On the upstream inputs into the MODIS primary productivity products using biometric data from oil palm plantations

This study evaluated the influence of upstream inputs into the Moderate Resolution Imaging Spectroradiometer (MODIS) primary productivity products, termed the MOD17, at tropical oil palm plantations (Elaeis guineensis Jacq.). Evaluation of MOD17 using oil palm plantations as test sites is ideal because the plantations are cultivated on large areas which are comparable with the size of MODIS pixels. It is difficult to find test sites covered by other single species in a whole pixel. The upstream inputs studied included (1) MODIS land cover, (2) the National Centers for Environmental Prediction–Department of Energy (NCEP-DOE) Reanalysis 2 meteorological data set, (3) MODIS leaf area index/fraction of photosynthetically active radiation (LAI/fPAR), and (4) MODIS maximum light-use efficiency (maximum LUE). Oil palm biometric and local meteorological data were utilized as ground data. Furthermore, scaling up oil palm LAI and fPAR from plot scale to regional scale (Peninsular Malaysia) was done empirically by correlating oil palm LAI derived from the hemispherical photography technique with radiance information from the Disaster Monitoring Constellation 2 satellite (UK-DMC 2). The upscaled LAI/fPAR developed in this study was used to evaluate the MODIS LAI/fPAR. The results showed that the MODIS land-cover product has an overall accuracy of 78.8% when compared to the Peninsular Malaysia land-use map produced by the Department of Agriculture, Malaysia. Regarding the NCEP-DOE Reanalysis 2 data set, vapour pressure deficit (VPD) and photosynthetically active radiation (PAR) contain large uncertainties in our study area. However, MODIS LAI and fPAR were correlated relatively well with the upscaled LAI (R² = 0.50) and the upscaled fPAR (R² = 0.60), respectively. The constant values of maximum LUE for croplands and evergreen broadleaf forest ecosystems are lower than the maximum LUE of oil palm. The relative predictive error assessment showed that the MOD17 net primary productivity (NPP) overestimated oil palm NPP derived from biometric methods by 142–204%. We replaced the upstream inputs of MOD17 by the local inputs for estimating oil palm GPP and NPP in Peninsular Malaysia. This was done by (1) assigning maximum LUE for oil palm plantations as a constant at 1.68 g C m^?2 day^?1, (2) utilizing meteorological data from local meteorological stations, and (3) using the upscaled fPAR of oil palm plantations. The amount of oil palm GPP and NPP for Peninsular Malaysia in 2010 were estimated to be ~0.09 Pg C year^?1 (or equivalent to ~0.33 Pg CO₂ year^?1) and ~0.03 Pg C year^?1 (~0.11 Pg CO₂ year^?1), respectively, indicating that oil palm plantations in Peninsular Malaysia can play an important role in global carbon sequestration. In the future there is likely to be a demand for MODIS GPP and NPP products that are more accurate than those currently generated by MOD17. We recommend future developments of the MOD17 processing system to allow improvements in the upstream input parameters, in the manner described in this article, both for global processing and for the production of more accurate values for GPP and NPP at regional and local scales.     

Comparison of MODIS gross primary production estimates for forests across the U.S.A. with those generated by a simple process model, 3-PGS

Since 2000, NASA's Moderate Resolution Imaging Spectro-radiometer (MODIS) has provided 1 × 1 km estimates of 8-day gross primary production (GPP). The MODIS algorithm computes GPP as a simple function of absorbed photosynthetically active radiation and a regionally assigned light-use conversion efficiency (LUE) that is reduced if temperature or atmospheric vapor pressure deficits are suboptimal. We compared MODIS-derived GPP estimates for forested areas across the United States of America (U.S.A.) with those generated by the 3-PGS (Physiological Principles Predicting Growth using Satellite data) model, the latter of which considers spatial variation in available soil water storage capacity (ASWC) and nitrogen content. We expected seasonal and annual MODIS GPP values to be in close agreement with those derived from the 3-PGS model in regions with adequate precipitation, soil water storage, and moderately fertile soils. 3-PGS was initially run with STATSGO-derived soils information provided by the Oak Ridge National Laboratory. The analysis was expanded to include sensitivity analyses with ASWC set at 50, 100, 300, and 400 mm to identify areas within nine major ecoregions where drought might prove to be a major limitation on GPP. The majority of forests across the U.S.A. were relatively insensitive to large variations in ASW storage. In areas where ASWC was assumed < 200 mm and average annual rainfall was < 100 mm yr^− 1, GPP was predicted to be reduced by > 60%. There was generally good agreement (within 20%) between MODIS and 3-PGS estimates of forest GPP across the U.S.A. GPP predicted by the MODIS model was generally higher in ecoregions with substantial drought and with relatively low soil fertility. The latter, which influences LUE, was more than twice as important as soil drought.     

Can a satellite-derived estimate of the fraction of PAR absorbed by chlorophyll (FAPARchl) improve predictions of light-use efficiency and ecosystem photosynthesis for a boreal aspen forest?

We used daily MODerate resolution Imaging Spectroradiometer (MODIS) imagery obtained over a five-year period to analyze the seasonal and inter-annual variability of the fraction of absorbed photosynthetically active radiation (FAPAR) and photosynthetic light use efficiency (LUE) for the Southern Old Aspen (SOA) flux tower site located near the southern limit of the boreal forest in Saskatchewan, Canada. To obtain the spectral characteristics of a standardized land area to compare with tower measurements, we scaled up the nominal 500 m MODIS products to a 2.5 km × 2.5 km area (5 × 5 MODIS 500 m grid cells). We then used the scaled-up MODIS products in a coupled canopy-leaf radiative transfer model, PROSAIL-2, to estimate the fraction of absorbed photosynthetically active radiation (APAR) by the part of the canopy dominated by chlorophyll (FAPAR_chl) versus that by the whole canopy (FAPAR_canopy). Using the additional information provided by flux tower-based measurements of gross ecosystem production (GEP) and incident PAR, we determined 90-minute averages for APAR and LUE (slope of GEP:APAR) for both the physiologically active foliage (APAR_chl, LUE_chl) and for the entire canopy (APAR_canopy, LUE_canopy).The flux tower measurements of GEP were strongly related to the MODIS-derived estimates of APAR_chl (r² = 0.78) but only weakly related to APAR_canopy (r² = 0.33). Gross LUE between 2001 and 2005 for LUE_chl was 0.0241 µmol C µmol^− 1 PPFD whereas LUE_canopy was 36% lower. Time series of the 5-year normalized difference vegetation index (NDVI) were used to estimate the average length of the core growing season as days of year 152-259. Inter-annual variability in the core growing season LUE_chl (µmol C µmol^− 1 PPFD) ranged from 0.0225 in 2003 to 0.0310 in 2004. The five-year time series of LUE_chl corresponded well with both the seasonal phase and amplitude of LUE from the tower measurements but this was not the case for LUE_canopy. We conclude that LUE_chl derived from MODIS observations could provide a more physiologically realistic parameter than the more commonly used LUE_canopy as an input to large-scale photosynthesis models.     

Regional mapping of gross light-use efficiency using MODIS spectral indices

Direct estimation of photosynthetic light-use efficiency (LUE) from space would be of significant benefit to LUE-based models which use inputs from remote sensing to estimate terrestrial productivity. The Photochemical Reflectance Index (PRI) has shown promise in tracking LUE at the leaf- to small canopy levels, but its use at regional to global scales still remains a challenge. In this study, we used different formulations of PRI calculated from the MODIS ocean band centered at 531 nm and a set of alternative reference bands at 488, 551, and 678 nm to explore the relationship between PRI and LUE where LUE was measured at eight eddy covariance flux towers located in the boreal forest of Saskatchewan, Canada. The magnitude and variability of LUE was significantly lower at the times when useful MODIS ocean band images were available (i.e. around midday under clear-sky conditions) relative to the rest of the growing season. PRI₆₇₈ (reference band at 678 nm) showed the strongest relationship (r² = 0.70) with LUE_90a (i.e. 90-minute mean LUE calculated using Absorbed Photosynthetically Active Radiation, APAR), but only when all sites were combined. Overall, the relationships between the MODIS PRIs and LUE_90a were always stronger for observations closer to the backscatter direction and there were no significant differences in the strength of the correlations whether LUE was calculated based on incident PAR or on APAR. Predictions of ecosystem photosynthesis at the time of the MODIS overpasses were significantly improved by multiplying either PAR or APAR by MODIS PRI (r² improved from 0.09 to 0.44 and 0.54 depending on the PRI formulation).We used our PRI-LUE model to create a regional LUE_90a map for the three cover types covering 47,500 km² around the flux sites. The MODIS PRI-derived LUE_90a map appeared to capture more realistic spatial heterogeneity of LUE across the landscape compared to a daily LUE map derived using the look-up table in the MODIS GPP (MOD17) algorithm. While our LUE map is only a snapshot of minimum regional LUE_90a values, with appropriate gap-filling methods it could be used to improve regional-scale monitoring of GPP. Moreover, the strong relationship between midday and daily LUE on clear days (r² = 0.93) indicates that instantaneous MODIS observations of LUE_90a could be used to estimate daily LUE. Finally, pixel shadow fraction from the 5-Scale geometric-optical model was closely related to both MODIS PRI and tower-derived LUE suggesting that differences in stand leaf area and in diffuse illumination among flux sites play a role in the relationship we observed between LUE and MODIS PRI.     

Impact of nutrients on peatland GPP estimations using MODIS time series data

Time series of satellite sensor-derived data can be used in the light use efficiency (LUE) model for gross primary productivity (GPP). The LUE model and a closely related linear regression model were studied at an ombrotrophic peatland in southern Sweden. Eddy covariance and chamber GPP, incoming and reflected photosynthetic photon flux density (PPFD), field-measured spectral reflectance, and data from the Moderate Resolution Imaging Spectroradiometer (MODIS) were used in this study. The chamber and spectral reflectance measurements were made on four experimental treatments: unfertilized control (Ctrl), nitrogen fertilized (N), phosphorus fertilized (P), and nitrogen plus phosphorus fertilized (NP). For Ctrl, a strong linear relationship was found between GPP and the photosynthetically active radiation absorbed by vegetation (APAR) (R² = 0.90). The slope coefficient (ε_s, where s stands for “slope”) for the linear relationship between seasonal time series of GPP and the product of the normalized difference vegetation index (NDVI) and PPFD was used as a proxy for the light use efficiency factor (ε). There were differences in ε_s depending on the treatments with a significant effect for N compared to Ctrl (ANOVA: p = 0.042, Tukey's: p ≤ 0.05). Also, ε_s was linearly related to the cover degree of vascular plants (R² = 0.66). As a sensitivity test, the regression coefficients (ε_s and intercept) for each treatment were used to model time series of 16-day GPP from the product of MODIS NDVI and PPFD. Seasonal averages of GPP were calculated for 2005, 2006, and 2007, which resulted in up to 19% higher average GPP for the fertilization treatments compared to Ctrl. The main conclusion is that the LUE model and the regression model can be applied in peatlands but also that temporal and spatial changes in ε or the regression coefficients should be considered.     

Comparison of MODIS, eddy covariance determined and physiologically modelled gross primary production (GPP) in a Douglas-fir forest stand

Quantification of the magnitude of net terrestrial carbon (C) uptake, and how it varies inter-annually, is an important question with future potential sequestration influenced by both increased atmospheric CO₂ and changing climate. However the assessment of differences in measured and modeled C accumulation is a challenging task due to the significant fine scale variation occurring in terrestrial productivity due to soil, climate and vegetation characteristics as well as difficulties in measuring carbon accumulation over large spatial areas. The Moderate Resolution Imaging Spectroradiometer (MODIS) offers a means of monitoring gross primary production (GPP), both spatially and temporally, routinely from space. However it is critical to compare and contrast the temporal dynamics of the C and water fluxes with those measured from ground-based networks, or estimated using physiological models. In this paper, using a number of approaches, our objective is to determine if any systematic biases exists in either the MODIS, or the modeled estimates of fluxes, relative to the measurements made over an evergreen, needleleaf temperate rainforest on Vancouver Island, Canada. Results indicate that 8-day GPP as predicted with a simple physiological model (3PGS), forced using local meteorology and canopy characteristics, matched measured fluxes very well (r² = 0.86, p < 0.001) with no significant difference between eddy covariance (EC) and modeled GPP (p < 0.001). In addition, modeled water supply closely matched measured relative available soil water content at the site. Using canopy characteristics from the MODIS fraction of photosynthetically active radiation (fPAR) algorithm, slightly reduced the correspondence of the predictions due to a large number of unsuccessful retrievals (83%) due to sun angle, snow and cloud. Predictions of GPP based on the MODIS GPP algorithm, forced using local meteorology and canopy characteristics, were also highly correlated with EC measurements (r² = 0.89, p < 0.001) however these estimates were biased under predicting GPP. Estimates of GPP based on the most recent MODIS reprocessing (collection 4.5) remained highly correlated (r² = 0.88, p < 0.001) yet were also the most biased with the estimates being 30% less than the EC-measured GPP. Most of the variance in GPP at the site was explained by the absorbed photosynthetically active radiation. We also compared the nighttime respiration as measured over 2 years at the site with the minimum 8-day MODIS land surface temperature and found a significant relationship (r² = 0.57), similar to other studies.     

Evaluation of satellite based primary production modelling in the semi-arid Sahel

The MODIS (Moderate Resolution Imaging Spectroradiometer) primary productivity products are evaluated against observed Above-ground Net Primary Production (AGNPP) in the semi-arid Senegal 2001. MODIS net primary productivity (NPP) modelling is a light use efficiency (LUE) based approach incorporating constraints on vegetation productivity arising from simulated radiation, water demand and temperature data from NASA's Data Assimilation Office (DAO). Annually integrated MODIS PSN (MOD17A2 net photosynthesis, Collection 4) explains more of the observed biomass variation (r² = 0.77) than MODIS fAPAR (fraction Absorbed Photosynthetically Active Radiation, Collection 4) (r² = 0.72), indicating the effect of including the canopy stress scalar (ε_s) based on DAO data combined with modelled maintenance respiration costs (of leaf and fine roots). Annual MODIS NPP (MOD17A3, Collection 4 (C4) and Collection 4.5 (C4.5)) including growth respiration and live wood maintenance respiration costs and modified DAO input (C4.5) however increases the residual unexplained observed AGNPP variance (C4 NPP; r² = 0.49) (C4.5 NPP; r² = 0.37). The overall quality of the annual NPP MODIS C4 and C4.5 products are moderate for the semi-arid Senegal because of the annual respiration cost modelling and a change in C4.5 biome-specific parameters stored in a Biome Properties Look-Up Table (BPLUT) is the main contributor to the observed discrepancy between C4 and C4.5 NPP. The dynamic range of the values of all MOD17 products was too low when compared to observed AGNPP. An estimate of canopy water stress (SIWSI; Shortwave Infrared Water Stress Index) derived from MODIS channels 2 and 6 and photosynthetically active radiation (PAR) irradiance derived from geostationary METEOSAT data were tested for primary production modelling using a stepwise linear regression analysis. PAR irradiance was combined with MODIS fAPAR into APAR (Absorbed Photosynthetically Active Radiation) explaining 79% of the observed AGNPP variation. Introducing SIWSI significantly increased the explained variance of observed AGNPP (r² = 0.89). MODIS-derived percentage tree cover was tested as a predictor based on the hypothesis that tree cover provides information on differences in respiratory costs between trees and grasses thereby accounting for variations in the LUE conversion efficiency ε. No significant reduction in residual unexplained AGNPP variance was found. Earth observation based derivation of PAR and canopy water stress from SIWSI suggest potential improvements to primary production models in semi-arid biomes that can be implemented in general NPP modelling LUE methodology.     

Evaluation of MODIS Gross Primary Productivity (GPP) in tropical monsoon regions

Near real-time vegetation indices derived from MODIS (MODerate resolution Imaging Spectroradiometer) observations (http://modis.gsfc.nasa.gov) provide a first opportunity to monitor ecohydrological systems globally at a spatial resolution consistent with biophysical processes at the field scale. Here, we present work toward the quantitative estimation of the uncertainty associated with MODIS Gross Primary Productivity (GPP), an end-product that depends on several MODIS derived vegetation indices. GPP products, available at 8-day and 1-km resolutions, were evaluated in two representative tropical ecosystems: a mixed forest site in the humid tropics (the Marsyandi river basin in the Nepalese Himalayas), and an open shrubland site in a semi-arid region (the Sonora river basin in northern Mexico). The MODIS-GPP products were compared against simulations made with a process-based biochemical-hydrology model driven by flux tower meteorological observations. Whereas the temporal march of vegetation indices and GPP products is consistent between the model and the algorithm, our study indicates that that there is a positive bias in the case of the mixed forest biome in the Marsyandi basin, and a negative bias in the case of open shrublands in the Sonora basin. We examined the error contribution from the DAO meteorological data used in the standard MODIS GPP products. The bias between the GPP estimates using DAO and tower meteorology is − 2.77 gC/m²/day (i.e., − 77% of the mean of the tower-based GPP) in the Marsyandi, and 0.33 gC/m²/day (i.e., 18% of the mean of the tower-based GPP) in Sonora. Analysis of the temporal evolution of the discrepancies between the model and the MODIS algorithm points to the need for examining the light use efficiency parameterization, especially with regard to the representation of nonlinear functional dependencies on vapor pressure deficit (VPD), photosynthetically available radiation (PAR), and seasonal evolution of the productive capacity of vegetation as influenced by water stress.     

Comparison of multiple models for estimating gross primary production using MODIS and eddy covariance data in Harvard Forest

Gross primary production (GPP) defined as the overall rate of fixation of carbon through the process of vegetation photosynthesis is important for carbon cycle and climate change research. Three models, the Vegetation Photosynthesis Model (VPM), the Temperature and Greenness (TG) model and the Vegetation Index (VI) model have been compared for the estimation of GPP in Harvard Forest from 2003 to 2006 using climate variables acquired by eddy covariance (EC) measurements and Moderate Resolution Imaging Spectroradiometer (MODIS) satellite images. All these models provide more reliable estimates of GPP than that of MODIS GPP product. High Pearsons correlation coefficients r equal to 0.94, 0.92 and 0.90 are observed for the VPM, the TG and the VI model, respectively. Relationships between GPP and land surface temperature (LST, R² = 0.72), and vapor pressure deficit (VPD, R² = 0.45) indicate that climate variables are important for GPP estimation. Due to proper characterization of temperature, water stress and leaf age by three scalars, VPM best follows the seasonal variations of GPP. By incorporation of the MODIS surface reflectance and LST product, the TG model is the most suitable choice for areas without prior knowledge as it is based entirely on remote sensing observations. Results from the VI model demonstrate the possibility of using a single vegetation index for light use efficiency (LUE) estimation in deciduous forest that is of high spatial heterogeneity. The validation and comparison of models will be helpful in development of future GPP models using combinations of climate variables and/or remote sensing observations.     

基于涡度数据的东北草地光能利用率模型构建与验证

草地作为地球上分布最广的植被类型,在陆地碳循环中发挥着重要作用。草地生产力是估算产草量的基础,准确模拟生产力对草原资源合理利用及生态保护具有重要意义。以东北草地生产力为研究核心,利用涡度相关通量观测数据、遥感数据和气象数据,构建和检验东北草地光能利用率模型。东北草地光能利用率模型以归一化物候植被指数(NDPI)代表光合有效辐射吸收比例,以地表水分指数(LSWI)+0.5表示水分胁迫因子。基于44个草原站的通量数据对东北草地光能利用率模型进行验证,东北草地光能利用率模型的R²为0.855,高于MODIS GPP产品(R²=0.719),略高于VPM GPP产品(R²=0.848),东北草地光能利用率模型的MAE和RMSE分别为0.374 g Cm^-2和0.735 g Cm^-2,低于MODIS GPP产品(MAE=0.562 g Cm^-2,RMSE=1.026 g Cm^-2)和VPM GPP产品(MAE=0.667 g Cm^-2...     

A continuous measure of gross primary production for the conterminous United States derived from MODIS and AmeriFlux data

The quantification of carbon fluxes between the terrestrial biosphere and the atmosphere is of scientific importance and also relevant to climate-policy making. Eddy covariance flux towers provide continuous measurements of ecosystem-level exchange of carbon dioxide spanning diurnal, synoptic, seasonal, and interannual time scales. However, these measurements only represent the fluxes at the scale of the tower footprint. Here we used remotely sensed data from the Moderate Resolution Imaging Spectroradiometer (MODIS) to upscale gross primary productivity (GPP) data from eddy covariance flux towers to the continental scale. We first combined GPP and MODIS data for 42 AmeriFlux towers encompassing a wide range of ecosystem and climate types to develop a predictive GPP model using a regression tree approach. The predictive model was trained using observed GPP over the period 2000-2004, and was validated using observed GPP over the period 2005-2006 and leave-one-out cross-validation. Our model predicted GPP fairly well at the site level. We then used the model to estimate GPP for each 1 km × 1 km cell across the U.S. for each 8-day interval over the period from February 2000 to December 2006 using MODIS data. Our GPP estimates provide a spatially and temporally continuous measure of gross primary production for the U.S. that is a highly constrained by eddy covariance flux data. Our study demonstrated that our empirical approach is effective for upscaling eddy flux GPP data to the continental scale and producing continuous GPP estimates across multiple biomes. With these estimates, we then examined the patterns, magnitude, and interannual variability of GPP. We estimated a gross carbon uptake between 6.91 and 7.33 Pg C yr^− 1 for the conterminous U.S. Drought, fires, and hurricanes reduced annual GPP at regional scales and could have a significant impact on the U.S. net ecosystem carbon exchange. The sources of the interannual variability of U.S. GPP were dominated by these extreme climate events and disturbances.     

On intra-annual EVI variability in the dry season of tropical forest: A case study with MODIS and hyperspectral data

Considerable controversy is associated with dry season increases in the Enhanced Vegetation Index (EVI), observed using the Moderate Resolution Imaging Spectroradiometer (MODIS), compared with field-based estimates of decreasing plant productivity. Here, we investigate potential causes of intra-annual variability by comparing EVI from mature forest with field-measured Leaf Area Index (LAI) to validate space-based observations. EVI was calculated from 19 nadir and off-nadir Hyperion images in the 2005 dry season, and inspected for consistency with MODIS observations from 2004 to 2009. The objective was to evaluate the possible influence of the view-illumination geometry and of canopy foliage and leaf flush on the EVI. Spectral mixture models were used to evaluate the relationship between EVI and the shade fraction, a measure that varies with pixel brightness. MODIS LAI values were compared with LAI estimated using hemispherical photographs taken in two field campaigns in the dry season. To keep LAI and leaf flush conditions as constant variables and vary solar illumination, we used airborne Hyperspectral Mapper (Hymap) data acquired over mature forest from another region on the same day but with two distinct solar zenith angles (SZA) (29° and 53°). Results showed that intra-annual variability in MODIS and nadir Hyperion EVI in the dry season of tropical forest were driven by solar illumination effects rather than changes in LAI. The reflectance of the MODIS and Hyperion blue, red and near infrared (NIR) bands was higher at the end of the dry season because of the predominance of sunlit canopy components for the sensors due to decreasing SZA from June (44°) to September (26°). Because EVI was highly correlated with the reflectance of the NIR band used to generate it (r of + 0.98 for MODIS and + 0.88 for Hyperion), this vegetation index followed the general NIR pattern, increasing with smaller SZA towards the end of the dry season. Hyperion EVI was inversely correlated with the shade fraction (r = − 0.93). Changes in canopy foliage detected from MODIS LAI data were not consistent with LAI estimates from hemispherical photographs. Although further research is necessary to measure the impact of leaf flush on intra-annual EVI variability in the Querência region, analysis of Hymap data with fixed LAI and leaf flush conditions confirmed the influence of the illumination effects on the EVI.     

The effects of aggregated land cover data on estimating NPP in northern Wisconsin

Ecosystem models are routinely used to estimate net primary production (NPP) from the stand to global scales. Complex ecosystem models, implemented at small scales (< 10 km²), are impractical at global scales and, therefore, require simplifying logic based on key ecological first principles and model drivers derived from remotely sensed data. There is a need for an improved understanding of the factors that influence the variability of NPP model estimates at different scales so we can improve the accuracy of NPP estimates at the global scale. The objective of this study was to examine the effects of using leaf area index (LAI) and three different aggregated land cover classification products-two factors derived from remotely sensed data and strongly affect NPP estimates-in a light use efficiency (LUE) model to estimate NPP in a heterogeneous temperate forest landscape in northern Wisconsin, USA. Three separate land cover classifications were derived from three different remote sensors with spatial resolutions of 15, 30, and 1000 m. Average modeled net primary production (NPP) ranged from 402 gC m^− 2 year^− 1 (15 m data) to 431 gC m^− 2 year^− 1 (1000 m data), for a maximum difference of 7%. Almost 50% of the difference was attributed each to LAI estimates and land cover classifications between the fine and coarse scale NPP estimate. Results from this study suggest that ecosystem models that use biome-level land cover classifications with associated LUE coefficients may be used to model NPP in heterogeneous land cover areas dominated by cover types with similar NPP. However, more research is needed to examine scaling errors in other heterogeneous areas and NPP errors associated with deriving LAI estimates.     

Improvements of the MODIS terrestrial gross and net primary production global data set

MODIS primary production products (MOD17) are the first regular, near-real-time data sets for repeated monitoring of vegetation primary production on vegetated land at 1-km resolution at an 8-day interval. But both the inconsistent spatial resolution between the gridded meteorological data and MODIS pixels, and the cloud-contaminated MODIS FPAR/LAI (MOD15A2) retrievals can introduce considerable errors to Collection4 primary production (denoted as C4 MOD17) results. Here, we aim to rectify these problems through reprocessing key inputs to MODIS primary vegetation productivity algorithm, resulting in improved Collection5 MOD17 (here denoted as C5 MOD17) estimates. This was accomplished by spatial interpolation of the coarse resolution meteorological data input and with temporal filling of cloud-contaminated MOD15A2 data. Furthermore, we modified the Biome Parameter Look-Up Table (BPLUT) based on recent synthesized NPP data and some observed GPP derived from some flux tower measurements to keep up with the improvements in upstream inputs. Because MOD17 is one of the down-stream MODIS land products, the performance of the algorithm can be largely influenced by the uncertainties from upstream inputs, such as land cover, FPAR/LAI, the meteorological data, and algorithm itself. MODIS GPP fits well with GPP derived from 12 flux towers over North America. Globally, the 3-year MOD17 NPP is comparable to the Ecosystem Model-Data Intercomparison (EMDI) NPP data set, and global total MODIS GPP and NPP are inversely related to the observed atmospheric CO₂ growth rates, and MEI index, indicating MOD17 are reliable products. From 2001 to 2003, mean global total GPP and NPP estimated by MODIS are 109.29 Pg C/year and 56.02 Pg C/year, respectively. Based on this research, the improved global MODIS primary production data set is now ready for monitoring ecological conditions, natural resources and environmental changes.     

Performances of neural networks for deriving LAI estimates from existing CYCLOPES and MODIS products

This paper evaluates the performances of a neural network approach to estimate LAI from CYCLOPES and MODIS nadir normalized reflectance and LAI products. A data base was generated from these products over the BELMANIP sites during the 2001-2003 period. Data were aggregated at 3 km × 3 km, resampled at 1/16 days temporal frequency and filtered to reject outliers. VEGETATION and MODIS reflectances show very consistent values in the red, near infrared and short wave infrared bands. Neural networks were trained over part of this data base for each of the 6 MODIS biome classes to retrieve both MODIS and CYCLOPES LAI products.Results show very good performances of neural networks to estimate the original LAI products with an overall root mean square error (RMSE) around 0.5 for MODIS LAI from both MODIS and CYCLOPES normalized reflectances and a RMSE ranging between 0.12 (CYCLOPES reflectances) and 0.29 (MODIS reflectances) for CYCLOPES LAI. A drop of 15% of performance was found by training MODIS biome dependant algorithm by a single network over all the classes at the same time. More detailed analyses show that CYCLOPES and MODIS LAI values are very consistent for grasses and crops. Conversely, other biomes including shrubs, savanna, needleleaf and broadleaf forests show significant discrepancies, mainly due to differences between LAI definitions used between CYCLOPES (closer to effective LAI) and MODIS (closer to true LAI). However, products derived from the original CYCLOPES LAI products show a better agreement with both effective and true LAI ground measurements values. MODIS LAI products show more instability, partly because of the slightly shorter temporal resolution as compared to CYCLOPES.These results confirm the interest and versatility of neural networks for operational algorithms. This approach could be extended to other products or sensors, and may constitute a step forward for the fusion of data from several sensors, hence contributing to develop ‘virtual constellations’.     

Improved algorithm for MODIS satellite retrievals of aerosol optical thickness over land in dusty atmosphere: Implications for air quality monitoring in China

A new algorithm, using the Moderate Resolution Imaging Spectroradiometer (MODIS) satellite reflectance and aerosol single scattering properties simulated from a chemistry transport model (GEOS-Chem), is developed to retrieve aerosol optical thickness (AOT) over land in China during the spring dust season. The algorithm first uses a “dynamic lower envelope” approach to sample the MODIS dark-pixel reflectance data in low AOT conditions, to derive the local surface visible (0.65 μm)/near infrared (NIR, 2.1 μm) reflectance ratio. Joint retrievals of AOT at 0.65 μm and surface reflectance at 2.1 μm are then performed, based on the time, location, and spectral-dependent single scattering properties of the dusty atmosphere as simulated by the GEOS-Chem. A linearized vector radiative transfer model (VLIDORT) that simultaneously computes the top-of-atmosphere reflectance and its Jacobian with respect to AOT, is used in the forward component of the inversion of MODIS reflectance to AOT. Comparison of retrieved AOT results in April and May of 2008 with AERONET observations shows a strong correlation (R = 0.83), with small bias (0.01), and small RMSE (0.17); the figures are a substantial improvement over corresponding values obtained with the MODIS Collection 5 AOT algorithm for the same study region and time period. The small bias is partially due to the consideration of dust effect at 2.1 μm channel, without which the bias is − 0.05. The surface PM₁₀ (particulate matter with diameter less than 10 μm) concentrations derived using this improved AOT retrieval show better agreement with ground observations than those derived from GEOS-Chem simulations alone, or those inferred from the MODIS Collection 5 AOT. This study underscores the value of using satellite reflectance to improve the air quality modeling and monitoring.     

Estimation of gross primary production in Moso bamboo forest based on light-use efficiency derived from MODIS reflectance data

Assessing the contribution of Moso bamboo (Phyllostachys pubescens) forest to forest ecosystem carbon storage requires accurate estimation of gross primary production (GPP). Based on measurements of light-use efficiency (LUE), defined as the ratio of measured GPP to photosynthetically active radiation (PAR), from the eddy covariance flux tower, the linear regression model and partial least squares regression model were used for estimation of LUE using the Moderate-Resolution Imaging Spectroradiometer (MODIS) reflectance data. GPP estimates were then calculated by the product of LUE estimates and PAR (named the LUE-PAR model), which was compared with GPP from the GPP algorithm designed for the MODIS sensor aboard the Aqua and Terra platforms (MOD17A2 model) and the EC-LUE model. The results revealed the PLS model performed better than the linear regression model in LUE estimation but had lager uncertainties in high and low LUE values. GPP estimates driven by a MODIS-based radiation product with high spatial resolution was more accurate than those driven by Modern-Era Retrospective Analysis for Research and Applications (MERRA) radiation product from the NASA’s Global Modelling and Assimilation Office data set. The LUE-PAR model had the highest accuracy than the other two LUE models. The GPP values derived from the EC-LUE model driven by photosynthetically active radiation (PAR) from MERRA and maximum LUE from the EC data were overestimated due to the overestimation in MERRA radiation product. The GPP values derived from the MOD17A2 model driven by PAR from the MERRA and maximum LUE from the biome properties look-up table were underestimated due to underestimation in the maximum LUE of Moso bamboo forest. This study implied that the LUE-PAR model driven by LUE estimates from the PLS model and PAR from MERRA is a superior approach in improving GPP simulations, and PAR products with high spatial resolution and accurate species-specific maximum LUE are necessary for the LUE models in estimating GPP at regional scale.     

Improvements to a MODIS global terrestrial evapotranspiration algorithm

MODIS global evapotranspiration (ET) products by Mu et al. [Mu, Q., Heinsch, F. A., Zhao, M., Running, S. W. (2007). Development of a global evapotranspiration algorithm based on MODIS and global meteorology data. Remote Sensing of Environment, 111, 519-536. doi: 10.1016/j.rse.2007.04.015] are the first regular 1-km² land surface ET dataset for the 109.03 Million km² global vegetated land areas at an 8-day interval. In this study, we have further improved the ET algorithm in Mu et al. (2007a, hereafter called old algorithm) by 1) simplifying the calculation of vegetation cover fraction; 2) calculating ET as the sum of daytime and nighttime components; 3) adding soil heat flux calculation; 4) improving estimates of stomatal conductance, aerodynamic resistance and boundary layer resistance; 5) separating dry canopy surface from the wet; and 6) dividing soil surface into saturated wet surface and moist surface. We compared the improved algorithm with the old one both globally and locally at 46 eddy flux towers. The global annual total ET over the vegetated land surface is 62.8 × 10³ km³, agrees very well with other reported estimates of 65.5 × 10³ km³ over the terrestrial land surface, which is much higher than 45.8 × 10³ km³ estimated with the old algorithm. For ET evaluation at eddy flux towers, the improved algorithm reduces mean absolute bias (MAE) of daily ET from 0.39 mm day⁻¹ to 0.33 mm day⁻¹ driven by tower meteorological data, and from 0.40 mm day⁻¹ to 0.31 mm day⁻¹ driven by GMAO data, a global meteorological reanalysis dataset. MAE values by the improved ET algorithm are 24.6% and 24.1% of the ET measured from towers, within the range (10-30%) of the reported uncertainties in ET measurements, implying an enhanced accuracy of the improved algorithm. Compared to the old algorithm, the improved algorithm increases the skill score with tower-driven ET estimates from 0.50 to 0.55, and from 0.46 to 0.53 with GMAO-driven ET. Based on these results, the improved ET algorithm has a better performance in generating global ET data products, providing critical information on global terrestrial water and energy cycles and environmental changes.     

Development of a global evapotranspiration algorithm based on MODIS and global meteorology data

The objective of this research is to develop a global remote sensing evapotranspiration (ET) algorithm based on Cleugh et al.'s [Cleugh, H.A., R. Leuning, Q. Mu, S.W. Running (2007) Regional evaporation estimates from flux tower and MODIS satellite data. Remote Sensing of Environment 106, page 285-304- 2007 (doi: 10.1016/j.rse.2006.07.007).] Penman-Monteith based ET (RS-PM). Our algorithm considers both the surface energy partitioning process and environmental constraints on ET. We use ground-based meteorological observations and remote sensing data from the MODerate Resolution Imaging Spectroradiometer (MODIS) to estimate global ET by (1) adding vapor pressure deficit and minimum air temperature constraints on stomatal conductance; (2) using leaf area index as a scalar for estimating canopy conductance; (3) replacing the Normalized Difference Vegetation Index with the Enhanced Vegetation Index thereby also changing the equation for calculation of the vegetation cover fraction (F_C); and (4) adding a calculation of soil evaporation to the previously proposed RS-PM method.We evaluate our algorithm using ET observations at 19 AmeriFlux eddy covariance flux towers. We calculated ET with both our Revised RS-PM algorithm and the RS-PM algorithm using Global Modeling and Assimilation Office (GMAO v. 4.0.0) meteorological data and compared the resulting ET estimates with observations. Results indicate that our Revised RS-PM algorithm substantially reduces the root mean square error (RMSE) of the 8-day latent heat flux (LE) averaged over the 19 towers from 64.6 W/m² (RS-PM algorithm) to 27.3 W/m² (Revised RS-PM) with tower meteorological data, and from 71.9 W/m² to 29.5 W/m² with GMAO meteorological data. The average LE bias of the tower-driven LE estimates to the LE observations changed from 39.9 W/m² to − 5.8 W/m² and from 48.2 W/m² to − 1.3 W/m² driven by GMAO data. The correlation coefficients increased slightly from 0.70 to 0.76 with the use of tower meteorological data. We then apply our Revised RS-PM algorithm to the globe using 0.05° MODIS remote sensing data and reanalysis meteorological data to obtain the annual global ET (MODIS ET) for 2001. As expected, the spatial pattern of the MODIS ET agrees well with that of the MODIS global terrestrial gross and net primary production (MOD17 GPP/NPP), with the highest ET over tropical forests and the lowest ET values in dry areas with short growing seasons. This MODIS ET product provides critical information on the regional and global water cycle and resulting environment changes.