Remote Sensing of Environment

Ninety percent of projected global urbanization will be concentrated in low income countries (United-Nations, 2004). This will have considerable environmental, economic and public health implications for those populations. Objective and efficient methods of delineating urban extent are a cross-sectoral need complicated by a diversity of urban definition rubrics world-wide. Large-area maps of urban extents are becoming increasingly available in the public domain, as are a wide-range of medium spatial resolution satellite imagery. Here we describe the extension of a methodology based on Landsat ETM and Radarsat imagery to the production of a human settlement map of Kenya. This map was then compared with five satellite imagery-derived, global maps of urban extent at Kenya national-level, against an expert opinion coverage for accuracy assessment. The results showed the map produced using medium spatial resolution satellite imagery was of comparable accuracy to the expert opinion coverage. The five global urban maps exhibited a range of inaccuracies, emphasising that care should be taken with use of these maps at national and sub-national scale.

Surface climatic conditions are key determinants of arthropod vector distribution and abundance and consequently affect transmission rates of any diseases they may carry. Remotely sensed observations by satellite sensors are the only feasible means of obtaining regional and continental scale measurements of climate at regular intervals for real-time epidemiological applications such as disease early warning systems. The potential of Pathfinder AVHRR Land (PAL) data to provide surrogate variables for near-surface air temperature and vapour pressure deficit (VPD) over Africa and Europe were assessed in this context. For the years 1988-1990 and 1992, correlations were examined between meteorological ground measurements (monthly mean air temperature and VPD(grd)) and variables derived from Advanced Very High Resolution Radiometer (AVHRR) data (LST and VPD(sat)). The AVHRR indices were derived from both daily and composite PAL data so that their relative performance could be determined. Furthermore, the ground observations were divided into African and European subsets, so that the relative performance of the satellite data at tropical/sub-tropical and temperate latitudes could be assessed.Significant correlations were shown between air temperature and LST in all months. Temporal variability existed in the strength of correlations throughout any twelve-month period, with the pattern of variability consistent between years. The adjusted r(2) values increased when elevation and the Normalised Difference Vegetation Index (NDVI) were included, in addition to LST, as predictor variables of air temperature. Attempts to derive monthly estimates of atmospheric moisture availability resulted in an over-estimation of VPD(sat) compared to ground observations, VPD(grd). The use of daily PAL data to derive monthly mean climatic indices was shown to be more accurate than those obtained using monthly maximum values from 10-day composite data. A subset of the 1992 data was then used to build linear regression models for the direct retrieval of monthly mean air temperature from PAL data. The accuracy of retrieved estimates was greatest when NDVI was included with LST as predictor variables, with root mean square errors varying from 1.83°C to 3.18 °C with a mean of 2.38 °C over the twelve months.

The Sentinel-1 will carry onboard a C-band radar instrument that will map the European continent once every four days and the global land surface at least once every twelve days with finest 5×20m spatial resolution. The high temporal sampling rate and operational configuration make Sentinel-1 of interest for operational soil moisture monitoring. Currently, updated soil moisture data are made available at 1km spatial resolution as a demonstration service using Global Mode (GM) measurements from the Advanced Synthetic Aperture Radar (ASAR) onboard ENVISAT. The service demonstrates the potential of the C-band observations to monitor variations in soil moisture. Importantly, a retrieval error estimate is also available; these are needed to assimilate observations into models. The retrieval error is estimated by propagating sensor errors through the retrieval model.

Monitoring freeze-thaw transitions of the soil and vegetation in high latitude terrestrial ecosystems is useful for determining the length of the growing season, and for monitoring potential damage to living plants due to freezing and frost drought. The authors present a technique for monitoring freeze-thaw cycles using repeat-pass SAR data from the European Remote Sensing Satellite, ERS-1, 100 km in swath width, and mosaicked together along a north-south Alaskan transect 1400 km in length. Freezing of the soil and vegetation is detected based on a 3 dB decrease in radar cross-section σ⁰ relative to a known thawed state of the landscape. The decrease in σ⁰ is explained by radar backscatter models as resulting from a large decrease of the dielectric constant of the soil and vegetation with freezing. The technique is validated using air-temperature recordings from three forest stands along the Tanana River near Manley Hot Springs, and from local weather stations along the transect. The technique does not apply to areas of standing water, but is independent of the type of vegetation cover, and permits the spatial and temporal monitoring of freezing of the natural landscape at the regional scale

The authors measured the vertical pattern of the Shuttle Imaging Radar C (SIR-C) antenna using images of the Amazon rain forest, the largest flat and homogeneous forest in the world. Its scattering coefficient is nearly independent of the incidence angle. The images produced by JPL use the standard SIR-C CEOS format. The authors must use images without radiometric correction to get the antenna pattern, as the radiometric correction depends on preflight patterns. They can only give a small sample of the patterns for the SIR-C antenna in this paper, as the number of Amazon passes was limited, and only a few of the many modes were available. For both L-band and C-band all possible linear polarizations could be used. In addition, electronic beam steering up to ±20° from the boresight direction could modify the patterns. Also nine “beam spoiling” modes could change the patterns. The authors previously reported on the SIR-B and X-SAR patterns obtained by the same method. If all modes had been tested over the Amazon or some other suitable homogenous area, we would have been able to give details of all patterns. Unfortunately the number of passes over suitable areas was very limited, so the authors can only report on those passes

For radar incidence angles ranging from 20 to 60°, modeled canopy volume scattering from a flooded forest in the Amazon dominated C-HH, C-VV, and L-VV backscatter. Between 20 and 35° incidence angles and at L-HH polarization, modeled canopy-ground term and canopy volume scattering contributed roughly equally to the total backscatter, and there was some contribution from the trunk-ground term. Canopy-ground (at small incidence angles) and trunk-ground terms (at large incidence angles) were dominant at P-HH and P-VV backscatter. The dominant canopy volume scattering at C-HH, C-VV, and L-VV showed that radar systems with these parameters could not penetrate the canopies in the stands. When trunk-ground and canopy-ground terms dominated or contributed significantly to the total backscatter, microwave energy penetrated the canopies. Thus, a L-HH, P-HH, or P-VV radar system could penetrate the canopies, and, potentially, detect boundaries between flooded and nonflooded forests.

Phenological differences among broadly defined vegetation types can be a basis for global scale classification at a very coarse spatial scale. Using the annual sequence of composited Normalized Vegetation Index (NDVI) values in an AVHRR data set, DeFries and Townshend (1994) distinguished eleven global cover types and classified global land cover with a maximum likelihood classifier. The present research presents a neural network architecture called fuzzy ARTMAP to classify the same data set. Classification results are analyzed and compared with those obtained from DeFries and Townshend (1994). First, classification accuracy on the training data set is 100% compared with the corresponding accuracy of 86.8%. Second, when fuzzy ARTMAP is trained using 80% of data set and tested on the remaining (unseen) 20% data set, classification accuracy is more than 80%. Third, classification results vary with and without latitude as an input variable similar to those of DeFries and Townshend

An exploratory study on the the spatial variance of the bidirectional reflectance over discontinuous plant canopies indicates that the spatial variance of the bidirectional reflectance calculated from ASAS images has peak values at the hotspot and near nadir. This behavior can be explained by the geometric effect of discontinuous tree crowns and the regularization effect. Validation of Jupp and Woodcock's two component geometric optical (GO) model shows that it captures the basic features of the spatial variance of the bidirectional reflectance over discontinuous plant canopies. Their two-component GO model is modified to account for the spatial interactions of four scene components. Validation shows that the modified GO model improves the model prediction. This exploratory study helps lay the foundation for future retrieval of surface information from remote sensing satellite data

Interferometric coherence maps from 1-day repeat ERS and 44, 88 and 132-day repeat JERS data are derived and compared to biomass density estimates from a tropical forest area in Brazil. Results show that phase coherence adds significantly to the information available from these data

A procedure for determination of conifer needle spectral optical properties (transmittance, T_λ; reflectance, R_λ; λ=wavelength) was developed to support field measurements acquired in the Boreal Ecosystem-Atmosphere Study (BOREAS). This was a revision of a protocol, C.S. Daughtry et al. (1989), which uses an indirect and labor-intensive step (involving painting of needles) to estimate the inter-needle light transmittance gaps, or gap fraction (GF), in each sample. The present authors procedure uses a direct image capture method to calculate GF, enabling measurements on both dorsal and ventral surfaces of all samples, and is 3-4 times faster than. However, for either method T_λ underestimates often result, including negative T_λ in the visible (VIS) spectrum, especially when GF is large (>25%). The authors performed controlled experiments to evaluate the general effect of GF, and errors in GF estimation, on the calculated T_λ and R_λ spectra. T_λ was found to be inversely related to GF, with larger coefficients associated with VIS than near-infrared (NIR) λ. Consequently, GF overestimates also yielded T_λ underestimates. Using these results, they developed a correction algorithm for their BOREAS measurements. “Corrected” estimates of the fraction of absorbed photosynthetically active radiation (APAR) were ~80-84% for jack pine and 75-78% for black spruce. Correction reduced the Simple Ratio (SR=VIS/NIR) by 40-60% in most cases. SR calculated from corrected T_λ spectra were significantly higher than those determined from R_λ spectra

In order to determine the significance of organic matter content on the microwave emissivity of soils when estimating soil moisture, a series of field experiments were conducted in which 1.4 GHz microwave emissivity data were collected over test plots of sandy loam soil with different organic matter levels (1.8%, 4.0%, and 6.1%) for a range of soil moisture values. Analyses of the observed data showed only minor variation in microwave emissivity due to a change in organic matter content at a given moisture level for soils with similar texture and structure. Predictions of microwave emissivity made using a dielectric model for aggregated soils exhibited the same trends and type of response as the measured data when adjusted values for the input parameters were utilized.

The ESA Sentinels will be the first series of operational satellites to meet the Earth observation needs of the European Union-ESA Global Monitoring for Environment and Security (GMES) programme. The pair of Sentinel-2 satellites will routinely provide high-resolution (10-20 m) optical images globally with frequent revisits tailored to the needs of GMES land and emergency services. Sentinel-2 aims at ensuring continuity of Spot- and Landsat-type data, with improvements to allow service evolution. The first launch is expected in 2012.

Not Available

This article describes the error sources for three in-flight sensor calibration methods used by the Remote Sensing Group of the Optical Sciences Center at the University of Arizona. The three methods are the reflectance-, improved reflectance-, and radiance-based methods, which all reference the earth-atmosphere system. The sources of error or uncertainty for each method are discussed, and an estimate of the percent uncertainty associated with each source is made for conditions similar to those actually used for calibrations at White Sands, New Mexico. The results of in-flight calibrations are compared to those of the on-board lamp calibration system for a SPOT HRV camera.

Near-infrared analysis (NIRA) methodology was applied to the reflectance spectra of arid and semiarid soils in the visible and near infrared (VIS-NIR; 0.4–1.1 μm) spectral region. The method is termed visible and near-infrared analysis (VNIRA). Although the spectra of the soils were characterized as monotonous and featureless, the methodology has yielded a prediction equation for estimating several soil constituents from the reflectance curves. The constituents were: CaCO3, Fe2O3, Al2O3, SiO2, LOI (lost-on-ignition), Fed (free iron oxides), and K2O. Several mathematic manipulations were applied to the raw data in order to derive the optimal prediction equation. Spectral compression into 6, 8, 15, 71, and 350 spectral bands and a spectral derivation technique were applied to four separate soil groups, which were selected on the basis of their chemical characteristics. The wavelengths selected by the method for the optimal prediction equation were assigned to constituents that consisted of spectral features within the VIS-NIR region. An intercorrelation between spectrally featureless constituents and constituents with special features was found to be the major mechanism by which to predict constituents that had no spectral features within this part of the spectrum. It was shown that low spectral resolution is not necessarily a limiting factor in obtaining quantitative information about the chemistry of soil samples. All the examined soil constituents except Fed (which needed 700 spectral bands) required between 15 and 350 spectral bands for optimal prediction. It was concluded that the VIS-NIR (0.4–1.1 μm) is a suitable spectral region for obtaining quantitative information about soil chemistry. Although the VNIRA performance is not as precise as the chemical performance, the precision obtained is likely to be useful for rapid soil characterization and remote-sensing applications. We strongly recommend the use of both the VNIRA and the NIRA methods to better interpret high resolution remote-sensing data.

Vegetation water content retrieval using passive remote sensing techniques in the 0.4–2.5 μm region (reflection of solar radiation) and the 8–14 μm region (emission of thermal radiation) has given rise to an abundant literature. The wavelength range in between, where the main water absorption bands are located, has surprisingly received very little attention because of the complexity of the radiometric signal that mixes both reflected and emitted fluxes. Nevertheless, it is now covered by the latest generation of passive optical sensors (e.g. SEBASS, AHS). This work aims at modeling leaf spectral reflectance and transmittance in the infrared, particularly between 3 μm and 5 μm, to improve the retrieval of vegetation water content using hyperspectral data. Two unique datasets containing 32 leaf samples each were acquired in 2008 at the USGS National Center, Reston (VA, USA) and the ONERA Research Center, Toulouse (France). Reflectance and transmittance were recorded using laboratory spectrometers in the spectral region from 0.4 μm to 14 μm, and the leaf water and dry matter contents were determined. It turns out that these spectra are strongly linked to water content up to 5.7 μm. This dependence is much weaker further into the infrared, where spectral features seem to be mainly associated with the biochemical composition of the leaf surface. The measurements show that leaves transmit light in this wavelength domain and that the transmittance of dry samples can reach 0.35 of incoming light around 5 μm, and 0.05 around 11 μm. This work extends the PROSPECT leaf optical properties model by taking into account the high absorption levels of leaf constituents (by the insertion of the complex Fresnel coefficients) and surface phenomena (by the addition of a top layer). The new model, PROSPECT-VISIR (VISible to InfraRed), simulates leaf reflectance and transmittance between 0.4 μm and 5.7 μm (at 1 nm spectral resolution) with a root mean square error (RMSE) of 0.017 and 0.018, respectively. Model inversion also allows the prediction of water (RMSE = 0.0011 g/cm²) and dry matter (RMSE = 0.0013 g/cm²) contents.

The MOMS-02 instrument is designed as a “pushbroom” scanning device for spectral data acquisition in the visible (VIS) and near infrared (NIR) range (Band 1: 449–511 nm, Band 2: 532–577 nm, Band 3: 645–677 nm, Band 4: 772–815 nm, IFOV: 13.5 m × 13.5 m), combined with along-track, simultaneously acquired forward, backward Bands 6/7: 524–763 nm (IFOV: 13.5 m × 13.5 m) and nadir Band 5: 512–765 nm (IFOV: 4.5 × 4.5 m) looking stereo capability. The location of the MS bands were done with respect to the suitibility of these spectral regions to analyze vegetation type and status. The calculations of the spectral significance for thematic data evaluation is based on spectroscopic field measurements of crops, balanced with an atmospheric transmission model (Lowtran 7), and the system parameters. The simulation results lead to the conclusion that the spatial, spectral, and radiometric resolution, combined with the on-track stereo capability of the MOMS-02 sensor, will result in substantial improvements in vegetation classification compared to concurrent remote sensing instruments like Landsat TM and Spot HRV.

COSMOS (Campaign for validating the Operation of Soil Moisture and Ocean Salinity), and NAFE (National Airborne Field Experiment) were two airborne campaigns held in the Goulburn River catchment (Australia) at the end of 2005. These airborne measurements are being used as benchmark data sets for validating the SMOS (Soil Moisture and Ocean Salinity) ground segment processor over prairies and crops. This paper presents results of soil moisture inversions and brightness temperature simulations at different resolutions from dual-polarisation and multi-angular L-band (1.4 GHz) measurements obtained from two independent radiometers. The aim of the paper is to provide a method that could overcome the limitations of unknown surface roughness for soil moisture retrievals from L-band data. For that purpose, a two-step approach is proposed for areas with low to moderate vegetation. Firstly, a two-parameter inversion of surface roughness and optical depth is used to obtain a roughness correction dependent on land use only. This step is conducted over small areas with known soil moisture. Such roughness correction is then used in the second step, where soil moisture and optical depth are retrieved over larger areas including mixed pixels. This approach produces soil moisture retrievals with root mean square errors between 0.034 m3 m− 3 and 0.054 m3 m− 3 over crops, prairies, and mixtures of these two land uses at different resolutions.

Results of radiometric measurements over bare and vegetated fields with dual-polarized microwave radiometers at 1.4-GHz and 5-GHz frequencies are presented. The measured brightness temperatures over bare fields are shown to compare favorably with those calculated from radiative transfer theory with two constant parameters characterizing surface roughness effect. The presence of vegetation cover is found to reduce the sensitivity to soil moisture variation. This sensitivity reduction is generally more pronounced the denser the vegetation cover and the higher the frequency of observation. The effect of vegetation cover is also examined with respect to the measured polarization factor at both frequencies. With the exception of dry corn fields, the measured polarization factor over vegetated fields is found appreciably reduced compared to that over bare fields. A much larger reduction in this factor is found at 5 GHz than at 1.4 GHz.

Ecosystem energy has been shown to be a strong correlate with biological diversity at continental scales. Early efforts to characterize this association used the normalized difference vegetation index (NDVI) to represent ecosystem energy. While this spectral vegetation index covaries with measures of ecosystem energy such as net primary production, the covariation is known to degrade in areas of very low vegetation or in areas of dense forest. Two of the new vegetation products from the MODIS sensor, derived by integrating spectral reflectance, climate data, and land cover, are thought to better approximate primary productivity than NDVI. In this study, we determine if the new MODIS derived measures of primary production, gross primary productivity (GPP) and net primary productivity (NPP) better explain variation in bird richness than historically used NDVI. Moreover, we evaluate if the two productivity measures covary more strongly with bird diversity in those vegetation conditions where limitations of NDVI are well recognized.

Four seasons (2004–2007) of snow surveys across the boreal forest of northern Manitoba were utilized to determine relationships between vertically polarized Advanced Microwave Scanning Radiometer (AMSR-E) brightness temperatures (TB) and ground measurements of snow water equivalent (SWE). Regression analysis identified moderate strength, yet statistically significant relationships between SWE and TB differences (36.5–18.7; 36.5–10.7; 18.7–10.7) for individual seasons. When multiple seasons were considered collectively, however, the 36.5–18.7 and 36.5–10.7 differences were insignificant because the seasonal linear relationships shifted from year to year over the same TB range regardless of SWE. This inter-seasonal consistency in TB was explained through significant correlations with vegetation density as characterized by a MODIS-derived forest transmissivity dataset. More encouraging results were found for the 18.7–10.7 difference: the relationship with SWE remained statistically significant when multiple years were considered together, and the 18.7–10.7 difference was not significantly associated with vegetation density. Additional snow survey data from the Northwest Territories (2005–2007) were used to verify the 18.7–10.7 relationship with SWE across the northern boreal forest. These results suggest use of the 18.7–10.7 TB difference, rather than the traditional 36.5–18.7 TB difference, is necessary to capture inter-seasonal SWE variability across forested regions.

Using 180° field-of-view (full-sky) imaging polarimetry, we measured the spatiotemporal change of the polarization pattern of the entire celestial hemisphere during the total solar eclipse of 11 August 1999 in Kecel, Hungary. We compared these patterns with the normal celestial polarization patterns measured at the same times on the subsequent day of the eclipse. As a second control sky, the celestial polarization pattern measured on 26 August 1999 in Tunisia was chosen with the same solar zenith distance as that at the Hungarian eclipse. We computed the corresponding theoretical celestial polarization patterns on the basis of the single-scattering Rayleigh model. The spectral characteristics of the polarization pattern in the sky during totality were also measured in the red (650 nm), green (550 nm), and blue (450 nm) ranges of the spectrum. A qualitative explanation was given for the origin of the angle of polarization (E-vector) pattern and the neutral point of skylight polarization near the zenith observed during totality. The relation of our results to earlier observations on skylight polarization during total eclipses was analyzed. The agreements with previous eclipse observations were discussed. The reasons for some disagreements with previous eclipse observations were explained in connection with the spectral dependence of skylight polarization and the fine structure of the celestial E-vector pattern during totality.

This article extends the calibration method of Mitchell, O'Brien, and Forgan (1992) to the calibration of the AVHRR on NOAA 11 using a data set obtained over the period January to May 1991. The method correlates AVHRR response against radiances computed from a radiative transfer model whose parameters are constrained by surface measurements. The data consist of cloud free A VHRR images over the ocean west of Tasmania and simultaneous sun photometer measurements of aerosol optical depth at Cape Grim, some 50 km from the ocean target. Ancilliary surface meteorological data are also employed. The best estimate of the mean gain in Channel 1 over the first half of 1991 is 1.75 ± 0.09 counts per W m−2 μm−1 sr−1, in good agreement with several other determinations using a variety of techniques. However, differences among results from the various methods suggests that ultimate accuracy in the determination of sensor gain is limited to ∼ ± 5%. Spurious results are obtained when the aerosol depth measured at Cape Grim does not characterize the airmass over the ocean target. This situation can arise when winds are light and variable in direction during the observing period. Under these circumstances it is not valid to use the sun photometer aerosol optical depths to model the upwelling radiance over the ocean. In addition, uncertainty in wind speed over the target site introduces significant error into the method. The technique is found to be unreliable in Channel 2 due to problems connected with water vapour absorption.

The radiative temperature difference between 3.7 and 11 μm (ΔAT) is investigated as a possible method for tracking dust outbreaks. Theoretical calculations indicate that the technique would be most sensitive to dust loading during the day when the day when ΔAT's are enhanced by reflection of 3.7 μm solar radiation. The feasibility of tracking dust outbreaks is demonstrated by comparing satellite observations with surface estimates of visibility. Theoretical calculations also demonstrate the potential of inferring the dust layer optical depth from these spectral measurements.

This paper examines the relationship between the microwave backscattering coefficient of a vegetation canopy, σcan0, and the canopy's leaf area index (LAI). The relationship is established through the development of one model for corn and sorghum and another for wheat. Both models are extensions of the cloud model of Attema and Ulaby (1978). Analysis of experimental data measured at 8.6, 13.0, 17.0, and 35.6 GHz indicates that most of the temporal variations of σcan0 can be accounted for through variations in green LAI alone, if the latter is greater than 0.5.

High spatial and spectral resolution thermal infrared imagery (8.0–13.5 μm) from the SEBASS airborne sensor was used to analyze and map tree canopy spectral features at the State Arboretum of Virginia, near Boyce, Virginia. Fifty tree species were analyzed and about half were directly identified with varying degrees of success on the basis of spectral matched filtering that utilized laboratory-measured leaf spectra as the target signatures. Spectral averages of pixels extracted from SEBASS emissivity data compared favorably with laboratory spectra of leaves collected from individual tree species. Best results were obtained from species having relatively strong spectral contrast, wide and flat leaves, closed planophile canopies, and/or large canopy areas. Tree species having small leaves or unfavorable leaf orientations showed spectral attenuation likely resulting from cavity blackbody effects. Increased spatial resolution and better image calibration and atmospheric correction might lead to further improvements in thermal infrared plant species identification.

The directional emissivity of snow and ice surfaces in the 8–14 μm thermal infrared (TIR) atmospheric window was determined from spectral radiances obtained by field measurements using a portable Fourier transform infrared spectrometer in conjunction with snow pit work. The dependence of the directional emissivity on the surface snow type (grain size and shape) was examined. We obtained emissivity spectra for five different surface types, i.e., fine dendrite snow, medium granular snow, coarse grain snow, welded sun crust snow, and smooth bare ice. The derived emissivities show a distinct spectral contrast at wavelengths λ = 10.5–12.5 μm which is enhanced with increasing the snow grain size. For example, emissivities at both 10.5 μm and 12.5 μm for the nadir angle were 0.997 and 0.984 for the fine dendrite snow, 0.996 and 0.974 for the medium granular snow, 0.995 and 0.971 for the coarse grain snow, 0.992 and 0.968 for the sun crust, and 0.993 and 0.949 for the bare ice, respectively. In addition, the spectral contrast exhibits a strong angular dependence, particularly for the coarser snow and bare ice, e.g., the emissivity at λ = 12.5 μm for the off-nadir angle of 75° reaches down to 0.927, 0.896, and 0.709 for the coarse grain snow, sun crust, and bare ice cases, respectively. The angular dependent emissivity spectra of the bare ice were quite consistent with the spectra predicted by the Fresnel reflectance theory. The observed results firmly demonstrate that the directional emissivity of snow in the TIR can vary depending upon the surface snow type. The high variability of the spectral emissivity of snow also suggests the possibility to discriminate between snow and ice types from space using the brightness temperature difference in the atmospheric window.

The two variants of the box method (with one and two lids) has been rigorously analyzed. As a result of this analysis, a correction factor that takes into account the nonideality of the materials used for the box, as well as its geometry, has been derived A simple method for determining the effective downward atmospheric temperature that uses only the temperature measurement at the zenith also has been proposed. Finally, by using one of the two variants of the box method, 72 in situ emissivity measurements in the 8–14 pin wave-band region of typical vegetation, soils, and rocks of Europe and South America, has been obtained. The use of these data for the ernissivity correction of satellite thermal measurements has been analyzed, and emissivities for the Landsat Thematic Mapper band 6 and NOAA Advanced Very High Resolution Radiometer channels 4 and 5 have been derived.

There is growing interest in the mid-infrared spectral region (8–14 μm) as both a laboratory and a remote sensing tool in geology, because this portion of the spectrum contains the characteristic, fundamental, molecular vibration bands for silicates and other mineral groups. However, it is necessary to understand the relationship between the spectra of mineral mixtures and those of individual minerals in the mixture in order to completely interpret and predict mineral abundances from infrared data. Results of this study show quantitatively for the first time that the spectra of particulate mixtures of silicate minerals in this wavelength region combine linearly by volume within a very small error, as long as particles are much larger than the wavelength so that volume scattering is insignificant compared to surface scattering. Results here apply specifically to mineral samples in the 75–250 μm size range. They imply that we can predict the spectral response of a rock if the constituent minerals and their abundances are known. More importantly, our results indicate that the relative quantities of minerals in simple mixtures can be predicted to within 12% in the worst case, and more typically to within 5%. Consequently, geologists should be able to unmix the composite spectra of rocks to determine mineral abundances. This is important for both laboratory rock identification and remote sensing applications. By better understanding how component mineral spectra mix in the spectrum of a rock, we can also better choose spectral band positions and resolutions in infrared remote sensing for compositional identification.

Recent day/night land surface temperature (LST) algorithms can recover both emissivity and temperature but require some assumptions about the relative optical properties of natural materials in the thermal infrared. We constructed a goniometer and measured the spectral, angular emissivity, and bidirectional reflectance of sands and soils in the 3–14 μm range. In this paper, we present the results for five diverse surfaces and examine the validity of the LST assumptions. We conclude that the change in emissivity with angle was small across the entire range for all of the materials except sand. In addition, for all of the materials, the measured variation in bidirectional reflectance anisotropy with wavelength was small enough to neglect.

Mid-infrared reflectance spectra were measured (using an interferometer spectrometer with a 4/cm resolution) of a number of mineral mixtures (including quartz-microcline, quartz-hornblende, microcline-albite, microcline-hornblende, albite-hornblende, and albite-augite mixtures containing different proportions of the components, each in the 75-250 micron size range), and the spectra were compared with those of single components. It is shown that the mid-infrared reflectance spectra of particulate mixtures of these minerals combine linearly by volume within a very small error, as long as particles are much larger than the wavelength.

We developed a new 6-year daily, daytime and nighttime, NOAA-14 AVHRR based land surface temperature (LST) dataset over continental Africa for the period 1995 through 2000. The processing chain was developed within the Global Inventory Modeling and Mapping System (GIMMS) at NASA's Goddard Space Flight Center. This paper describes the processing methodology used to convert the Global Area Coverage Level-1b data into LST and collateral data layers, such as sun and view geometries, cloud mask, local time of observation, and latitude and longitude. We used the Ulivieri et al. [Ulivieri, C., M.M. Castronuovo, R. Francioni, and A. Cardillo (1994), A split window algorithm for estimating land surface temperature from satellites, Adv. Space Research, 14(3):59–65.] split window algorithm to determine LST values. This algorithm requires as input values of surface emissivity in AVHRR channels 4 and 5. Thus, we developed continental maps of emissivity using an ensemble approach that combines laboratory emissivity spectra, MODIS-derived maps of herbaceous and woody fractional cover, and the UNESCO FAO soil map. A preliminary evaluation of the resulting LST product over a savanna woodland in South Africa showed a bias of < 0.3 K and an uncertainty of < 1.3 K for daytime retrievals (< 2.5 K for night). More extensive validation is required before statistically significant uncertainties can be determined. The LST production chain described here could be adapted for any wide field of view sensor (e.g., MODIS, VIIRS), and the LST product may be suitable for monitoring spatial and temporal temperature trends, or as input to many process models (e.g., hydrological, ecosystem).

At the behest of NASA's Mission to Planet Earth, the National Research Council recently conducted a review on the current status and future directions for earth science information provided by spaceborne synthetic aperture radars. As part of this process, a panel of 16 scientists met to review the utility of SAR for monitoring ecosystem processes. The consensus of this ecology panel was that the demonstrated capabilities of imaging radars for investigating terrestrial ecosystems could best be organized into four broad categories: 1) classification and detection of change in land cover; 2) estimation of woody plant biomass; 3) monitoring the extent and timing of inundation; and 4) monitoring other temporally-dynamic processes. The major conclusions from this panel were: 1) Multichannel radar data provide a means to classify land-cover patterns because of its sensitivity to variations in vegetation structure and vegetation and ground-layer moisture. The relative utility of data from imaging radars versus multispectral scanner data has yet to be determined in a rigorous fashion over a wide range of biomes for this application. 2) Imaging radars having the capability to monitor variations in biomass in forested ecosystems. This capability is not consistent among different forest types. The upper levels of sensitivity for L-band and C-band systems such as SIR-C range between <100 t ha−1 for complex tropical forest canopies to ∼250 t ha−1 for simpler forests dominated by a single tree species. Best performance for biomass estimation is achieved using lower frequency (P- and L-band) radar systems with a cross-polarized (HV or VH) channel. 3) Like-polarized imaging radars (HH or VV) are well suited for detection of flooding under vegetation canopies. Lower frequency radars (P- and L-band) are most optimal for detecting flooding under forests, whereas higher frequency radars (C-band) work best for wetlands dominated by herbaceous vegetation. 4) It has been shown that spaceborne radars that have been in continuous operation for several years [such as the C-band (VV) ERS-1 SAR] provide information on temporally dynamic processes, such as monitoring a) variations in flooding in nonwooded wetlands, b) changes in the frozen/thawed status of vegetation, and c) relative variations in soil moisture in areas with low amounts of vegetation cover. These observations have been shown to be particularly important in studying ecosystems in high northern latitudes.

Vegetation indices, including the simple ratio (SR) and the normalized difference vegetation index (NDVI), from Landsat TM data were correlated to ground-based measurements of LAI, effective LAI, and the crown closure in boreal conifer forests located near Candle Lake and Prince Albert, Saskatchewan and near Thompson, Manitoba, as part of the Boreal Ecosystem-Atmosphere Study (BOREAS). The measurements were made using two optical instruments: the Plant Canopy Analyzer (LAI-2000, LI-COR) and the TRAC (Tracing Radiation and Architecture of Canopies). The TRAC was recently developed to quantify the effect of canopy architecture on optical measurements of leaf area index. The stands were located on georeferenced Landsat TM images using global positioning system (GPS) measurements. It is found that late spring Landsat images are superior to summer images for determining overstory LAI in boreal conifer stands because the effect of the understory is minimized in the spring before the full growth of the understory and moss cover. The effective LAI, obtained from gap fraction measurements assuming a random distribution of foliage spatial positions, was found to be better correlated to SR and NDVI than LAI. The effective LAI is less variable and easier to measure than LAI, and is also an intrinsic attribute of plant canopies. It is therefore suggested to use effective LAI as the most important parameter for radiation interception considerations.

Rainfall estimation from passive microwave satellite data has been used widely over oceans but has been less successful over land. This is because over land surfaces, the high spatial and temporal variability in emissivity, coupled with relatively low contrast between surface and rain cloud microwave emissions, make the rainfall signal more difficult to extract. The variability of emissivity is mainly due to variations in vegetation cover and soil moisture. Major improvements in reliability of rainfall estimates are possible if emissivities could be measured routinely at appropriate scales. The possibility of estimating emissivity at the frequencies relevant to rainfall from vegetation and soil moisture measurements is explored in this paper using data from airborne sensors over a semiarid area in Spain. Results show a good correlation between vegetation cover (represented by Normalized Difference Vegetation Index) and emissivity in dry conditions. This relationship is not significantly affected by vegetation type. Under wet conditions, the correlation is greatly reduced possibly due to the difficulty in accounting for cloud effects at higher frequencies. Attempts to quantify the effect of soil moisture using the Antecedent Precipitation Index were partially successful but more accurate measurements would be needed for reliable retrieval of emissivities. The use of a soil-adjusted vegetation index produced a higher correlation with emissivity than did the nonadjusted Normalized Difference Vegetation Index.

The accuracy of the Moderate Resolution Imaging Spectroradiometer (MODIS) 16-day albedo product (MOD43) is assessed using ground-based albedo observations from automatic weather stations (AWS) over spatially homogeneous snow and semihomogeneous ice-covered surfaces on the Greenland ice sheet. Data from 16 AWS locations, spanning the years 2000–2003, were used for this assessment. In situ reflected shortwave data were corrected for a systematic positive spectral sensitivity bias of between 0.01 and 0.09 on a site-by-site basis using precise optical black radiometer data. Results indicate that the MOD43 albedo product retrieves snow albedo with an average root mean square error (RMSE) of ±0.07 as compared to the station measurements, which have ±0.035 RMSE uncertainty. If we eliminate all satellite retrievals that rely on the backup algorithm and consider only the highest quality results from the primary bidirectional reflectance distribution function (BRDF) algorithm, the MODIS albedo RMSE is ±0.04, slightly larger than the in situ measurement uncertainty. There is general agreement between MODIS and in situ observations for albedo <0.7, while near the upper limit, a −0.05 MODIS albedo bias is evident from the scatter of the 16-site composite.

This work extends the previous study of Trishchenko et al. [Trishchenko, A. P., Cihlar, J., & Li, Z. (2002). Effects of spectral response function on surface reflectance and NDVI measured with moderate resolution satellite sensors. Remote Sensing of Environment 81 (1), 1–18] that analyzed the spectral response function (SRF) effect for the Advanced Very High Resolution Radiometer (AVHRR) onboard the NOAA satellites NOAA-6 to NOAA-16 as well as the Moderate Resolution Imaging Spectroradiometer (MODIS), the VEGETATION sensor (VGT) and the Global Imager (GLI). The developed approach is now applied to cover three new AVHRR sensors launched in recent years on NOAA-17, 18, and METOP-A platforms. As in the previous study, the results are provided relative to the reference sensor AVHRR NOAA-9. The differences in reflectance among these three radiometers relative to the AVHRR NOAA-9 are similar to each other and range from − 0.015 to 0.015 (− 20% to + 2% relative) for visible (red) channel, and from − 0.03 to 0.02 (− 5% to 5%) for the near infrared (NIR) channel. The absolute change in the Normalized Difference Vegetation Index (NDVI) ranged from − 0.03 to + 0.06. Due to systematic biases of the visible channels toward smaller values and the NIR channels toward slightly larger values, the overall systematic biases for NDVI are positive. The polynomial approximations are provided for the bulk spectral correction with respect to the AVHRR NOAA-9 for consistency with previous study. Analysis was also conducted for the SRF effect only among the AVHRR-3 type of radiometer on NOAA-15, 16, 17, 18 and METOP-A using AVHRR NOAA-18 as a reference. The results show more consistency between sensors with typical correction being under 5% (or 0.01 in absolute values). The AVHRR METOP-A reveals the most different behavior among the AVHRR-3 group with generally positive bias for visible channel (up to + 5%, relative), slightly negative bias for the NIR channel (1%–2% relative), and negative NDVI bias (− 0.02 to + 0.005). Polynomial corrections are also suggested for normalization of AVHRR on NOAA-15, 16, 17 and METOP-A to AVHRR NOAA-18.

The Meteorological Service of Canada (MSC) has developed an operational snow water equivalent (SWE) retrieval algorithm suite for western Canada that can be applied to both Scanning Multichannel Microwave Radiometer (SMMR) and Special Sensor Microwave/Imager (SSM/I) data. Separate algorithms derive SWE for open environments, deciduous, coniferous, and sparse forest cover. A final SWE value represents the area-weighted average based on the proportional land cover within each pixel. The combined SSM/I and SMMR time series of dual polarized, multichannel, spaceborne passive microwave brightness temperatures extends back to 1978, providing a lengthy time series for algorithm assessment. In this study, 5-day average (pentad) passive microwave-derived SWE imagery for 18 winter seasons (December, January, February 1978/79 through 1995/96) was compared to SWE estimates taken from a distributed network of surface measurements throughout western Canada.Results indicated both vegetative and snowpack controls on the performance of MSC algorithms. In regions of open and low-density forest cover, the in situ and passive microwave SWE data exhibited both strong agreement and similar levels of interannual variability. In locations where winter season SWE typically exceeded 75 mm, and/or dense vegetative cover was present, dataset agreement weakened appreciably, with little interannual variability in the passive microwave SWE retrievals. These results have important implications for extending the SWE monitoring capability of the MSC algorithm suite to northern regions such as the Mackenzie River basin.

In situ hyperspectral data measured above sunlit and shaded sides of canopies using a high spectral resolution radiometer were analyzed for identification of six conifer tree species. An artificial neural network algorithm was assessed for the identification purpose. Linear discrimination analysis was compared with the neural network algorithm. The hyperspectral with the neural data were further processed to smoothed reflectance and first derivative spectra and were separately used in tree species identification. Tree species recognition with data collected front six study sites was tested in seven experiments. The average accuracy of species recognition was obtained at every site. The overall performance of the neural network algorithm was better than that of linear discriminant analysis for species recognition when the same number of training samples and test samples were used. The discriminant analysis produced better accuracy than neural network at one site where many samples (10) were taken from six individual trees. Use of the average spectra of all samples for a particular tree species in training may not result in higher accuracy than use of individual spectral samples in training. Use of sunlit samples alone resulted in an overall accuracy of greater than 91%. The effects of site background including illuminating conditions on tree species specra were large. Neural networks are sensitive to subtle spectral details and can be trained to separate samples front the same species at different sites. Our experiments indicate that the discriminating power of visible bands is stronger than that of near-infrared bands. Higher recognition accuracies can be obtained in the blue to green or the red-edge spectral region as compared with four other spectral regions. A smaller set of selected bands can generate more accurate identification than all spectral bands.

Surface reflectivities for horizontally and vertically polarized emission at 19 GHz and 37 GHz are calculated for two locations over sand deserts, one location each over a rainforest, and a savanna from daytime observations by the Special Sensor Microwave Imager (SSM/I) on board the DMSP-F8 satellite for the period January 1988 to December 1989. The calculated reflectivities over the deserts are compared against predictions from the Fresnel equations, while the reflectivities over the rainforest are compared against predictions from a radiative transfer model and with field observations. The temporal variation of the reflectivities over the savanna is discussed in relation to those over deserts and rainforest, and in terms of biomass growth and decay. These SSM/I observations can be used to study seasonal and interannual changes of vegetation cover.

The general capability of synthetic aperture radar (SAR) for monitoring forest ecosystems is well documented. However, the majority of SAR studies of forest dynamics use only imagery acquired by one SAR system and are thus limited to the lifecycle of a particular satellite. The synergistic analysis of SAR data from one of the earliest spaceborne SAR missions, the SEASAT mission, with the Japanese JERS-1 satellite-borne SAR is presented. Biophysical parameters frequently retrieved from SAR are tree biomass using backscatter and tree height from the interferometric phase. One potential application that has not been thoroughly examined is mapping of incremental tree growth from SAR backscatter changes. Tree growth measures biomass changes over time, and is correlated to the amount of carbon sequestered by the trees. This paper examines the retrieval of tree growth from multitemporal spaceborne L-band SAR. A SEASAT SAR image from 1978 and a JERS-1 SAR image from 1997 over Thetford forest, UK are used to retrieve tree growth of Corsican Pine stands. Incremental growth was estimated from the changes in backscatter coefficient, and compared to the expected tree growth from general yield class models used by the UK Forestry Commission. The accuracy of the retrieval algorithm depends on the minimum forest stand size included in the analysis. For managed forest plantations, multitemporal L-band SAR has some potential for detecting incremental biomass to support sustainable forest management.

Surface mining and reclamation is the dominant driver of land cover land use change (LCLUC) in the Central Appalachian Mountain region of the Eastern U.S. Accurate quantification of the extent of mining activities is important for assessing how this LCLUC affects ecosystem services such as aesthetics, biodiversity, and mitigation of flooding. We used Landsat imagery from 1976, 1987, 1999 and 2006 to map the extent of surface mines and mine reclamation for eight large watersheds in the Central Appalachian region of West Virginia, Maryland and Pennsylvania. We employed standard image processing techniques in conjunction with a temporal decision tree and GIS maps of mine permits and wetlands to map active and reclaimed mines and track changes through time. For the entire study area, active surface mine extent was highest in 1976, prior to implementation of the Surface Mine Control and Reclamation Act in 1977, with 1.76% of the study area in active mines, declining to 0.44% in 2006. The most extensively mined watershed, Georges Creek in Maryland, was 5.45% active mines in 1976, declining to 1.83% in 2006. For the entire study area, the area of reclaimed mines increased from 1.35% to 4.99% from 1976 to 2006, and from 4.71% to 15.42% in Georges Creek. Land cover conversion to mines and then reclaimed mines after 1976 was almost exclusively from forest. Accuracy levels for mined and reclaimed cover was above 85% for all time periods, and was generally above 80% for mapping active and reclaimed mines separately, especially for the later time periods in which good accuracy assessment data were available. Among other implications, the mapped patterns of LCLUC are likely to significantly affect watershed hydrology, as mined and reclaimed areas have lower infiltration capacity and thus more rapid runoff than unmined forest watersheds, leading to greater potential for extreme flooding during heavy rainfall events.

Unusually cloud-free NOAA satellite thermal infrared images of the Sea of Japan reveal patterns of surface temperature important to understanding surface circulation. A 26 March image of almost the entire ocean basin, cloud-free, shows the appropriate temperature structure of the expected classical surface circulation as formulated by Uda (1934). These include the two-branched pattern of the Tsushima Current, the polar front, and the cold, isothermal surface water of the Maritime Provinces Cold Current (north-central Sea of Japan). Additionally, these observational data show the sea to be nearly filled with mesoscale eddies, as suggested by Yoon and Suginohara (1977) from numerical modeling studies. A 28 March image is supportive of the conclusions of Lim (1971) that the Tsushima Current waters are formed in the outer East China Sea shelf by mixing of Kuroshio waters with outer shelf and slope waters. Satellite-measured surface temperatures differed from surface ship measurements (JMA, Ten-Day Marine Report) by 1–2°C, an offset observed previously under similar atmospheric conditions. Images of this quality can be expected in late winter and early spring, as regional surface cooling facilitates penetration of polar continental air deep into the Sea of Japan and western Pacific.

We present time-series observations by two spaceborne sensors over three desert regions, the Kalahari (in southern Africa), and the Great Victoria Desert and the Great Sandy Desert (in western Australia). The observations are by the Advanced Very High Resolution Radiometer on board the NOAA-7 satellite from April 1982 to December 1984, and by the Scanning Multichannel Microwave Radiometer on board the Nimbus-7 satellite from January 1979 to February 1985. The objective was to compare and contrast seasonal and interannual variation of vegetation over these three deserts using the normalized difference vegetation index and the 37 GHz brightness temperature. The seasonal variation from both sensors was found to be most pronounced over the Kalahari, followed by the Great Sandy Desert and the Great Victoria Desert. The normalized difference vegetation index was roughly identical over the two Australian deserts and was significantly higher for the Kalahari. Additionally, there was no consistent change from both sensors over the two Australian deserts, but a consistent decrease from 1979 to 1984 over the Kalahari was found in the 37 GHz microwave data.

Seasonal snow cover in South America was examined in this study using passive microwave satellite data from the Scanning Multichannel Microwave Radiometer (SMMR) on board the Nimbus-7 satellite and the Special Sensor Microwave Imagers (SSM/I) on board Defense Meteorological Satellite Program (DMSP) satellites. For the period from 1979–2006, both snow cover extent and snow water equivalent (snow mass) were investigated during the coldest months (May–September), primarily in the Patagonia area of Argentina and in the Andes of Chile, Argentina and Bolivia, where most of the seasonal snow is found. Since winter temperatures in this region are often above freezing, the coldest winter month was found to be the month having the most extensive snow cover and usually the month having the deepest snow cover as well. Sharp year-to-year differences were recorded using the passive microwave observations. The average snow cover extent for July, the month with the greatest average extent during the 28-year period of record, is 321,674 km2. In July of 1984, the average monthly snow cover extent was 701,250 km2 — the most extensive coverage observed between 1979 and 2006. However, in July of 1989, snow cover extent was only 120,000 km2. The 28-year period of record shows a sinusoidal like pattern for both snow cover and snow mass, though neither trend is significant at the 95% level.

As cropland and pasture have replaced forest and cerrado in Brazilian Amazônia, concern has mounted over the effects of changing the biogeochemical and hydrological properties of one of the world's great storehouses of biomass and biodiversity. Although much recent effort has focused on the location, effects, and causes of deforestation and cerrado conversion, much less is known about the basin-wide spatial distribution and density of the land use following conversion for crops or pasture.In this paper, we use census and satellite records to develop maps of the distribution and abundance of major agricultural land uses across 4.5×108 ha of Brazilian Amazônia in 1980 and 1995. Results indicate an overall expansion of 7.0×106 ha in total agricultural area in Brazilian Amazônia between 1980 and 1995. The net change during this period is estimated for three different land-use types: croplands (an increase of 0.8×106 ha), natural pastures (a decrease of 8.4×106 ha), and planted pastures (an increase of 14.7×106 ha). These estimates, the first spatially explicit quantifications of agricultural land-use activities in 1980 and 1995 across Brazilian Amazônia, are shown to be consistent with the results of applying a land use change and secondary regrowth model to published deforestation rates for the period.The resulting time slices, presented for each land-use category at 5-min (∼9 km) spatial resolution, allow for the quantification of land-use changes in this region for biogeochemical, demographic and economic models. Several foci of agricultural change existed within Brazilian Amazônia during this period: in the state of Pará, cropland was lost and planted pasture increased markedly; in Mato Grosso, both cropland and planted pasture increased; in Rondônia, planted pasture replacing forest was the primary route to agricultural expansion.

An 18-year time series of monthly NOAA-AVHRR Pathfinder Land burned area was analyzed for the region of tropical Africa, from July 1981 to June 1999. The transition period between NOAA-11 and NOAA-14 platforms from July 1993 to June 1995 was not included due to missing and outlier data. Stability of the time series was addressed for the input variables in the burned area algorithm, reflectance and temperature channels.A Seasonal AutoRegressive Integrated Moving-Average (SARIMA) model was developed for forecasting potential burned area. The SARIMA model identified an autoregressive regular term with 1-month lag and an autoregressive 12-month seasonal term with one season (12 months) component. A cross-correlation between Southern Oscillation Index (SOI) and burned area was statistically significant predictor variable in a time series with 20-month lag. Results show that the SARIMA model with this predictor improved both, fitting and forecasting, residual variance, by 4.1% and 5.6%, respectively, thereby, demonstrating potential relationship between SOI and burned area for the study region. Forecasting was estimated by considering only the first 16 years of the monthly burned area in the time series, from July 1981 to June 1997. The prediction for the following 24 months (from July 1997 to June 1999) was within the 95% confidence level indicating that the forecast was a valid characterization of the modeled process.

Spring–summer (November, December, January) ice sheet and sea ice regional surface albedo, surface temperature, sea ice concentration and sea ice extent averages and trends from 1981 to 2000 have been calculated for the Antarctic area. In this research the AVHRR Polar Pathfinder 5-km EASE-Grid Composites and the combined SMMR and SSMI data sets from the National Snow and Ice Data Center (NSIDC), Boulder, Colorado have been employed. A regional analysis has been made for five longitudinal sectors around Antarctica: the Weddell Sea (WS), the Indian Ocean (IO), the Pacific Ocean (PO), the Ross Sea (RS) and the Bellingshausen–Amundsen Sea (BS). The IO and PO sectors show ice sheet albedos of 0.85 and temperatures of − 25 °C. The corresponding values in the RS and BS sectors are 0.80 and − 16 °C respectively. The sea ice albedo is about 0.60 in the RS, BS and WS sectors and 0.55 in the IO and PO sectors. The average sea ice temperature varies around − 12 °C. All the sectors show slight increasing spring–summer albedo trends and decreasing spring–summer temperature trends and similar interannual variability in albedo and surface temperature. The steepest ice sheet albedo trend of 0.0019 ± 0.0009/yr is found in the RS sector. The steepest sea ice albedo trend of 0.0044 ± 0.0017 /yr occurs in the PO sector. The steepest temperature trends for both the ice sheet and sea ice occur in the BS sector, having values of − 0.075 ± 0.040 °C/yr and − 0.107 ± 0.027 °C/yr respectively. The sea ice concentration shows slight increasing trends, the highest being in the PO sector (0.3 ± 0.12%/yr), whereas the sea ice extent trends are near zero with the exception of the RS sector (14,700 ±8600 km2/yr) and the BS sector (− 13,000 ± 6400 km2/yr).

Measurements of spring phenological dates in boreal regions using NDVI can be affected by snowmelt. This impacts the analysis of interannual variations in phenology and the estimates of annual carbon fluxes. For these two objectives, snowmelt effect must be removed from the phenological detection. We propose a methodology for determining the date of onset of greening in the 1982–2004 period using SPOT-VEGETATION (VGT) and NOAA Advanced Very High Resolution Radiometer (AVHRR) data. From 1998 onwards, the date of onset of greening is taken as the date at which the Normalized Difference Water Index (NDWI), calculated from SPOT-VGT near and short-wave infrared bands, starts increasing. This index decreases with snowmelt but increases with vegetation greening. For the 1982–2001 period, the date of onset of greening is the date at which AVHRR-NDVI equals a pixel specific threshold (PST), determined using the results of the NDWI method in the years common to the two datasets. The methods are validated using in situ measurements of the dates of leaf appearance. RMSE of 6.7 and 7.8 days, respectively, is found using NDWI-VGT and PST-NOAA methodologies, and the difference between the two methodologies in the common years is small. Very importantly, the dates are not biased. The interannual variations of the 23-year spring phenology dataset on the study area in northern Eurasia are analysed. In average over the study area, an advance of 8 days and a delay of 3.6 days are, respectively, found over the periods 1982–1991 and 1993–2004. These results confirm and complete previous studies about the greening trend, remove the uncertainty due to snow, and may improve carbon budget calculations.