Linking satellite-based spring phenology to temperate deciduous broadleaf forest photosynthesis activity

Satellite-based remote sensed phenology has been widely used to assess global climate change. However, it is constrained by uncertain linkages with photosynthesis activity. Two dynamic threshold methods were employed to retrieve spring phenology metrics from four Moderate Resolution Imaging Spectroradiometer (MODIS) products, including fraction of Absorbed Photosynthetically Active Radiation (fAPAR), Leaf Area Index (LAI), Normalized Difference Vegetation Index (NDVI), and Enhanced Vegetation Index (EVI) for three temperate deciduous broadleaf forests in North America between 2001 and 2009. These MODIS-based spring phenology metrics were subsequently linked to the photosynthetic curves (daily gross primary productivity, GPP) measured by an eddy covariance flux tower. The 20% dynamic threshold spring onset metrics from MODIS products were closer to the photosynthesis onset metrics at the date of 2% GPP increase for NDVI and fAPAR, and closer to the date of 5% and 10% increase of GPP for EVI and LAI, respectively. The 50% dynamic threshold onset metrics were closer to the photosynthesis onset metrics at the date of 10% GPP increase for NDVI, and closer to the date of 20% GPP increase for fAPAR, LAI and EVI, respectively. These results can improve our knowledge on the photosynthesis activity status of remotely sensed spring phenology metrics.     

Major forest changes and land cover transitions based on plant functional types derived from the ESA CCI Land Cover product

Land use and land cover change are of prime concern due to their impacts on CO₂ emissions, climate change and ecological services. New global land cover products at 300 m resolution from the European Space Agency (ESA) Climate Change Initiative Land Cover (CCI LC) project for epochs centered around 2000, 2005 and 2010 were analyzed to investigate forest area change and land cover transitions. Plant functional types (PFTs) fractions were derived from these land cover products according to a conversion table. The gross global forest loss between 2000 and 2010 is 172,171 km², accounting for 0.6% of the global forest area in year 2000. The forest changes are mainly distributed in tropical areas such as Brazil and Indonesia. Forest gains were only observed between 2005 and 2010 with a global area of 9844 km², mostly from crops in Southeast Asia and South America. The predominant PFT transition is deforestation from forest to crop, accounting for four-fifths of the total increase of cropland area between 2000 and 2010. The transitions from forest to bare soil, shrub, and grass also contributed strongly to the total areal change in PFTs. Different PFT transition matrices and composition patterns were found in different regions. The highest fractions of forest to bare soil transitions were found in the United States and Canada, reflecting forest management practices. Most of the degradation from grassland and shrubland to bare soil occurred in boreal regions. The areal percentage of forest loss and land cover transitions generally decreased from 2000–2005 to 2005–2010. Different data sources and uncertainty in the conversion factors (converting from original LC classes to PFTs) contribute to the discrepancy in the values of change in absolute forest area.     

Integration of ground and satellite data to model Mediterranean forest processes

The current work presents the testing of a modeling strategy that has been recently developed to simulate the gross and net carbon fluxes of Mediterranean forest ecosystems. The strategy is based on the use of a NDVI-driven parametric model, C-Fix, and of a biogeochemical model, BIOME-BGC, whose outputs are combined to simulate the behavior of forest ecosystems at different development stages. The performances of the modeling strategy are evaluated in three Italian study sites (San Rossore, Lecceto and Pianosa), where carbon fluxes are being measured through the eddy correlation technique. These sites are characterized by variable Mediterranean climates and are covered by different types of forest vegetation (pine wood, Holm oak forest and Macchia, respectively). The results of the tests indicate that the modeling strategy is generally capable of reproducing monthly GPP and NEE patterns in all three study sites. The highest accuracy is obtained in the most mature, homogenous pine wood of San Rossore, while the worst results are found in the Lecceto forest, where there are the most heterogeneous terrain, soil and vegetation conditions. The main error sources are identified in the inaccurate definition of the model inputs, particularly those regulating the site water budgets, which exert a strong control on forest productivity during the Mediterranean summer dry season. In general, the incorporation of NDVI-derived fAPAR estimates corrects for most of these errors and renders the forest flux simulations more stable and accurate.     

Greenhouse gas emissions from tropical forest degradation: an underestimated source

Background

The degradation of forests in developing countries, particularly those within tropical and subtropical latitudes, is perceived to be an important contributor to global greenhouse gas emissions. However, the impacts of forest degradation are understudied and poorly understood, largely because international emission reduction programs have focused on deforestation, which is easier to detect and thus more readily monitored. To better understand and seize opportunities for addressing climate change it will be essential to improve knowledge of greenhouse gas emissions from forest degradation.

Results

Here we provide a consistent estimation of forest degradation emissions between 2005 and 2010 across 74 developing countries covering 2.2 billion hectares of forests. We estimated annual emissions of 2.1 billion tons of carbon dioxide, of which 53% were derived from timber harvest, 30% from woodfuel harvest and 17% from forest fire. These percentages differed by region: timber harvest was as high as 69% in South and Central America and just 31% in Africa; woodfuel harvest was 35% in Asia, and just 10% in South and Central America; and fire ranged from 33% in Africa to only 5% in Asia. Of the total emissions from deforestation and forest degradation, forest degradation accounted for 25%. In 28 of the 74 countries, emissions from forest degradation exceeded those from deforestation.

Conclusions

The results of this study clearly demonstrate the importance of accounting greenhouse gases from forest degradation by human activities. The scale of emissions presented indicates that the exclusion of forest degradation from national and international GHG accounting is distorting. This work helps identify where emissions are likely significant, but policy developments are needed to guide when and how accounting should be undertaken. Furthermore, ongoing research is needed to create and enhance cost-effective accounting approaches.

     

Forest classification and impact of BIOMASS resolution on forest area and aboveground biomass estimation

The European Space Agency (ESA) is currently implementing the BIOMASS mission as 7th Earth Explorer satellite. BIOMASS will provide for the first time global forest aboveground biomass estimates based on P-band synthetic aperture radar (SAR) imagery. This paper addresses an often overlooked element of the data processing chain required to ensure reliable and accurate forest biomass estimates: accurate identification of forest areas ahead of the inversion of radar data into forest biomass estimates.The use of the P-band data from BIOMASS itself for the classification into forest and non-forest land cover types is assessed in this paper. For airborne data in tropical, hemi-boreal and boreal forests we demonstrate that classification accuracies from 90 up to 97% can be achieved using radar backscatter and phase information. However, spaceborne data will have a lower resolution and higher noise level compared to airborne data and a higher probability of mixed pixels containing multiple land cover types. Therefore, airborne data was reduced to 50 m, 100 m and 200 m resolution. The analysis revealed that about 50–60% of the area within the resolution level must be covered by forest to classify a pixel with higher probability as forest compared to non-forest. This results in forest omission and commission leading to similar forest area estimation over all resolutions. However, the forest omission resulted in a biased underestimated biomass, which was not equaled by the forest commission. The results underline the necessity of a highly accurate pre-classification of SAR data for an accurate unbiased aboveground biomass estimation.     

Soil erosion modelled with USLE and PESERA using QuickBird derived vegetation parameters in an alpine catchment

The focus of soil erosion research in the Alps has been in two categories: (i) on-site measurements, which are rather small scale point measurements on selected plots often constrained to irrigation experiments or (ii) off-site quantification of sediment delivery at the outlet of the catchment. Results of both categories pointed towards the importance of an intact vegetation cover to prevent soil loss. With the recent availability of high-resolution satellites such as IKONOS and QuickBird options for detecting and monitoring vegetation parameters in heterogeneous terrain have increased. The aim of this study is to evaluate the usefulness of QuickBird derived vegetation parameters in soil erosion models for alpine sites by comparison to Cesium-137 (Cs-137) derived soil erosion estimates. The study site (67 km²) is located in the Central Swiss Alps (Urseren Valley) and is characterised by scarce forest cover and strong anthropogenic influences due to grassland farming for centuries. A fractional vegetation cover (FVC) map for grassland and detailed land-cover maps are available from linear spectral unmixing and supervised classification of QuickBird imagery. The maps were introduced to the Pan-European Soil Erosion Risk Assessment (PESERA) model as well as to the Universal Soil Loss Equation (USLE). Regarding the latter model, the FVC was indirectly incorporated by adapting the C factor. Both models show an increase in absolute soil erosion values when FVC is considered. In contrast to USLE and the Cs-137 soil erosion rates, PESERA estimates are low. For the USLE model also the spatial patterns improved and showed “hotspots” of high erosion of up to 16 t ha⁻¹ a⁻¹. In conclusion field measurements of Cs-137 confirmed the improvement of soil erosion estimates using the satellite-derived vegetation data.     

The potential of the greenness and radiation (GR) model to interpret 8-day gross primary production of vegetation

Remote sensing of vegetation gross primary production (GPP) is an important step to analyze terrestrial carbon (C) cycles in response to changing climate. The availability of global networks of C flux measurements provides a valuable opportunity to develop remote sensing based GPP algorithms and test their performances across diverse regions and plant functional types (PFTs). Using 70 global C flux measurements including 24 non-forest (NF), 17 deciduous forest (DF) and 29 evergreen forest (EF), we present the evaluation of an upscaled remote sensing based greenness and radiation (GR) model for GPP estimation. This model is developed using enhanced vegetation index (EVI) and land surface temperature (LST) from the Moderate Resolution Imaging Spectroradiometer (MODIS) and global course resolution radiation data from the National Center for Environmental Prediction (NCEP). Model calibration was achieved using statistical parameters of both EVI and LST fitted for different PFTs. Our results indicate that compared to the standard MODIS GPP product, the calibrated GR model improved the GPP accuracy by reducing the root mean square errors (RMSE) by 16%, 30% and 11% for the NF, DF and EF sites, respectively. The standard MODIS and GR model intercomparisons at individual sites for GPP estimation also showed that GR model performs better in terms of model accuracy and stability. This evaluation demonstrates the potential use of the GR model in capturing short-term GPP variations in areas lacking ground measurements for most of vegetated ecosystems globally.     

Remote sensing of fire severity: assessing the performance of the normalized burn ratio

Several studies have used satellite data to map different levels of fire severity present within burned areas. Increasingly, fire severity has been estimated using a spectral index called the normalized burn ratio (NBR). This letter assesses the performance of the NBR against ideal requirements of a spectral index designed to measure fire severity. According to index theory, the NBR would be optimal for quantifying fire severity if the trajectory in spectral feature space caused by different levels of severity occurred perpendicular to the NBR isolines. We assess how well NBR meets this condition using reflectance data sensed before and shortly after fires in the South African savanna, Australian savanna, Russian Federation boreal forest, and South American tropical forest. Although previous studies report high correlation between fire severity measured in the field- and satellite-derived NBR, our results do not provide evidence that the performance of the NBR is optimal in describing fire severity shortly after fire occurrence. Spectral displacements due to burning occur in numerous directions relative to the NBR index isolines, suggesting that the NBR may not be primarily and consistently sensitive to fire severity. Findings suggest that the development of the next generation of methods to estimate fire severity remotely should incorporate knowledge of how fires of different severity displace the position of prefire vegetation in multispectral space.     

Warming trends in Patagonian subantartic forest

The forests in the Aysén region (ca. 43–49 °S, Chile) have a high degree of wilderness and cover more than 4.8 million hectares, making it one of the largest areas of subantarctic forest in the Southern Hemisphere. The impact of global warming on this region is poorly documented. The main objective of this work was to analyze the normalized difference vegetation index (NDVI), land surface temperature (LST) and precipitation over Aysen forests in the context of ongoing global warming. We used average monthly images of LST and NDVI derived from the MODIS sensor covering the period 2001–2016 and precipitation from gridded datasets. The Aysén region was divided into three nested spatial scales: i) regional, ii) regional considering only forests, iii) local scale considering an evergreen subantarctic forest area covering around 5 × 5 km and a local deciduous forest area (dominated by Nothofagus pumilio). Trend analysis showed a warming rate of +0.78 K/decade (p ≤ 0.05) over the subantarctic forest zone, greening of +0.01/decade for NDVI (p ≤ 0.05) over the western zone, and a drying trend (p ≤ 0.05) over the eastern zone. The minimum temperature anomalies showed an increase of about 4.5 K during the period under analysis. LST, NDVI and precipitation were also analyzed here. The recent trends in temperature, greening and precipitation over the forests of Aysén detected in this research contribute to a better understanding of global warming impacts on subantarctic forests in the southern tip of South America. Nevertheless, to get a better estimation of the impact of global warming at multiple scales is needed to have better quality and quantity of data in situ.     

Use of principal components of UAV-acquired narrow-band multispectral imagery to map the diverse low stature vegetation fAPAR

The fraction of absorbed photosynthetically active radiation (fAPAR) is an important plant physiological index that is used to assess the ability of vegetation to absorb PAR, which is utilized to sequester carbon in the atmosphere. This index is also important for monitoring plant health and productivity, which has been widely used to monitor low stature crops and is a crucial metric for food security assessment. The fAPAR has been commonly correlated with a greenness index derived from spaceborne optical imagery, but the relatively coarse spatial or temporal resolution may prohibit its application on complex land surfaces. In addition, the relationships between fAPAR and remotely sensed greenness data may be influenced by the heterogeneity of canopies. Multispectral and hyperspectral unmanned aerial vehicle (UAV) imaging systems, conversely, can provide several spectral bands at sub-meter resolutions, permitting precise estimation of fAPAR using chemometrics. However, the data pre-processing procedures are cumbersome, which makes large-scale mapping challenging. In this study, we applied a set of well-verified image processing protocols and a chemometric model to a lightweight, frame-based and narrow-band (10 nm) UAV imaging system to estimate the fAPAR over a relatively large cultivated land area with a variety of low stature vegetation of tropical crops along with native and non-native grasses. A principal component regression was applied to 12 bands of spectral reflectance data to minimize the collinearity issue and compress the data variation. Stepwise regression was employed to reduce the data dimensionality, and the first, third and fifth components were selected to estimate the fAPAR. Our results indicate that 77% of the fAPAR variation was explained by the model. All bands that are sensitive to foliar pigment concentrations, canopy structure and/or leaf water content may contribute to the estimation, especially those located close to (720 nm) or within (750 nm and 780 nm) the near-infrared spectral region. This study demonstrates that this narrow-band frame-based UAV system would be useful for vegetation monitoring. With proper pre-flight planning and hardware improvement, the mapping of a narrow-band multispectral UAV system could be comparable to that of a manned aircraft system.     

Estimation of daily average net radiation and its variation over West Bengal,India using MODIS products

A simple and exclusively satellite-based instantaneous net radiation (INR) and its components have been estimated and it is based on the method proposed. The method has been used to check its suitability over West Bengal, India on clear sky days which is eliminating the need for ground-based observations. The method explicitly recognizes the need for spatially varied input parameters to estimate net radiation using remote sensing information and provides a distributed daily average net radiation map with finer spatial resolution. The satellite-derived radiative flux components were compared with the Eddy Flux Tower measurements located at Bonnie camp within mangrove forest. The bias, root-mean square error and R² between modelled and measured INR are 52.07, 49.5 W m^?2 and 0.74, respectively which are within the acceptable limit. The errors in the downward short wave and long wave as well as upward short-wave and long-wave radiative flux are 13.78, 3.54, 12.41 and 3.54%, respectively. Further, variations of all the radiative flux components were analysed over different eco-systems.     

Integrating global land cover products for improved forest cover characterization: an application in North America

Six widely used coarse-resolution global land cover data-sets – Global Land Cover Characterization (GLCC), Global Land Cover 2000 (GLC2000), GlobCover land cover product (GlobCover), MODIS land cover product (MODIS LC), the University of Maryland land cover product (UMD LC), and the MODIS Vegetation Continuous Fields tree cover layer (MODIS VCF) disagree substantially in their estimates of forest cover. Employing a regression tree model trained on higher-resolution, Landsat-based data, these multisource multiresolution maps were integrated for an improved characterization of forest cover over North America. Evaluated using a withheld test sample, the integrated percent forest cover (IPFC) data-set has a root mean square error of 11.75% – substantially better than the 17.37% of GLCC, 17.61% of GLC2000, 17.96% of GlobCover, 15.23% of MODIS LC, 19.25% of MODIS VCF, and 15.15% of UMD LC, respectively. Although demonstrated for forest, this approach based on integration of multiple products has potential for improved characterization of other land cover types as well.     

Use of local and global maps of forest canopy height and aboveground biomass to enhance local estimates of biomass in miombo woodlands in Tanzania

Field surveys are often a primary source of aboveground biomass (AGB) data, but plot-based estimates of parameters related to AGB are often not sufficiently precise, particularly not in tropical countries. Remotely sensed data may complement field data and thus help to increase the precision of estimates and circumvent some of the problems with missing sample observations in inaccessible areas. Here, we report the results of a study conducted in a 15,867 km² area in the dry miombo woodlands of Tanzania, to quantify the contribution of existing canopy height and biomass maps to improving the precision of canopy height and AGB estimates locally. A local and a global height map and three global biomass maps, and a probability sample of 513 inventory plots were subject to analysis. Model-assisted sampling estimators were used to estimate mean height and AGB across the study area using the original maps and then with the maps calibrated with local inventory plots. Large systematic map errors – positive or negative – were found for all the maps, with systematic errors as great as 60–70 %. After being calibrated locally, the maps contributed substantially to increasing the precision of both mean height and mean AGB estimates, with relative efficiencies (variance of the field-based estimates relative to the variance of the map-assisted estimates) of 1.3–2.7 for the overall estimates. The study, although focused on a relatively small area of dry tropical forests, illustrates the potential strengths and weaknesses of existing global forest height and biomass maps based on remotely sensed data and universal prediction models. Our results suggest that the use of regional or local inventory data for calibration can substantially increase the precision of map-based estimates and their applications in assessing forest carbon stocks for emission reduction programs and policy and financial decisions.     

Comparison of eMODIS and MOD/MYD13A2 NDVI products during 2012–2014 spring green-up periods in Alaska and northwest Canada

Accurate monitoring of vegetation dynamics is required to understand the inter-annual variability and long term trends in terrestrial carbon exchange in tundra and boreal ecoregions. In western North America, two Normalized Vegetation Index (NDVI) products based on spectral reflectance data from the Moderate Resolution Imaging Spectroradiometer (MODIS) are available. The MOD/MYD13A2 NDVI product is available as a 16-day composite product in a sinusoidal projection as global hdf tiles. The eMODIS Alaska NDVI product is available as a 7-day composite geotif product in a regional equal area conic projection covering Alaska and the entire Yukon River Basin. These two NDVI products were compared for the 2012–2014 late May–late June spring green-up periods in Alaska and the Yukon Territory. Relative to the MOD/MYD13A2 NDVI product, it is likely that the eMODIS NDVI product contained more cloud-contaminated NDVI values. For example, the MOD/MYD13A2 product flagged substantially fewer pixels as “good quality” in each 16-day composite period compared to the corresponding MODIS Alaska NDVI product from a 7-day composite period. During the spring green-up period, when field-based NDVI increases, the eMODIS NDVI product averaged 43 percent of pixels that declined by at least 0.05 NDVI between 2 composite periods, consistent with cloud-contamination problems, while the MOD/MYD13A2 NDVI averaged only 6 percent of pixels. Based on a cloudy Landsat-8 scene, the eMODIS compositing process selected 23 percent pixels, while the MOD/MYD13A2 compositing process selected less than 0.003 percent pixels. Based on the results, it appears that the MOD/MYD13A2 NDVI product is superior for scientific applications based on NDVI phenology in the tundra and boreal regions of northwestern North America.     

Modeling three-dimensional forest structures to drive canopy radiative transfer simulations of bidirectional reflectance factor

Three-dimensional (3-D) Monte Carlo-based radiative transfer (MCRT) models are usually used for benchmarking in intercomparisons of the canopy radiative transfer (RT) simulations. However, the 3-D MCRT models are rarely applied to develop remote sensing algorithms to estimate essential climate variables of forests, due mainly to the difficulties in obtaining realistic stand structures for different forest biomes over regional to global scales. Fortunately, some of important tree structure parameters such as canopy height and tree density distribution have been available globally. This enables to run the intermediate complexities of the 3-D MCRT models. We consequently developed a statistical approach to generate forest structures with intermediate complexities depending on the inputs of canopy height and tree density. It aims at facilitating applications of the 3-D MCRT models to develop remote sensing retrieval algorithms. The proposed approach was evaluated using field measurements of two boreal forest stands at Estonia and USA, respectively. Results demonstrated that the simulations of bidirectional reflectance factor (BRF) based on the measured forest structures agreed well with the BRF based on the generated structures from the proposed approach with the root mean square error (RMSE) and relative RMSE (rRMSE) ranging from 0.002 to 0.006 and from 0.7% to 19.8%, respectively. Comparison of the computed BRF with corresponding MODIS reflectance data yielded RMSE and rRMSE lower than 0.03 and 20%, respectively. Although the results from the current study are limited in two boreal forest stands, our approach has the potential to generate stand structures for different forest biomes.     

Boreal forest coherence-based measures of interferometric pair suitability for operational stem volume retrieval

The performance of interferometric synthetic aperture radar (INSAR)-based boreal forest stem volume retrieval is strongly affected by weather conditions around the time of the SAR image acquisitions. Since weather conditions cannot be controlled, the suitability of a particular interferometric pair for stem volume retrieval can only be assessed afterward. In this letter, four objective measures based on observed forest coherence were compared in assessing the suitability of interferometric pairs for stem volume retrieval. These suitability measures can be used to identify the best and worst pairs, i.e., the ones with the most and least favorable weather conditions. Stem volume retrievals were performed using single European Remote Sensing (ERS-1/2) Tandem interferometric pairs by inverting a backscattering-coherence model for boreal forests. A total of 14 ERS Tandem image pairs acquired in varying weather conditions were studied, and the stem volume retrieval performance was assessed against ground-based stem volume estimates on 134 boreal forest stands. Stem volume retrieval performance as measured by R/sup 2/-values between INSAR-estimated stem volumes and ground truth was found to be directly proportional to boreal forest coherence. The interferometric coherence-contrast (ICC), i.e., the difference in coherence between sparsest and densest boreal forest stands was found to be the best of the four studied suitability measures. The ICC could be used as a suitability parameter in the selection of the best interferometric pairs for operational boreal forest stem volume retrieval.     

Developing a gross primary production model for coniferous forests of northeastern USA from MODIS data

Accurate estimation of ecosystem carbon fluxes is crucial for understanding the feedbacks between the terrestrial biosphere and the atmosphere and for making climate-policy decisions. A statistical model is developed to estimate the gross primary production (GPP) of coniferous forests of northeastern USA using remotely sensed (RS) radiation (land surface temperature and near-infra red albedo) and ecosystem variables (enhanced vegetation index and global vegetation moisture index) acquired by the Moderate Resolution Imaging Spectroradiometer (MODIS) sensor. This GPP model (called R-GPP-Coni), based only on remotely sensed data, was first calibrated with GPP estimates derived from the eddy covariance flux tower of the Howland forest main tower site and then successfully transferred and validated at three other coniferous sites: the Howland forest west tower site, Duke pine forest and North Carolina loblolly pine site, which demonstrate its transferability to other coniferous ecoregions of northeastern USA. The proposed model captured the seasonal dynamics of the observed 8-day GPP successfully by explaining 84–94% of the observed variations with a root mean squared error (RMSE) ranging from 1.10 to 1.64 g C/m²/day over the 4 study sites and outperformed the primary RS-based GPP algorithm of MODIS.     

Use of local and global maps of forest canopy height and aboveground biomass to enhance local estimates of biomass in miombo woodlands in Tanzania

Field surveys are often a primary source of aboveground biomass (AGB) data, but plot-based estimates of parameters related to AGB are often not sufficiently precise, particularly not in tropical countries. Remotely sensed data may complement field data and thus help to increase the precision of estimates and circumvent some of the problems with missing sample observations in inaccessible areas. Here, we report the results of a study conducted in a 15,867 km² area in the dry miombo woodlands of Tanzania, to quantify the contribution of existing canopy height and biomass maps to improving the precision of canopy height and AGB estimates locally. A local and a global height map and three global biomass maps, and a probability sample of 513 inventory plots were subject to analysis. Model-assisted sampling estimators were used to estimate mean height and AGB across the study area using the original maps and then with the maps calibrated with local inventory plots. Large systematic map errors – positive or negative – were found for all the maps, with systematic errors as great as 60–70 %. The maps contributed nothing or even negatively to the precision of mean height and mean AGB estimates. However, after being calibrated locally, the maps contributed substantially to increasing the precision of both mean height and mean AGB estimates, with relative efficiencies (variance of the field-based estimates relative to the variance of the map-assisted estimates) of 1.3–2.7 for the overall estimates. The study, although focused on a relatively small area of dry tropical forests, illustrates the potential strengths and weaknesses of existing global forest height and biomass maps based on remotely sensed data and universal prediction models. Our results suggest that the use of regional or local inventory data for calibration can substantially increase the precision of map-based estimates and their applications in assessing forest carbon stocks for emission reduction programs and policy and financial decisions.     

Evaluations and comparisons of rule-based and machine-learning-based methods to retrieve satellite-based vegetation phenology using MODIS and USA National Phenology Network data

Vegetation phenology is a sensitive indicator that reflects the vegetation–atmosphere interactions and vegetation processes under global atmospheric changes. Fast-developing remote sensing technologies that monitor the land surface at high spatial and temporal resolutions have been widely used in vegetation phenology retrieval and analysis at a large scale. While researchers have developed many phenology retrieving methods based on remote sensing data, the relationships and differences among the phenology retrieving methods are unclear, and there is a lack of evaluation and comparison with the field phenology recoding data. In this study, we evaluated and compared eight phenology retrieving methods using Moderate Resolution Imaging Spectroradiometer (MODIS) and the USA National Phenology Network data from across North America. The studied phenology retrieving methods included six commonly used rule-based methods (i.e., amplitude threshold, the first-order derivative, the second-order derivative, the third-order derivative, the relative change curvature, and the curvature change rate) and two newly developed machine learning methods (i.e., neural network and random forest). At the large scale, the start of the season (SOS) values, derived by all methods, had similar spatial distributions; however, the retrieved values had large uncertainties in each pixel, and the end of the season (EOS) inverted values were largely different among methods. At the site scale, the SOS and EOS values extracted by the rule-based methods all had significant positive correlations with the field phenology observations. Among the rule-based methods, the amplitude threshold method performed the best. The machine learning methods outperformed the rule-based methods in terms of retrieving the SOS when assessed using the field observations. Our study highlighted that there were large differences among the methods in retrieving the vegetation phenology from satellite data and that researchers must be cautious in selecting an appropriate method for analyzing the satellite-retrieved phenology. Our results also demonstrated the importance of field phenology observations and the usefulness of the machine learning methods in understanding the satellite-based land surface phenology. These findings provide a valuable reference for the future development of global and regional phenology products.     

Updating Landsat-based forest cover maps with MODIS images using multiscale spectral-spatial-temporal superresolution mapping

With the high deforestation rates of global forest covers during the past decades, there is an ever-increasing need to monitor forest covers at both fine spatial and temporal resolutions. Moderate Resolution Imaging Spectroradiometer (MODIS) and Landsat series images have been used commonly for satellite-derived forest cover mapping. However, the spatial resolution of MODIS images and the temporal resolution of Landsat images are too coarse to observe forest cover at both fine spatial and temporal resolutions. In this paper, a novel multiscale spectral-spatial-temporal superresolution mapping (MSSTSRM) approach is proposed to update Landsat-based forest maps by integrating current MODIS images with the previous forest maps generated from Landsat image. Both the 240 m MODIS bands and 480 m MODIS bands were used as inputs of the spectral energy function of the MSSTSRM model. The principle of maximal spatial dependence was used as the spatial energy function to make the updated forest map spatially smooth. The temporal energy function was based on a multiscale spatial-temporal dependence model, and considers the land cover changes between the previous and current time. The novel MSSTSRM model was able to update Landsat-based forest maps more accurately, in terms of both visual and quantitative evaluation, than traditional pixel-based classification and the latest sub-pixel based super-resolution mapping methods The results demonstrate the great efficiency and potential of MSSTSRM for updating fine temporal resolution Landsat-based forest maps using MODIS images.