Carbon dioxide and methane emissions are the two primary anthropogenic climate-forcing agents and an important source of uncertainty in the global carbon budget. Uncertainties are further magnified when emissions occur at fine spatial scales (<1 km), making attribution challenging. We present the first observations from NASA’s Earth Surface Mineral Dust Source Investigation (EMIT) imaging spectrometer showing quantification and attribution of fine-scale methane (0.3 to 73 tonnes CH ₄ hour ⁻¹ ) and carbon dioxide sources (1571 to 3511 tonnes CO ₂ hour ⁻¹ ) spanning the oil and gas, waste, and energy sectors. For selected countries observed during the first 30 days of EMIT operations, methane emissions varied at a regional scale, with the largest total emissions observed for Turkmenistan (731 ± 148 tonnes CH ₄ hour ⁻¹ ). These results highlight the contributions of current and planned point source imagers in closing global carbon budgets.

Accurate and spatially-explicit maps of tropical forest carbon stocks are needed to implement carbon offset mechanisms such as REDD+ (Reduced Deforestation and Degradation Plus). The Random Forest machine learning algorithm may aid carbon mapping applications using remotely-sensed data. However, Random Forest has never been compared to traditional and potentially more reliable techniques such as regionally stratified sampling and upscaling, and it has rarely been employed with spatial data. Here, we evaluated the performance of Random Forest in upscaling airborne LiDAR (Light Detection and Ranging)-based carbon estimates compared to the stratification approach over a 16-million hectare focal area of the Western Amazon. We considered two runs of Random Forest, both with and without spatial contextual modeling by including—in the latter case—x, and y position directly in the model. In each case, we set aside 8 million hectares (i.e., half of the focal area) for validation; this rigorous test of Random Forest went above and beyond the internal validation normally compiled by the algorithm (i.e., called “out-of-bag”), which proved insufficient for this spatial application. In this heterogeneous region of Northern Peru, the model with spatial context was the best preforming run of Random Forest, and explained 59% of LiDAR-based carbon estimates within the validation area, compared to 37% for stratification or 43% by Random Forest without spatial context. With the 60% improvement in explained variation, RMSE against validation LiDAR samples improved from 33 to 26 Mg C ha⁻¹ when using Random Forest with spatial context. Our results suggest that spatial context should be considered when using Random Forest, and that doing so may result in substantially improved carbon stock modeling for purposes of climate change mitigation.

High-resolution mapping of tropical forest carbon stocks can assist forest management and improve implementation of large-scale carbon retention and enhancement programs. Previous high-resolution approaches have relied on field plot and/or light detection and ranging (LiDAR) samples of aboveground carbon density, which are typically upscaled to larger geographic areas using stratification maps. Such efforts often rely on detailed vegetation maps to stratify the region for sampling, but existing tropical forest maps are often too coarse and field plots too sparse for high-resolution carbon assessments. We developed a top-down approach for high-resolution carbon mapping in a 16.5 million ha region (> 40%) of the Colombian Amazon – a remote landscape seldom documented. We report on three advances for large-scale carbon mapping: (i) employing a universal approach to airborne LiDAR-calibration with limited field data; (ii) quantifying environmental controls over carbon densities; and (iii) developing stratification- and regression-based approaches for scaling up to regions outside of LiDAR coverage. We found that carbon stocks are predicted by a combination of satellite-derived elevation, fractional canopy cover and terrain ruggedness, allowing upscaling of the LiDAR samples to the full 16.5 million ha region. LiDAR-derived carbon maps have 14% uncertainty at 1 ha resolution, and the regional map based on stratification has 28% uncertainty in any given hectare. High-resolution approaches with quantifiable pixel-scale uncertainties will provide the most confidence for monitoring changes in tropical forest carbon stocks. Improved confidence will allow resource managers and decision makers to more rapidly and effectively implement actions that better conserve and utilize forests in tropical regions.

Significance: Land use is a principal driver of carbon emissions, either directly through land change processes such as deforestation or indirectly via transportation and industries supporting natural resource use. To minimize the effects of land use on the climate system, natural ecosystems are needed to offset gross emissions through carbon sequestration. Managing this critically important service must be achieved tactically if it is to be cost-effective. We have developed a high-resolution carbon mapping approach that can identify biogeographically explicit targets for carbon storage enhancement among all landholders within a country. Applying our approach to Perú reveals carbon threats and protections, as well as major opportunities for using ecosystems to sequester carbon. Our approach is scalable to any tropical forest country.