Land use (LU) and land cover (LC) are two complementary pieces of cartographic information used for urban planning and environmental monitoring. In the context of New Caledonia, a biodiversity hotspot, the availability of up-to-date LULC maps is essential to monitor the impact of extreme events such as cyclones and human activities on the environment. With the democratization of satellite data and the development of high-performance deep learning techniques, it is possible to create these data automatically. This work aims at determining the best current deep learning configuration (pixel-wise vs. semantic labelling architectures, data augmentation, image prepossessing, …), to perform LULC mapping in a complex, subtropical environment. For this purpose, a specific data set based on SPOT6 satellite data was created and made available for the scientific community as an LULC benchmark in a tropical, complex environment using five representative areas of New Caledonia labelled by a human operator: four used as training sets, and the fifth as a test set. Several architectures were trained and the resulting classification was compared with a state-of-the-art machine learning technique: XGboost. We also assessed the relevance of popular neo-channels derived from the raw observations in the context of deep learning. The deep learning approach showed comparable results to XGboost for LC detection and over-performed it on the LU detection task (61.45% vs. 51.56% of overall accuracy). Finally, adding LC classification output of the dedicated deep learning architecture to the raw channels input significantly improved the overall accuracy of the deep learning LU classification task (63.61% of overall accuracy). All the data used in this study are available on line for the remote sensing community and for assessing other LULC detection techniques.

Trichodesmium is the major nitrogen-fixing species in the western tropical South Pacific (WTSP) region, a hot spot of diazotrophy. Due to the paucity of in situ observations, remote-sensing methods for detecting Trichodesmium presence on a large scale have been investigated to assess the regional-to-global impact of this organism on primary production and carbon cycling. A number of algorithms have been developed to identify Trichodesmium surface blooms from space, but determining with confidence their accuracy has been difficult, chiefly because of the scarcity of sea-truth information at the time of satellite overpass. Here, we use a series of new cruises as well as airborne surveys over the WTSP to evaluate their ability to detect Trichodesmium surface blooms in the satellite imagery. The evaluation, performed on MODIS data at 250 m and 1 km resolution acquired over the region, shows limitations due to spatial resolution, clouds, and atmospheric correction. A new satellite-based algorithm is designed to alleviate some of these limitations, by exploiting optimally spectral features in the atmospherically corrected reflectance at 531, 645, 678, 748, and 869 nm. This algorithm outperforms former ones near clouds, limiting false positive detection and allowing regional-scale automation. Compared with observations, 80 % of the detected mats are within a 2 km range, demonstrating the good statistical skill of the new algorithm. Application to MODIS imagery acquired during the February-March 2015 OUTPACE campaign reveals the presence of surface blooms northwest and east of New Caledonia and near 20 ^∘ S–172 ^∘ W in qualitative agreement with measured nitrogen fixation rates. Improving Trichodesmium detection requires measuring ocean color at higher spectral and spatial (<250 m) resolution than MODIS, taking into account environment properties (e.g., wind, sea surface temperature), fluorescence, and spatial structure of filaments, and a better understanding of Trichodesmium dynamics, including aggregation processes to generate surface mats. Such sub-mesoscale aggregation processes for Trichodesmium are yet to be understood.

Trichodesmium is the main nitrogen-fixing species in the South Pacific region, a hotspot for diazotrophy. Due to the paucity of in situ observations, methods for detecting Trichodesmium presence on a large scale have been investigated to assess the regional-to-global impact of these species on primary production and carbon cycling. A number of satellite-derived algorithms have been developed to identify Trichodesmium surface blooms, but determining with confidence their accuracy has been difficult, chiefly because of the scarcity of sea-truth information at time of satellite overpass. Here, we use a series of new cruises as well as airborne observational surveys in the South Pacific to quantify statistically the ability of these algorithms to discern correctly Trichodesmium surface blooms in the satellite imagery. The evaluation, performed on MODIS data at 250 m and 1 km resolution acquired over the South West Pacific, shows limitations due to spatial resolution, clouds, and atmospheric correction. A new satellite-based algorithm is designed to alleviate some of these limitations, by exploiting optimally spectral features in the atmospherically corrected reflectance at 531, 645, 678, 748, and 869 nm. This algorithm outperforms former ones near clouds, limiting false positive detection, and allowing regional scale automation. Compared with observations, 80 % of the detected mats are within a 2 km range, demonstrating the good statistical skill of the new algorithm. Application to MODIS imagery acquired during the February–March 2015 OUTPACE campaign reveals the presence of surface blooms Northwest and East of New Caledonia and near 20° S–172° W in qualitative agreement with measured nitrogen fixation rates. The new algorithm, however, fails to detect sub-surface booms evidenced in trichome counts. Improving Trichodesmium detection requires measuring ocean color at higher spectral and spatial (< 250 m) resolution than MODIS, taking into account environment properties (wind, sea surface temperature, ...), fluorescence, and spatial structure of filaments, and a better understanding of Trichodesmium dynamics, including aggregation processes to generate surface mats. Such sub-mesoscales aggregation processes for Trichodesmium are yet to be understood.