable with a small number of labeled datasets by fine-tuning self-pretrained models and incorporating it with new labels. This approach not only showcases the model’s ability to efficiently learn from limited data, but also highlights the potential of adapting pre-existing models for land-cover classification tasks.