The uncertainty of the terrestrial carbon cycle is a consequence of sparse and irregular observations on the ground. National forest inventories of countries in the boreal zone are limited by adverse conditions during winter-time and vastity of the forest area to be surveyed. The synoptic view by satellite remote sensing fills this temporal and spatial gap. The record of backscattered intensity measured since 1991 by satellite scatterometers operating at C-band (wavelength of 6 cm) spans almost three decades by now. The observations have a spatial resolution of 25 km on the ground and an almost daily frequency, allowing for almost unbiased retrievals of above-ground biomass (AGB), and thereof carbon, when averaging over one year of measurements. The uncertainty of our AGB estimates is around 30% of the estimated value at the pixel level, providing strong confidence in multi-decadal AGB trends in biomes. Almost 30 years of AGB estimates for the boreal zone reveal a rather constant average AGB at around 55 Mg ha-1. The temporal patterns of the AGB stocks mimic those of the forest area estimated from the Climate Change Initiative Land Cover dataset, with an increase in the 1990 and 2000 decade and a decrease in the 2010 decade. The variability of the AGB stocks is, however, small (+/- 2 Pg) for an average of 75 Pg throughout the three decades. Yet, our estimates indicate different AGB dynamics in space. The AGB stocks of forests in Russia increased by approximately 4% between 1992 and 2020 whereas Canadian forests have recovered towards the end of the 2010s the significant loss of almost 10 Pg since 1992 due primarily to insect damages and forest fires. Our results indicate that C-band scatterometer observations from space have strong potential to contribute to the quantification of terrestrial carbon stocks.

Russian forest is of global importance on climate change agenda considering its potential for absorption and accumulation of the atmospheric carbon. However, existing quantitative estimates of Russian forests’ carbon budget are of significant level of uncertainty. One of the most obvious reasons for such uncertainty is not sufficiently reliable and up-to-date information on the forests and their dynamics. Information necessary for carbon budget estimation includes data on land cover and forest characteristics (growing stock volume, species composition, age, site-index) as well as data on natural and anthropogenic forest disturbances. The existing remote sensing methods can provide significant part of missing information about the land cover and forest characteristics for the national-scale carbon budget estimation and monitoring.

The multi-annual time series of 250 m MODIS data were used to develop the set of harmonised thematic maps characterising the entire Russian forests during the period of 2006-2020, including:

- land cover - 25 thematic classes of forest and non-forest lands;

- dominant forest tree species - 12 main tree species;

- forest growing stock volume;

- forest site-index classes;

- forest age classes;

- horizontal forest structure (forest area fraction, canopy closure, relative stock);

- forest burnt area and severity;

- biogenic forest mortality (limited only by evergreen coniferous forests).

The retrieved the forest stock volume time series (considering both growing and dead trees) shows significant positive trend in range between 112,1x109 m3 and 117,1x109 m3 with mean annual increment at about 314,5x106 m3 . y-1. The forest site-index and age maps were developed based on assimilation of the derived time-series data on growing stock volume into empirical forest growth models. The retrieved multi-annual estimates of carbon balance in Russian forests show its significant variability strongly depending on fire disturbances.

The study was supported by the Russian Science Foundation (project no. RSF-19-77-30015).

Since the collapse of the Soviet Union and transition to a new forest inventory system, Russia has reported to the Forest Resources Assessment of the Forest and Agriculture Organization almost no changes in growing stock (+1.8%) and biomass (+0.6%). Yet remote sensing (RS) products indicate increased vegetation productivity, tree cover and above-ground biomass. We calibrated two global RS-based growing stock volume (GSV) maps: GlobBiomass and CCI Biomass with ca 9,000 National Forest Inventory forest plots first time made available for this study. Our GSV estimate for the year 2014 is 111.13±1.2 109 m3 for the forest area reported in the State Forest Register (SFR) or 39% higher compared to SFR. Additional 7.1 109 m3 can be attributed to the larger forest area observed from the space, including expansion of forests to the north and higher elevations, in abandoned arable land, as well as the inclusion of parks, gardens and other trees outside of forest, which were not counted as forest in the SFR. Using the last Soviet Union report (1988) as a reference, we can judge that Russian forests have accumulated 1163×106 m3 yr-1 of GSV between 1988–2014. Our estimate of the growing stock of managed forests is 94.2 109 m3, which corresponds to sequestration of 354 Tg C yr-1 in live biomass over 1988–2014, or 47% higher than reported in the National Greenhouse Gases Inventory.

Remotely sensed land cover (LC) products map spatial distributions of different forest types which are needed to project future changes in forest cover and to improve carbon budget estimation. Since 2000, a large number of forest extent maps have been published, yet many global LC-products use simple tree cover (referred to also as ‘forest cover’ or even ‘vegetation cover’) definition to differentiate forests from non-forests. Although numerous papers have investigated and compared forest extent data, inter-comparison of different forest classifications employed by varied LC-products is often overlooked due to challenges involved (i.e. in non-standard and overlapping LC-class definitions). In this study, we assessed forest cover representations of three different annual (year of 2017) global LC-products (i.e. MODIS VCF (MOD44B, Collection 6 (C6)), MCD12Q1 (C6), and CCI LC (v.2.1.1)) using Finnish Multi-Source National Forest Inventory (MS-NFI) data for the year 2017 as a reference. To accomplish the task, we first developed an inter-comparison approach for analyzing spatial representations of coniferous and deciduous species forest cover values. We used Finland as a case study area, although the method is suited for even larger geographical regions. Inter-comparison of the LC-products revealed notable differences in conifer and deciduous species spatial distributions. Based on our results forest cover was either overestimated or underestimated depending on the LC-product. Classification accuracy of different products varied between 42 % and 75 %. Overall, spatial mapping of coniferous and deciduous tree covers was the best represented by the CCI LC-product as compared with the reference MS-NFI data. Considering that these global LC products are employed by e.g., today’s land surface models, more attention should be paid to tree cover classifications and spatial mappings of coniferous and deciduous species of different LC data products.

The Arctic-Boreal Vulnerability Experiment (ABoVE), a field campaign sponsored and initiated by NASA’s Terrestrial Ecology Program, is a large-scale study of changes to terrestrial and freshwater ecosystems in the Arctic and boreal regions of western North America and the implications of these changes for local, regional, and global social-ecological systems. Over 110 projects and 792 scientists are affiliated with ABoVE, and research from the campaign has resulted in over 300 publications. Many of the science questions for ABoVE address issues of carbon cycling and carbon biogeochemistry. As part of ABoVE, an airborne campaign (the ABoVE Airborne Campaign, or AAC) was initially conducted from April through November 2017, with follow-on measurements made in 2018 and 2019. The 2017 AAC involved ten aircraft in more than 200 science flights and surveyed over 4 million km2 in Alaska and northwestern Canada. The airborne strategy involved collecting domain-wide measurements with L- and P- band synthetic aperture radar (SAR), imaging spectroscopy, full waveform LIDAR, and atmospheric carbon dioxide and methane with focused studies using Ka-band SAR and solar induced chlorophyll fluorescence. Additional measurements were coordinated with the NEON Airborne Observing Platform, the ASCENDS instrument suite, and the ATom investigation.

Many flights during the AAC were coordinated with same-day ground-based measurements to link process-level studies with geospatial data products derived from satellite sensors. The data collected spans the critical intermediate space and time scales that are essential for a comprehensive understanding of scaling issues across the ABoVE Study Domain and extrapolation to the pan-Arctic. A recent result using this scaling strategy showed the fine-scale spatial distribution of intense methane emission hotspots near water bodies within the ABoVE domain (Elder et al., 2020). Datasets from many of the airborne instruments are freely available for download and use, and links to these datasets will be provided.

Measurements of the atmospheric sources and sinks of carbon dioxide (CO2) and methane (CH4) in the pan-Arctic domain are extremely sparse that limits our knowledge of carbon cycling over this dramatically sensitive environment and making predictions about a fate of carbon conserved in currently frozen ground. Especially critical are the gaps in the arctic latitudes of Siberia, covered by the vast permafrost underlain tundra, where only few continuous atmospheric observation stations are operational.

We present the temporal variations of atmospheric CO2 and CH4 based on the first two years (2018 - 2020) of accurate continuous observations at the new atmospheric carbon observation station located near the Dikson settlement (73.33° N, 80.34° E) on the seashore of the Taimyr Peninsula in Siberia. Atmospheric mole fractions of CO2, CH4, and H2O are continuously measured by a CRDS analyser (G2301-f, Picarro Inc.), which is regularly calibrated against WMO-traceable reference gases. Associated meteorological variables permit a background for screening the greenhouse gases (GHG) data records and provide a description of the climate variability for the environment. Observation records deal with the daytime mixed layer and considered as representative throughout the vast area. The strict filtering allowed identifying most of the values that are representative of well-mixed air in study area and not affected by local contamination events.

Here we summarize the scientific rationale of the new site, give technical details of the instrumental setup, analyse the local environments and present CO2 and CH4 fluctuations in the arctic atmosphere. Along with the temporal variability, we provide an overview of the angular distribution of detected GHG signals and their input to the atmospheric fluctuations on the measurement site. The reported study was funded by RFBR, Krasnoyarsk Territory and Krasnoyarsk Regional Fund of Science, project number 20-45-242908, RFBR project 18-05-60203 and by the Max Planck Society (Germany).

Bud development and leaf senescence represent important phenological events of forest ecosystems and define the growing period of trees. Satellite remote sensing and phenocams based near-surface remote sensing data provides continuous coverage to carry out time-series analysis of canopy greenness to measure leaf and canopy phenology. Questions remain on accurate calibrations of continuous records of canopy greenness collected by optical satellite and phenocams data to represent variations in the bud phenology of evergreen species, which show limited seasonal changes in canopy greenness. In this study, we compared bud phenology of black spruce [Picea mariana (Mill.) B.S.P] with the canopy greenness, represented by phenocam derived Green Chromatic Coordinate (GCC), and bi-weekly MODIS satellite derived Normalized Difference Vegetation Index (NDVI) during 2015, 2017 and 2018 in a black spruce stands of Quebec, Canada. We applied double-logistic functions and calibrated using ordinal logit models to estimate the probability of observing the sequential phenological phases of bud burst and bud set along the GCC and NDVI. The beginning of bud burst was observed in mid-May for both NDVI and GCC, when their values reached 80.5% and 72% of its maximum amplitude, respectively. However, ending of bud burst occurred at the beginning of September for NDVI and in mid-June for GCC, when their values reached 92.2% and 94% of its maximum amplitude, respectively. These NDVI and GCC values are reliable thresholds indicating the boundaries of the growing period. Our study builds a bridge between field observations and satellite and near surface remote sensing data, providing a statistically sound protocol for reliable timings of bud phenological events using calibrated GCC and NDVI time series data.

It is important to understand the physically based causality of observed satellite products for a variety of reasons. These can be such as developing field-sampling-independent vegetation characterization models and better understanding of the radiation balance of land covers. The forest reflectance and transmittance model (FRT) is a relatively well known and accepted radiative transfer model that helps advance towards this objective. For the first time, we tried to quantify the efficacy of this reflectance model over a large region by carrying out a simulation study over a large number (~24,000) of georeferenced forested plots in Southern Finland. The choice of these plots was such that high-quality Landsat-8 imagery was also available for them, within ± 6 months of the establishment and inventory of the plots. We then compared the FRT simulated BRF (bidirectional reflectance factor) with the satellite-estimated surface reflectance values. The general trend seen over almost all bands and for all three tree species studied is that of FRT overestimating BRF, compared to satellite estimates. Moreover, we found that seedling, sapling and low-timber-volume stands were especially affected by this bias present in FRT. We also noticed a similar trend of overestimation for most birch (Betula pendula and Betula pubescens) stands, which is mostly connected with their broadleaved nature. We also discuss some spatial and temporal trends in bias seen over Southern Finland, which would be useful for further tuning and development of the FRT model and its associated input parameters. In conclusion, our results indicate that the combination of a large dataset of forest plots and freely available Landsat satellite data is a compelling verification dataset for radiative transfer models.

Estimation of tree species distributions and carbon stocks in the boreal forest of interior Alaska is critical to our understanding of how climate change can impact ecosystem services in this region, including carbon sequestration/emissions, habitat quality, and bioenergy/timber production. However, the high cost and extreme logistical challenges associated with establishing in situ field observations in this remote forest biome requires inventory designs that can efficiently leverage remote sensing measurements – from multiple sources – to reduce the number of expensive field plots. In the period 2014-2020, the US Forest Service and NASA implemented a new forest monitoring system in interior Alaska using a combination of field inventory measurements and high-resolution airborne remote sensing data, including imaging spectroscopy, 3-cm RGB stereo imagery and laser scanning, allowing for estimation of forest attributes in the relatively more accessible east-central regions of interior Alaska (Tanana, Susitna, and Copper river watersheds). However, given the increased costs of field sampling in western Alaska, it is expected that the relative contributions from field sampling and remote sensing in the forest monitoring design will be adjusted, with increased reliance on remote sensing. In order to determine the optimal combination of field and remote sensing data to meet monitoring goals in western Alaska, we carried out a pilot project in the Susitna-Copper region, where a unique set of remote sensing measurements (lidar-derived forest structure measurements and photo-interpreted forest attributes) were compared to field-based forest condition attributes, including forest type, disturbance class, stand size class, stand density, canopy cover, and land cover class. Using the results from this pilot project, we were able to assess how changes to the forest monitoring design criteria (plot grid sampling intensity/configuration, remote sensing acquisition parameters, etc.) in western Alaska would affect the quality of the monitoring products, with particular focus on the the precision of estimates for area and forest biomass by condition class.

Spatially explicit information on forest management at a global scale is critical for understanding the status of forests and for planning sustainable forest management and restoration. However, such data are currently lacking as there is no map available with consistent information on intact forests, managed forests with natural regeneration, planted forests, short rotation woody plantations, oil palms, and agroforestry. To address this issue, we launched a series of expert and crowdsourcing campaigns using the Geo-Wiki (https://www.geo-wiki.org/) application, involving forest experts from different world regions, to classify tree cover at 226k locations using very high-resolution images and remote-sensing data. We then combined the expert reference samples with time series from PROBA-V satellite imagery to create a global wall-to-wall map of forest management at a 100 m resolution for the year 2015. The overall accuracy of the map is 82±0.01%. The map present the status of forest ecosystems and can be used for assessments of protected and production forests, potential use of forest resources, habitats, and biodiversity.