Moderate Resolution Imaging Spectroradiometer Normalized Difference Vegetation Index Research Articles

Development of long‐term spatiotemporal continuous NDVI products for alpine grassland from 1982 to 2020 in the Qinghai–Tibet Plateau, China

AbstractBackgroundThe time‐series data of the Normalized Difference Vegetation Index (NDVI) is a crucial indicator for global and regional vegetation monitoring. However, the current assessment of global and regional long‐term vegetation changes is subject to large uncertainties due to the lack of spatiotemporally continuous time‐series data sets.MethodsIn this study, a long time‐series monthly NDVI data set with a spatial resolution of 250 m from 1982 to 2020 was developed by combining Moderate Resolution Imaging Spectroradiometer (MODIS) and AVHRR (Advanced Very High‐Resolution Radiometer) time‐series NDVI products using the Random Forest (RF) downscaling model.ResultsCompared to the MODIS NDVI product, the fused product shows RMSE and mean absolute error ranging from 0 to 0.075 and from 0 to 0.05, respectively, with R2 values mostly above 0.7.ConclusionsThe long time‐series NDVI products generated in this study are reliable in terms of accuracy and have great potential for long‐term dynamic monitoring of terrestrial ecosystems on the Qinghai–Tibet Plateau.

Mapping and classification of Liao River Delta coastal wetland based on time series and multi-source GaoFen images using stacking ensemble model

The precise mapping of coastal wetlands holds great significance for monitoring carbon sequestration and storage within coastal ecosystems, particularly in light of climate change and human-induced activities. Time series and multi-source remote sensing data offer distinct advantages in spatial and temporal land use mapping, particularly in wetland systems encompassing various vegetation types. The primary aim of this study was to delineate the spatial distribution of land use within the Liao River Delta (LRD) wetland. This was achieved by employing a stacking ensemble model that integrates time-series GaoFen-1 (GF-1) optical imagery, GaoFen-3 (GF-3) synthetic aperture radar (SAR) imagery, and Moderate Resolution Imaging Spectroradiometer (MODIS) Normalized Difference Vegetation Index (NDVI) imagery. The first step involved the application of an enhanced spatial and temporal adaptive reflectance fusion model (ESTARFM) to fuse the GF-1 NDVI and MODIS NDVI datasets, producing time-series NDVI data. Subsequently, the parameters pertaining to vegetation phenology were obtained by employing the threshold method on the time-series NDVI data. We compiled feature datasets encompassing GF-1 spectral bands, indices, phenological parameters, and GF-3 SAR features. A Recursive Feature Elimination and Cross-Validation (RFECV) model was employed to identify and select significant features to mitigate data redundancy. Finally, a stacking ensemble model was constructed by combining five base models [K-Nearest Neighbors (KNN), Random Forest (RF), Adaptive Boosting (AdaBoost), Extreme Gradient Boosting (XGBoost), and a Light Gradient Boosting Machine (LightGBM)] to perform wetland classification. The findings were as follows: (1) ESTARFM was able to successfully fuse GF-1 NDVI and MODIS NDVI data, resulting in spatiotemporal fusion with a coefficient of determination (R2) of 0.85 and a root mean square error (RMSE) of 0.07. (2) the RFECV algorithm was employed to select relevant features in the wetland classification process. Specifically, 75 spectral band features, 89 spectral index features, 13 SAR features, and seven phenological parameters were identified as significant for this task. (3) A stacking ensemble model was constructed using the aforementioned multi-source features. This model exhibited a robust and consistent performance in wetland classification, achieving the highest overall accuracy of 94.33%. Notably, this accuracy improvement ranged from approximately 0.09% to 10.02% compared to the individual base models. Thus, the present study has the potential to be utilized for fine-scale wetland monitoring, thereby offering valuable assistance in the field of wetland environmental research.

Spatial-temporal dynamics of paddy productivity on the north coast of Java Island, Indonesia based on the principal component analysis of MODIS NDVI anomaly data

Comprehending the dynamics of paddy productivity is imperative for enhancing the efficacy of agricultural land developments. This study provides the application of principal component analysis (PCA) as a method for visualizing the spatial-temporal changes in paddy productivity. The analysis is conducted using the 8-day NDVI (normalized difference vegetation index) anomaly data of MODIS (Moderate Resolution Imaging Spectroradiometer) data spanning the period from 2000 to 2020. The regencies of Karawang, Subang, and Indramayu on the north coast of Java island are chosen as the study area because of their top rice production areas in Indonesia. The results show that the first leading PCA of the NDVI anomaly is related to the interannual variability of paddy productivity with 3-4 year cycles. The spatial and temporal dynamics of the first mode of eigenvectors and principal component time series can generally be grouped into nine categories. Two important categories to note are category-1 (1 January – 19 March) and category-8 (12 September – 16 December). In category-1, the NDVI anomalies move from north to middle and middle to north areas in Karawang and Subang regencies, respectively. In Indramayu Regency, the NDVI anomalies relatively remain in almost all areas. In contrast, in category-8 the NDVI anomalies move from the middle to northern areas in Karawang, Subang, and Indramayu regencies.

Evaluating the Effect of Vegetation Index Based on Multiple Tree-Ring Parameters in the Central Tianshan Mountains

Combining tree ring data with remote sensing data can help to gain a deeper understanding of the driving factors that influence vegetation change, identify climate events that lead to vegetation change, and improve the parameters of global vegetation index reconstruction models. However, it is currently not well understood how climate change at different elevations in the central Tianshan Mountains affects radial tree growth and the dynamics of forest canopy growth. We selected Schrenk spruce (Picea schrenkiana) tree core samples from different elevations in the central Tianshan Mountains. We analyzed the relationships of various tree-ring parameters, including tree-ring width, maximum latewood density (MXD), and minimum earlywood density (MID) chronologies, with 1982–2012 GIMMS (Global Inventory Modelling and Mapping Studies) NDVI (Normalized Difference Vegetation Index), 2001–2012 MODIS (moderate resolution imaging spectroradiometer) NDVI, and meteorological data. (1) There were strong correlations between tree-ring width chronologies and the lowest temperatures, especially in July. Tree-ring width chronologies at higher altitudes were positively correlated with temperature; the opposite pattern was observed at lower altitudes. MID chronologies were positively correlated with July temperature in high-altitude areas and mean temperature and highest temperature from May to September in low-altitude areas, and negatively correlated with precipitation during this period. MXD chronologies were mainly negatively correlated with precipitation. MXD chronologies were mainly positively correlated with temperature in April and May. (2) The correlations between MXD chronologies at each sampling point and NDVI in each month of the growing season were strong. Both MID and MXD chronologies were negatively correlated with GIMMS NDVI in July. The overall correlations between tree-ring parameters and MODIS NDVI were stronger than the correlations between tree-ring parameters and GIMMS NDVI in high-altitude areas; the opposite pattern was observed in low-altitude areas. Drought stress may be the main factor affecting tree ring parameters and NDVI. In the future, we should combine tree ring parameters with vegetation index to investigate a larger scale of forests.

Spatiotemporally consistent global dataset of the GIMMS Normalized Difference Vegetation Index (PKU GIMMS NDVI) from 1982 to 2022

Abstract. Global products of remote sensing Normalized Difference Vegetation Index (NDVI) are critical to assessing the vegetation dynamic and its impacts and feedbacks on climate change from local to global scales. The previous versions of the Global Inventory Modeling and Mapping Studies (GIMMS) NDVI product derived from the Advanced Very High Resolution Radiometer (AVHRR) provide global biweekly NDVI data starting from the 1980s, being a reliable long-term NDVI time series that has been widely applied in Earth and environmental sciences. However, the GIMMS NDVI products have several limitations (e.g., orbital drift and sensor degradation) and cannot provide continuous data for the future. In this study, we presented a machine learning model that employed massive high-quality global Landsat NDVI samples and a data consolidation method to generate a new version of the GIMMS NDVI product, i.e., PKU GIMMS NDVI (1982–2022), based on AVHRR and Moderate-Resolution Imaging Spectroradiometer (MODIS) data. A total of 3.6 million Landsat NDVI samples that were well spread across the globe were extracted for vegetation biomes in all seasons. The PKU GIMMS NDVI exhibits higher accuracy than its predecessor (GIMMS NDVI3g) in terms of R2 (0.97 over 0.94), root mean squared error (RMSE: 0.05 over 0.09), mean absolute error (MAE: 0.03 over 0.07), and mean absolute percentage error (MAPE: 9 % over 20 %). Notably, PKU GIMMS NDVI effectively eliminates the evident orbital drift and sensor degradation effects in tropical areas. The consolidated PKU GIMMS NDVI has a high consistency with MODIS NDVI in terms of pixel value (R2 = 0.956, RMSE = 0.048, MAE = 0.034, and MAPE = 6.0 %) and global vegetation trend (0.9×10-3 yr−1). The PKU GIMMS NDVI product can potentially provide a more solid data basis for global change studies. The theoretical framework that employs Landsat data samples can facilitate the generation of remote sensing products for other land surface parameters. The PKU GIMMS NDVI product is open access and available under a Creative Commons Attribution 4.0 License at https://doi.org/10.5281/zenodo.8253971 (Li et al., 2023).

Exploring the Spatiotemporal Dynamics of CO2 Emissions through a Combination of Nighttime Light and MODIS NDVI Data

Climate change caused by CO2 emissions is posing a huge challenge to human survival, and it is crucial to precisely understand the spatial and temporal patterns and driving forces of CO2 emissions in real time. However, the available CO2 emission data are usually converted from fossil fuel combustion, which cannot capture spatial differences. Nighttime light (NTL) data can reveal human activities in detail and constitute the shortage of statistical data. Although NTL can be used as an indirect representation of CO2 emissions, NTL data have limited utility. Therefore, it is necessary to develop a model that can capture spatiotemporal variations in CO2 emissions at a fine scale. In this paper, we used the nighttime light and the Moderate Resolution Imaging Spectroradiometer (MODIS) normalized difference vegetation index (NDVI), and proposed a normalized urban index based on combination variables (NUI-CV) to improve estimated CO2 emissions. Based on this index, we used the Theil–Sen and Mann–Kendall trend analysis, standard deviational ellipse, and a spatial economics model to explore the spatial and temporal dynamics and influencing factors of CO2 emissions over the period of 2000–2020. The experimental results indicate the following: (1) NUI-CV is more suitable than NTL for estimating the CO2 emissions with a 6% increase in average R2. (2) The center of China’s CO2 emissions lies in the eastern regions and is gradually moving west. (3) Changes in industrial structure can strongly influence changes in CO2 emissions, the tertiary sector playing an important role in carbon reduction.

An improved change detection method for tacking remote sensing time series trends

ABSTRACT To improve the accuracy of detecting changes in remote sensing time series, an improved algorithm based on the combination of the antileakage least-squares spectral analysis (ALLSSA) algorithm and detecting breakpoints and estimating segments in trends (DBEST) algorithm is proposed and applied. The method uses the ALLSSA algorithm to decompose the time series and identify the trend components in the time series. Then, the trend segmentation mechanism of the DBEST algorithm is used to detect the changes in the trend component. In this paper, the improved algorithm is evaluated using a simulated time series data set, a time series data set with multiple change points, and data set based on the moderate resolution imaging spectroradiometer (MODIS) normalized difference vegetation index (NDVI) remote sensing time series. The results demonstrate that the average detection accuracies of the improved algorithm and DBEST algorithm are 98.4% and 85.2%, respectively, for the simulated time series data set. For the time series data set with multiple change points, the average root mean square errors (RMSEs) of the trend data for the improved and DBEST algorithms are 0.0386 and 0.0331, respectively. The mean normalized residual norms (MNRNs) of the improved and DBEST algorithms are 0.0252 and 0.0351, respectively. Finally, the improved algorithm, DBEST algorithm, and breaks for additive season and trend (BFAST) algorithm are applied to MODIS NDVI data, and their performance with remote sensing data is compared. The improved algorithm has higher detection accuracy and a smaller MNRN, indicating that more information is included in the trend and seasonal components. Therefore, the proposed method is useful for analysing trends in remote sensing time series data.

The response and sensitivity of global vegetation to water stress: A comparison of different satellite-based NDVI products

Normalized Difference Vegetation Index (NDVI) is an effective proxy for vegetation activity and terrestrial productivity, and the widely used global NDVI products include the third generation Global Inventory Modelling and Mapping Studies Advanced Very High-Resolution Radiometer NDVI (NDVI3g), Satellite Pour l’Observation de la Terre VEGETATION NDVI (NDVIsp) and Terra Moderate Resolution Imaging Spectroradiometer NDVI (NDVItr). Here, for the first time, we investigated the differences between the three NDVIs in response and sensitivity to water stress represented by root zone soil moisture, vapor pressure deficit, self-calibrating Palmer Drought Severity Index and Standardized Precipitation Evapotranspiration Index across climate zones, humidity gradients, vegetation types and tree cover gradients. Over the temperate and arid climates region, the three NDVIs show higher consistent response fractions (CRFs) to water stress in the water-limited regions, with their sensitivity being in the order of NDVI3g < NDVIsp < NDVItr. The three NDVIs tend to exhibit inconsistent responses to water stress in the areas with higher humidity and tree cover, and their sensitivities to water stress fall as humidity and tree cover increase. In addition, the three NDVIs display higher CRFs and sensitivities to water stress over the non-forested areas. Our results also show that, in the cold climates region, the three NDVIs respond in the opposite way to water stress as that in water-limited areas, and NDVI3g has higher sensitivity to water stress than NDVIsp and NDVItr. This study highlights the differences between three NDVIs products for qualifying and quantifying their response and sensitivity to water stress.

Swidden Agriculture Landscape Mapping Using MODIS Vegetation Index Time Series and Its Spatio-Temporal Dynamics in Northern Laos

Swidden agriculture or shifting cultivation is still being widely practiced in tropical developing countries and Laos has spared no effort to eradicate it since the mid-1990s. So far, the development of swidden agriculture in this land-locked mountainous country during the 2000–2020 bi-decade remains poorly examined. Moderate-resolution Imaging Spectroradiometer (MODIS) time-series products have shown potential in monitoring vegetative status; however, only extremely limited cases of remote sensing of swidden agriculture landscapes have been reported. Taking northern Laos as a study area and using 2001–2020 MODIS vegetation indices products, the Savitzky–Golay filter, the Mann–Kendall trend test and a threshold method were employed to delineate and monitor annual patterns and dynamics of swidden agriculture landscape at the village level. The results showed that: MODIS Normalized Difference Vegetation Index (NDVI) time series perform better in delineating the temporal development of swidden agriculture. The swidden agriculture landscape has shown a general descending trend in the past decades, especially in the 2010s, with an annual average of 14.70 × 104 ha. The total number of swidden-practicing villages (or districts) also displayed a declining trend and there were 957 villages or 91 districts practicing it continuously between 2001 and 2020. An average of 32 villages per year or two districts per decade highlights the difficulty in ending swidden agriculture in Laos, although the government of Laos has established a number of policies for the eradication of swidden agriculture by 2020. This study provides a necessary methodological reference for monitoring a two-decade evolution and transformation of swidden agriculture in the tropics.

Linear and Non-Linear Vegetation Trend Analysis throughout Iran Using Two Decades of MODIS NDVI Imagery

Vegetation is the main component of the terrestrial Earth, and it plays an imperative role in carbon cycle regulation and surface water/energy exchange/balance. The coupled effects of climate change and anthropogenic forcing have undoubtfully impacted the vegetation cover in linear/non-linear manners. Considering the essential benefits of vegetation to the environment, it is vital to investigate the vegetation dynamics through spatially and temporally consistent workflows. In this regard, remote sensing, especially Normalized Difference Vegetation Index (NDVI), has offered a reliable data source for vegetation monitoring and trend analysis. In this paper, two decades (2000 to 2020) of Moderate Resolution Imaging Spectroradiometer (MODIS) NDVI datasets (MOD13Q1) were used for vegetation trend analysis throughout Iran. First, the per-pixel annual NDVI dataset was prepared using the Google Earth Engine (GEE) by averaging all available NDVI values within the growing season and was then fed into the PolyTrend algorithm for linear/non-linear trend identification. In total, nearly 14 million pixels (44% of Iran) were subjected to trend analysis, and the results indicated a higher rate of greening than browning across the country. Regarding the trend types, linear was the dominant trend type with 14%, followed by concealed (11%), cubic (8%), and quadratic (2%), while 9% of the vegetation area remained stable (no trend). Both positive and negative directions were observed in all trend types, with the slope magnitudes ranging between −0.048 and 0.047 (NDVI units) per year. Later, precipitation and land cover datasets were employed to further investigate the vegetation dynamics. The correlation coefficient between precipitation and vegetation (NDVI) was 0.54 based on all corresponding observations (n = 1785). The comparison between vegetation and precipitation trends revealed matched trend directions in 60% of cases, suggesting the potential impact of precipitation dynamics on vegetation covers. Further incorporation of land cover data showed that grassland areas experienced significant dynamics with the highest proportion compared to other vegetation land cover types. Moreover, forest and cropland had the highest positive and negative trend direction proportions. Finally, independent (from trend analysis) sources were used to examine the vegetation dynamics (greening/browning) from other perspectives, confirming Iran’s greening process and agreeing with the trend analysis results. It is believed that the results could support achieving Sustainable Development Goals (SDGs) by serving as an initial stage study for establishing conservation and restoration practices.

Photosynthesis phenology, as defined by solar-induced chlorophyll fluorescence, is overestimated by vegetation indices in the extratropical Northern Hemisphere

Vegetation phenology is highly sensitive to climate change, although the data and methods used to estimate key phenological states can influence this sensitivity. Because of its direct relation to leaf photosynthetic carbon uptake, remotely sensed solar-induced chlorophyll fluorescence (SIF) can provide new insight assessing changes in vegetation phenology. Here, we investigated the potential of using a SIF time series product named contiguous SIF (CSIF) to estimate spring, summer, and autumn phenology in the extratropical Northern Hemisphere (>30°N) and compared the results with those based on Moderate Resolution Imaging Spectroradiometer (MODIS) Normalized Difference Vegetation Index (NDVI) for the period 2001–2018. Overall, we found similar spatial patterns in phenological states. However, specific dates of key phenological events differed when using CSIF vs. MODIS NDVI data. NDVI data indicated that the growing season started earlier (by 10.1 days on average) and ended later (11.5 days on average) relative to CSIF data. This implies that actual periods for photosynthetic activity are shorter (by 21.6 days on average) than those estimated from vegetation indices more directly related to changes in canopy structure. These large differences between results from NDVI and that from CSIF suggest that vegetation indices such as NDVI seem to overestimate the period for active photosynthesis over the extratropical Northern Hemisphere. Furthermore, while phenology of the early growing season is dominated by temperature for both NDVI and CSIF data, phenology of the late growing season is mainly controlled by temperature for NDVI but by precipitation for CSIF. Our findings were further confirmed by other SIF (GOME-2 SIF) and vegetation index (MODIS EVI) datasets. Phenology modes in Earth system modelling are often parameterized using leaf unfolding and senescence from either station or satellite observations. Our results imply that canopy structure-based parameterization schemes may have overestimated photosynthesis active period, and thus productivity responses. We conclude that SIF data offers a novel and unique approach for assessing phenological change - one that is more directly tied to the carbon cycle and how it is being influenced by climate change.

A Global 250-m Downscaled NDVI Product from 1982 to 2018

Satellite-based normalized difference vegetation index (NDVI) time series data are useful for monitoring the changes in vegetation ecosystems in the context of global climate change. However, most of the current NDVI products cannot effectively reconcile high spatial resolution and continuous observations in time. Here, to produce a global-scale, long-term, and high-resolution NDVI database, we developed a simple and new data downscaling approach. The downscaling algorithm considers the pixel-wise ratios of the coefficient of variation (CV) between the coarse- and fine-resolution NDVI data and relative changes in the NDVI against a baseline period. The algorithm successfully created a worldwide monthly NDVI database with 250 m resolution from 1982 to 2018 by translating the fine spatial information from MODIS (Moderate-resolution Imaging Spectroradiometer) data and the long-term temporal information from AVHRR (Advanced Very High Resolution Radiometer) data. We employed the evaluation indices of root mean square error (RMSE), mean absolute error (MAE), and Pearson’s correlation coefficient (Pearson’s R) to assess the accuracy of the downscaled data against the MODIS NDVI. Both the RMSE and MAE values at the regional and global scales are typically between 0 and 0.2, whereas the Pearson’s R values are mostly above 0.7, which implies that the downscaled NDVI product is similar to the MODIS NDVI product. We then used the downscaled data to monitor the NDVI changes in different plant types and places with significant vegetation heterogeneity, as well as to investigate global vegetation trends over the last four decades. The Google Earth Engine platform was used for all the data downscaling processes, and here we provide a code for users to easily acquire data corresponding to any part of the world. The downscaled global-scale NDVI time series has high potential for the monitoring of the long-term temporal and spatial dynamics of terrestrial ecosystems under changing environments.

Reconstructing High-Spatiotemporal-Resolution (30 m and 8-Days) NDVI Time-Series Data for the Qinghai–Tibetan Plateau from 2000–2020

As the largest and highest alpine ecoregion in the world, the Qinghai–Tibetan Plateau (QTP) is extremely sensitive to climate change and has experienced extraordinary warming during the past several decades; this has greatly affected various ecosystem processes in this region such as vegetation production and phenological change. Therefore, numerous studies have investigated changes in vegetation dynamics on the QTP using the satellite-derived normalized-difference vegetation index (NDVI) time-series data provided by the Moderate-Resolution Imaging Spectroradiometer (MODIS). However, the highest spatial resolution of only 250 m for the MODIS NDVI product cannot meet the requirement of vegetation monitoring in heterogeneous topographic areas. In this study, therefore, we generated an 8-day and 30 m resolution NDVI dataset from 2000 to 2020 for the QTP through the fusion of 30 m Landsat and 250 m MODIS NDVI time-series data. This dataset, referred to as QTP-NDVI30, was reconstructed by employing all available Landsat 5/7/8 images (&gt;100,000 scenes) and using our recently developed gap-filling and Savitzky–Golay filtering (GF-SG) method. We improved the original GF-SG approach by incorporating a module to process snow contamination when applied to the QTP. QTP-NDVI30 was carefully evaluated in both quantitative assessments and visual inspections. Compared with reference Landsat images during the growing season in 100 randomly selected subregions across the QTP, the reconstructed 30 m NDVI images have an average mean absolute error (MAE) of 0.022 and a spatial structure similarity (SSIM) above 0.094. We compared QTP-NDVI30 with upscaled cloud-free PlanetScope images in some topographic areas and observed consistent spatial variations in NDVI between them (averaged SSIM = 0.874). We further examined an application of QTP-NDVI30 to detect vegetation green-up dates (GUDs) and found that QTP-NDVI30-derived GUD data show general agreement in spatial patterns with the 250 m MODIS GUD data, but provide richer spatial details (e.g., GUD variations at the subpixel scale). QTP-NDVI30 provides an opportunity to monitor vegetation and investigate land-surface processes in the QTP region at fine spatiotemporal scales.

Estimation of fractional vegetation cover dynamics and its drivers based on multi-sensor data in Dera Ismail Khan, Pakistan

Time series study about the changes in Fractional Vegetation Cover (FVC) with its drivers is studied. From 2003 to 2013, FVC was calculated using remotely sensed data from the Moderate Resolution Imaging Spectroradiometer-Normalized Difference Vegetation Index (MODIS-NDVI), Landsat-7, Defense Meteorological Satellite Program/Operational Linescan System (DMSP/OLS), and Compound Night Light Index (CNLI). The mean value of FVC is 0.34 (34 %). The findings indicate that the highest annual mean values are between 34% and 37%, and 31% and 33%, that correspond to levelof 70% and 40%, respectively. The trend analysis by MODIS and Landsat-7 has the low (10%) FVC in the northern parts due to increase in industrial development, urbanization, changes in landuse, landcover and other linked environmental factors. The southwestern-southeastern and eastern areas showed medium (40 %) and high (70 %) FVC respectively. This is because of water availability and less human disturbance.

SPATIAL DOWNSCALING OF SMAP SOIL MOISTURE USING THE MODIS AND SRTM OBSERVATIONS

Abstract. The main objective of this study is the spatial downscaling of Soil Moisture Active Passive (SMAP) soil moisture (36 km) using the Moderate Resolution Imaging Spectroradiometer (MODIS) and Shuttle Radar Topography Mission (SRTM) products. The study was conducted over India during the post-monsoon (i.e., Rabi) season Daily SMAP soil moisture (SM) data was composited to 3 days to cover the entire study area. MODIS data for the Normalized Difference Vegetation Index (NDVI), Normalized Difference Water Index (NDWI), Albedo, and Land Surface Temperature (LST) were similarly obtained by constructing a three day composite. SMAP soil moisture was used as a dependent variable, whereas, MODIS NDVI, NDWI, Albedo, LST, and SRTM elevation were used as independent variables in a regression analysis for downscaling of SMAP soil moisture. The coefficient of determination (R2) was used to evaluate the performance of multi-variate linear regression (MLR), support vector regression (SVR), and random forest regression (RFR). Each method was used to test the performance of monthly and seasonal models. RFR outperformed MLR and SVR for monthly and seasonal models. Furthermore, a comparison of monthly and seasonal models revealed that the model created on Jan. data performed best (R2=0.80), while R2 of 0.73, 0.61, 0.75, and 0.76 were attained using RFR for seasonal, Dec., Feb., and Mar. models, respectively. In addition, in-situ soil moisture data was used to validate downscaled soil moisture (1 km). Comparison between downscaled soil moisture and in-situ soil moisture showed good agreement with a difference ranging between −9.3 to 7.4 %.

ESTIMATION OF NDVI FOR CLOUDY PIXELS USING MACHINE LEARNING

Abstract. The Normalized Difference Vegetation Index (NDVI) is a useful index for vegetation monitoring. However, due to cloud cover the observations of NDVI are discrete and vary in the intensity. Therefore, there is a need to estimate the NDVI during cloud cover using alternative sources of satellite observations. The main objective of this study is to estimate NDVI during cloudy conditions using moderate resolution multi-spectral and synthetic aperture radar (SAR) observations. Two approaches were identified: 1) pixel replacement and 2) machine learning based regression analysis to estimate cloud free NDVI. Moderate Resolution Imaging Spectroradiometer (MODIS) 8-day NDVI composite, Sentinel-1 SAR and cloud masked Sentinel-2 multi-spectral observations were collected for entire cropping season. The satellite observations were selected only for agricultural areas by applying the agriculture, non-agriculture land use land cover mask. Machine learning algorithms such as Linear Regression (LR), Random Forest Regression (RFR), and Support Vector Regression (SVR) were used for NDVI estimation. Regression analysis was performed using Sentinel-2 NDVI as an independent variable and VV, VH, Cross Ratio (i.e., VV/VH), and MODIS NDVI as dependent variables. NDVI of the cloudy pixel was estimated using the trained regression models over the agriculture areas. A regression model was trained and applied to each Sentinel-2 tile that covers an area of 100 km × 100 km. The RFR and SVR showed the highest R2 of 0.73 and a RMSE of 0.12. A visual comparison of time series graphs showed good alignment between actual (Sentinel-2) and predicted NDVI and usual crop growth trend.

Improved identification of cotton cultivated areas by applying instance-based transfer learning on the time series of MODIS NDVI

Cotton is an important cash crop and strategic material in the world, as the main source of natural and renewable fiber for textiles. Accurate and timely cotton distribution maps are crucial for monitoring and managing cotton cultivation system. Remotely sensed data have been widely used in cropland mapping, whereas relatively less attention has been paid specifically to cotton mapping partly due to the difficulty in obtaining the training samples over large regions. To resolve this issue, this study introduced the instance-based transfer learning to identify the cotton cultivated areas using the remotely sensed images. The annual time series of normalized difference vegetation index (NDVI) derived from Moderate Resolution Imaging Spectroradiometer (MODIS) images was adopted as input features. The random forest (RF) and long short-term memory (LSTM) algorithms integrated with the Transfer AdaBoost (TrAdaBoost) algorithm were adopted to generate the RF-based and LSTM-based TrAdaBoost approaches, respectively. The experiments were conducted in Arkansas State of the United States, where the cropland data layer (CDL) was available and utilized as a source of auxiliary data. The results showed that both the RF-based and LSTM-based TrAdaBoost performed better than the original RF and LSTM under the condition that the training samples in the target domain were limited. The advantages of transfer learning were much greater when the percentage of training samples from the source domain equals 80%. The two transfer learning approaches were then applied to identify the cotton cultivated areas in Uzbekistan. The cotton areas detected by MODIS images in 2018 were agreed well with the statistics at the sub-national level, with the R2 values of 0.57 and 0.64, respectively. These results demonstrate the potential of the RF-based and LSTM-based TrAdaBoost approaches in generating the cotton distribution maps when there are few samples in the target domain.

Estimation of fractional vegetation cover dynamics based on satellite remote sensing in pakistan: A comprehensive study on the FVC and its drivers

The present investigation exposes the significant variations of Fractional Vegetation Cover (FVC) and drivers from year 2003 to 2013 in Pakistan. It is directly calculated by remote-sensing data using Moderate Resolution Imaging Spectroradiometer-Normalized Difference Vegetation Index (MODIS-NDVI) and Compound Night Light Index/Defense Meteorological Program/Operational Line-Scan System (CNLI/DMSP/OLS). The spatial patterns of FVC variation are mainly categorized into three levels, low (<10%), medium (40%) and high (70%), with consequent findings. The FVC time series analysis exhibited the fitting curve with straight lines at 0.29 (29%). These findings displayed the utmost annual mean values 26% to 31% and 28% and 29% correspond to level 70% and 40%, respectively. Among all levels, level 10% has the lowest mean. The trend analysis displayed the low FVC for the southwest and south-eastern regions. This is attributed to industrialization, urbanization, land use, land cover change, and related climatic factors: the northeast and the northwest areas displayed medium and high FVC due to less human disturbance. Using different remotely sensed data, human activities (industrialization and urbanization, etc.) and climatic factors such as rainfall and temperature are considered driving factors of FVC dynamics. Finally, the correlation value of the coefficient verified the link between climatic factors, FVC and CNLI. These findings present the positive correlation of FVC with rainfall and the negative with compounded night light index and temperature.

Spatiotemporal patterns of maize drought stress and their effects on biomass in the Northeast and North China Plain from 2000 to 2019

Droughts are characterized by limited water resources and extreme arid climates and have a notable influence on crop yield and food security. Therefore, evaluating the crop drought stress at a large spatiotemporal scale is important. In this study, maize planting areas in the Northeast China Plain (NECP) and North China Plain (NCP) over the past 20 years (2000–2019) were identified from the moderate resolution imaging spectroradiometer (MODIS) enhanced vegetation index, comprising 8-day time-series moving windows. The temperature vegetation drought index (TVDI) was retrieved using the MODIS normalized difference vegetation index (NDVI) and land surface temperature datasets to evaluate maize drought. Furthermore, by combining MODIS NDVI with measured meteorological data, monthly net primary productivity (NPP) was retrieved based on the Carnegie–Ames–Stanford approach (CASA) to characterize the maize biomass. Thus, the weight of drought stress in different growth stages of maize was obtained by combining the maize phenological information with TVDI and NPP. Moreover, TVDIdrought stress was obtained to comprehensively assess maize drought stress. Finally, we fitted TVDIdrought stress with NPP to establish a conceptual model for the response of maize biomass to drought stress in the NCP and NECP at a large spatiotemporal scale for the first time. The conceptual model revealed that the phenomenon of excessive moisture during maize growth was not evident (TVDIdrought stress < 0.375) and that most maize plants were still affected by drought stress (TVDIdrought stress > 0.375) in the NECP. Correspondingly, in the NCP, there was hardly any excessive moisture, and maize exhibited a certain resistance to low-intensity drought stress (TVDIdrought stress was approximately 0.330). However, after the drought stress exceeded the tolerance range (TVDIdrought stress > 0.375), the maize biomass was significantly affected by drought stress. Our findings illustrate the two response models of drought stress in the major maize producing areas of China for the first time, thereby providing new insights for preventing drought stress risk and improving maize yield on a large scale.

Mapping Winter Wheat with Optical and SAR Images Based on Google Earth Engine in Henan Province, China

The timely and accurate acquisition of winter wheat acreage is crucial for food security. This study investigated the feasibility of extracting the spatial distribution map of winter wheat in Henan Province by using synthetic aperture radar (SAR, Sentinel-1A) and optical (Sentinel-2) images. Firstly, the SAR images were aggregated based on the growth period of winter wheat, and the optical images were aggregated based on the moderate resolution imaging spectroradiometer normalized difference vegetation index (MODIS-NDVI) curve. Then, five spectral features, two polarization features, and four texture features were selected as feature variables. Finally, the Google Earth Engine (GEE) cloud platform was employed to extract winter wheat acreage through the random forest (RF) algorithm. The results show that: (1) aggregated images based on the growth period of winter wheat and sensor characteristics can improve the mapping accuracy and efficiency; (2) the extraction accuracy of using only SAR images was improved with the accumulation of growth period. The extraction accuracy of using the SAR images in the full growth period reached 80.1%; and (3) the identification effect of integrated images was relatively good, which makes up for the shortcomings of SAR and optical images and improves the extraction accuracy of winter wheat.