NDVI Model Research Articles

Mapping Field-Level Maize Yields in Ethiopian Smallholder Systems Using Sentinel-2 Imagery

Remote sensing offers a low-cost method for estimating yields at large spatio-temporal scales. Here, we examined the ability of Sentinel-2 satellite imagery to map field-level maize yields across smallholder farms in two regions in Oromia district, Ethiopia. We evaluated how effectively different indices, the MTCI, GCVI, and NDVI, and different models, linear regression and random forest regression, can be used to map field-level yields. We also examined if models improved by adding weather and soil data and how generalizable our models were if trained in one region and applied to another region, where no data were used for model calibration. We found that random forest regression models that used monthly MTCI composites led to the highest yield prediction accuracies (R2 up to 0.63), particularly when using only localized data for training the model. These models were not very generalizable, especially when applied to regions that had significant haze remaining in the imagery. We also found that adding soil and weather data did little to improve model fit. Our results highlight the ability of Sentinel-2 imagery to map field-level yields in smallholder systems, though accuracies are limited in regions with high cloud cover and haze.

Dynamics of land surface temperature: Insights into vegetation, elevation, and air pollution in Bengaluru

Bengaluru, among the five fastest-growing cities of India, is facing the challenge of urban heat islands (UHI) and temperature variations across its landscape. Monitoring the temperature fluctuations is crucial for understanding the city's microclimate and the impact of urbanization on its environment. Land Surface Temperature (LST) is a significant metric for this purpose. The present study examined the relationship of LST with vegetation, elevation, and air pollutants during the summer and winter seasons for 4-years (2019–2022). Air pollutants and weather parameters data from air quality monitoring stations was used. The study found negative relationship of LST with NDVI and Digital Elevation Model (DEM) during both seasons. The relationship between LST and individual air pollutants was variable during the summer and winter seasons and R2 ranged between 0.52 and 0.79. The relationship became stronger (R2 between 0.69 and 0.82 during summer and R2 between 0.74 and 0.80 during winter season) when all the parameters were considered simultaneously. The result from the Taguchi method showed that different air pollutants had variable impact over change in LST and the order of their effect was found to be PM2.5 > SO2 > O3 > CO > NO2. The results indicate that O3 (summer) and PM2.5 (winter) were the major contributors to rising LST. The present study has applications in future scientific research as well as it can be utilized to integrate mitigation strategies for increasing temperatures.

Assessing the accuracy of spectral indices obtained from Sentinel images using field research to estimate land degradation.

Wind erosion resulting from soil degradation is a significant problem in Iran's Baluchistan region. This study evaluated the accuracy of remote sensing models in assessing degradation severity through field studies. Sentinel-2 Multispectral Imager's (MSI) Level-1C satellite data was used to map Rutak's degradation severity in Saravan. The relationship between surface albedo and spectral indices (NDVI, SAVI, MSAVI, BSI, TGSI) was assessed. Linear regression establishes correlations between the albedo and each index, producing a degradation severity map categorized into five classes based on albedo and spectral indices. Accuracy was tested with 100 ground control points and field observations. The Mann-Whitney U-Test compares remote sensing models with field data. Results showed no significant difference (P > 0.05) between NDVI, SAVI, and MSAVI models with field data, while BSI and TGSI models exhibited significant differences (P ≤ 0.001). The best model, BSI-NDVI, achieves a regression coefficient of 0.86. This study demonstrates the advantage of remote sensing technology for mapping and monitoring degraded areas, providing valuable insights into land degradation assessment in Baluchistan. By accurately identifying severity levels, informed interventions can be implemented to mitigate wind erosion and combat soil degradation in the region.

Analisis Indeks Vegetasi dan Perancangan Sistem Monitoring Kawasan Ekowisata Mangrove Berbasis Hyper Spectral of Remote Sensing dan Design Thinking Framework

One of the challenges in decision-making for determining the direction of mangrove ecotourism development policies and programs on Tagalaya Island is the availability of mangrove databases at each observation station and remote sensing models used to identify mangrove forest density. This research offers the idea of combining the hyperspectral method of remote sensing with the information system design method design thinking framework. Meanwhile, the location of this research is the mangrove ecotourism area of Tagalaya Island, North Halmahera Regency, North Maluku Province, Indonesia. The findings of this study demonstrate the need to use remote sensing to locate and evaluate spatial data, particularly index values based on NDVI models for each Tagalaya Island observation station. The first and second observation stations are shown to be included in the unusual group by the computation results of the 2023 raster data in the mangrove area of Tagalaya Island. The third observation station is also included in the medium category. The monitoring information system for the Tagalaya mangrove ecotourism region is also a solution to the mangrove database challenge in deciding the direction of Tagalaya mangrove ecotourism policies and activities (SIMOKE Tagalaya). Thus, the problems and solutions encountered in enhancing the monitoring program of mangrove ecotourism areas on Tagalaya Island become relevant and contextual when applying the hyperspectral method of remote sensing and information system design using a design thinking framework.

Evaluation of the Use of the 12 Bands vs. NDVI from Sentinel-2 Images for Crop Identification.

Today, machine learning applied to remote sensing data is used for crop detection. This makes it possible to not only monitor crops but also to detect pests, a lack of irrigation, or other problems. For systems that require high accuracy in crop identification, a large amount of data is required to generate reliable models. The more plots of and data on crop evolution used over time, the more reliable the models. Here, a study has been carried out to analyse neural network models trained with the Sentinel satellite's 12 bands, compared to models that only use the NDVI, in order to choose the most suitable model in terms of the amount of storage, calculation time, accuracy, and precision. This study achieved a training time gain of 59.35% for NDVI models compared with 12-band models; however, models based on 12-band values are 1.96% more accurate than those trained with the NDVI alone when it comes to making predictions. The findings of this study could be of great interest to administrations, businesses, land managers, and researchers who use satellite image data mining techniques and wish to design an efficient system, particularly one with limited storage capacity and response times.

Analysis on the Rationality of Urban Green Space Distribution in Hangzhou City Based on GF-1 Data

With its ecological, economic, and social benefits, urban green space (UGS) plays an important role in urban planning. Accordingly, it is also an important indicator in the evaluation of urban liveability. However, the extraction and statistical analysis of UGS are difficult because urban land use involves complex types and UGS exhibits fragmented distribution and common vegetation extraction models such as the NDVI model and pixel bipartite model. In addition, there are few studies that analyze UGS in Hangzhou with a pixel decomposition model. Therefore, applying the mixed pixel decomposition model with GF-1 data, the following three objectives were set in this study: (1) analyzing the temporal changes of UGS in Shangcheng District, Hangzhou from 2018 to 2020; (2) analyzing the spatial distribution characteristics of UGS in the six main urban areas of Hangzhou in 2020; (3) analyzing the rationality and influencing factors of UGS distribution in Hangzhou. In Shangcheng District, the overall UGS area increased from 2018 to 2020 due to the increase in forest area rather than grassland area. As for the main built-up area in Hangzhou, medium and high coverage of UGS were primarily observed, with an overall high level of greening and a relatively uniform vegetation cover. Only a few areas showed very low UGS coverage. Some differences were observed among administrative regions under the influence of topography, but the overall coverage is high. At the same time, the distribution of UGS in Hangzhou is closely related to policy guidance, the needs of urban residents, and the requirements of economic development. This research not only can provide a new way to analyze UGS features in Hangzhou but also provides scientific guidance for governments in urban planning.

Estimation of agricultural burned affected area using NDVI and dNBR satellite-based empirical models

One of the major crucial issues that need worldwide attention is open stubble burning, which imposes a variety of adverse impacts on nature and human society, destroying the world's biodiversity. Many earth observation satellites render information to monitor and assess agricultural burning activities. In this study, different remotely sensed data (Sentinel-2A, VIIRS) has been employed to estimate the quantitative measurements of agricultural burned areas of the Purba Bardhaman district from October–December 2018. The multi-temporal image differencing techniques and indices (NDVI, NBR, and dNBR) and VIIRS active fires data (VNP14IMGT) have been utilized to spot agricultural burned areas. In the case of the NDVI technique, a prominent area, 184.82 km2 of agricultural burned area (7.85% of the total agriculture), was observed. The highest (23.04 km2) burned area was observed in the Bhatar block, located in the middle part of the district, and the lowest (0.11 km2) burned area was observed in the Purbasthali-II block, which is located in the eastern part of the district. On the other hand, the dNBR technique revealed that the agricultural burned areas enwrap 8.18% of the total agricultural area, which is 192.45 km2. As per the earlier NDVI technique, the highest agricultural burned areas (24.82 km2) were observed in the Bhatar block, and the lowest (0.13 km2) burn area occurred in the Purbashthali-II block. In both cases, it is observed that agricultural residue burning is high in the western part of the Satgachia block and the adjacent areas of the Bhatar block, which is in the middle part of Purba Bardhaman. The agricultural burned area was extracted using different spectral separability analyses, and the performance of dNBR was the most effective in spectral discrimination of burned and unburned surfaces. This study manifested that agricultural residue burning started in the central part of Purba Bardhaman. Later it spread all over the district due to the trend of early harvesting rice crops in this region. The performance of different indices for mapping the burned areas was evaluated and compared, revealing a strong correlation (R2) = 0.98. To estimate the campaign's effectiveness against the dangerous practice and plan the control of the menace, regular monitoring of crop stubble burning using satellite data is required.

Development of Hybrid Models to Estimate Gross Primary Productivity at a Near-Natural Peatland Using Sentinel 2 Data and a Light Use Efficiency Model

Peatlands store up to 2320 Mt of carbon (C) on only ~20% of the land area in Ireland; however, approximately 90% of this area has been drained and is emitting up to 10 Mt C per year. Gross primary productivity (GPP) is a one of the key components of the peatland carbon cycle, and detailed knowledge of the spatial and temporal extent of GPP under changing management practices is imperative to improve our predictions of peatland ecology and biogeochemistry. This research assesses the relationship between remote sensing and ground-based estimates of GPP for a near-natural peatland in Ireland using eddy covariance (EC) techniques and high-resolution Sen-tinel 2A satellite imagery. Hybrid models were developed using multiple linear regression along with six widely used conventional indices and a light use efficiency model. Estimates of GPP using NDVI, EVI, and NDWI2 hybrid models performed well using literature-based light use efficiency parameters and showed a significant correlation from 89 to 96% with EC-derived GPP. This study also reports additional site-specific light use efficiency parameters for dry and hydrologically normal years on the basis of light response curve methods (LRC). Overall, this research has demonstrated the potential of combining EC techniques with satellite-derived models to better understand and monitor key drivers and patterns of GPP for raised bog ecosystems under different climate scenarios and has also provided light use efficiency parameters values for dry and wetter conditions that can be used for the estimation of GPP using LUE models across various site and scales.

Establishment of a Monitoring Model for the Cotton Leaf Area Index Based on the Canopy Reflectance Spectrum

Cotton is the main economic crop in China and is important owing to its use as an industrial raw material and a cash crop. This experiment was conducted in the main cotton-producing area of Xinjiang, China. A hyperspectrometer was used to monitor the canopy spectral reflectance of cotton at different stages of growth. The results showed that the leaf area index (LAI) increased with the increase in the amount of nitrogen fertilizer added during the early full boll stage and decreased with the increase in nitrogen fertilization in the full and late boll stages. Insufficient or excessive fertilization led to a decrease in the LAI. The visible light band indicated that the canopy spectral reflectance decreased, and the amount of fertilizer increased in all the growth stages. The near-infrared band revealed that the canopy spectral reflectance increased with the amount of nitrogen applied during the bud stage, early boll stage, and the most vigorous period of boll growth. During the flowering period, the spectral reflectance followed the order N3 > N4 > N2 > N1 > N0. During the entire growth period of cotton, the values of the cotton LAI predicted using the ratio vegetation index (RVI) model were found to best fit the measured values. The LAI monitoring models of cotton in each growth stage were different. The TVI model is the best in the bud and early boll stages. The NDVI model is the best in the flowering stage, and the DVI model is the best in the full boll stage. This study provides a basis to accurately monitor the LAI in each growth period of cotton.

Assessment of soil erosion risks in response to land-use and land-cover changes in Coka watershed, Southern Ethiopia

ABSTRACT Soil erosion is a serious problem arising from agricultural intensification and other anthropogenic activities which can cause deterioration of soil fertility and productivity. This study aims to estimate total amounts of soil loss using the RUSLE and NDVI models and identify major areas of concern for management interventions through site specific in the Coka watershed, Southern Ethiopia. The RUSLE model required the integration of thematic factors such as rainfall, length and steepness of the slope, land-use and land-cover, soil erodibility, and control practices. These factors were calculated using remote sensing data and GIS. The results indicated that total area coverage from study of watershed under the degree of severity of erosion and priority was <5(very low), 5 − 10 (low), 10 − 18 (moderate), 18 − 50 (high), 50 − 100 (very high) and >100 t ha–1yr–1 (severe). Soil loss rates ranged between 2.33 in bare land and 237.16 t ha–1yr–1 in forest. Based on the findings, erosion risk having estimated soil loss (≥ 18 t ha–1yr–1) which is above tolerable range for the country containing 73% (2723 ha) from the total area coverage of watershed. Thus, appropriate soil and water conservation practices should be adopted to restore degraded lands and environment by giving attention for erosion hotspot areas.

Development of Modeler for Automated Mapping of Land Surface Temperature Using GIS and LANDSAT-8 Satellite Imagery

Land surface temperature (LST) can be described as the temperature of the earth’s surface and it is most important parameters in climate change, evapotranspiration, urban climate, vegetation monitoring and environmental studies. LST, calculated from remote sensing data, is used in many areas of science such as; hydrology, agriculture, forestry, oceanography etc. The main objective of this study was to develop a model making the LST retrieval process quite simple and automated. This model developed using the ArcGIS Desktop 10.3.1 with the Model Building. Without the model, the process of retrieving LST is very long, and it is susceptible to many mistakes. In this model when user inputs required bands (4,5 and 10) of Landsat-8 data then the model calculate automatically LST and display output. The model first makes the conversions to top of atmosphere (TOA) spectral radiance. Then NDVI is calculated based on band 4 and 5 (NIR and RED) reflectance. Then using the TOA and NDVI model calculates brightness temperature (BT) and Proportion of Vegetation respectively. After that it calculate Land Surface Emissivity with the help of NDVI and Proportion of Vegetation and finally, the model calculates land surface temperatures in degrees Celsius. The findings highlight the capabilities of GIS modelers for such spatial estimation. The developed model can be helpful to field engineers and researchers for using Landsat-8 images for direct estimation of LST, to be used for different other studies to derive LST based products.

Comparison of Machine-Learning and CASA Models for Predicting Apple Fruit Yields from Time-Series Planet Imageries

Apple (Malus domestica Borkh. cv. “Fuji”), an important cash crop, is widely consumed around the world. Accurately predicting preharvest apple fruit yields is critical for planting policy making and agricultural management. This study attempted to explore an effective approach for predicting apple fruit yields based on time-series remote sensing data. In this study, time-series vegetation indices (VIs) were derived from Planet images and analyzed to further construct an accumulated VI (∑VIs)-based random forest (RF∑VI) model and a Carnegie–Ames–Stanford approach (CASA) model for predicting apple fruit yields. The results showed that (1) ∑NDVI was the optimal predictor to construct an RF model for apple fruit yield, and the R2, RMSE, and RPD values of the RF∑NDVI model reached 0.71, 16.40 kg/tree, and 1.83, respectively. (2) The maximum light use efficiency was determined to be 0.499 g C/MJ, and the CASASR model (R2 = 0.57, RMSE = 19.61 kg/tree, and RPD = 1.53) performed better than the CASANDVI model and the CASAAverage model (R2, RMSE, and RPD = 0.56, 24.47 kg/tree, 1.22 and 0.57, 20.82 kg/tree, 1.44, respectively). (3) This study compared the yield prediction accuracies obtained by the models using the same dataset, and the RF∑NDVI model (RPD = 1.83) showed a better performance in predicting apple fruit yields than the CASASR model (RPD = 1.53). The results obtained from this study indicated the potential of the RF∑NDVI model based on time-series Planet images to accurately predict apple fruit yields. The models could provide spatial and quantitative information of apple fruit yield, which would be valuable for agronomists to predict regional apple production to inform and develop national planting policies, agricultural management, and export strategies.

Model building of vegetation cover changes in Abbasid sub-district using SABI and NDVI algorithms

The current study focuses on building a model for vegetation changes in the Abbasid sub-district using the SABI and NDVI algorithm, as the research focused on studying the natural and human factors that affect vegetation cover, and then building digital models that depend on the mathematical algorithms obtained through satellite visuals in a region Spectral, through ARC GIS-based software processors and the Model Builder Extension, to make these models ready tools for obtaining the density of vegetation cover in any geographical area. When applying the two models, it became clear that the SABI model for the derivation of vegetative indices and algae was more accurate than the NDVI model for also deriving vegetative indices. The second model on two scales are (Band 4, Band 5) in Landsat 8. In addition, the SABI model works on deriving aquatic plants as well, while the second model cannot. From the application of the two models to monitoring the change of vegetation cover, it became clear that the region was experiencing a significant change during (30) years for the period (1990-2020) at the level of its agricultural provinces.

Mapping the Urban Atmospheric Carbon Stock by LiDAR and WorldView-3 Data

Currently, the worsening impacts of urbanizations have been impelled to the importance of monitoring and management of existing urban trees, securing sustainable use of the available green spaces. Urban tree species identification and evaluation of their roles in atmospheric Carbon Stock (CS) are still among the prime concerns for city planners regarding initiating a convenient and easily adaptive urban green planning and management system. A detailed methodology on the urban tree carbon stock calibration and mapping was conducted in the urban area of Brussels, Belgium. A comparative analysis of the mapping outcomes was assessed to define the convenience and efficiency of two different remote sensing data sources, Light Detection and Ranging (LiDAR) and WorldView-3 (WV-3), in a unique urban area. The mapping results were validated against field estimated carbon stocks. At the initial stage, dominant tree species were identified and classified using the high-resolution WorldView3 image, leading to the final carbon stock mapping based on the dominant species. An object-based image analysis approach was employed to attain an overall accuracy (OA) of 71% during the classification of the dominant species. The field estimations of carbon stock for each plot were done utilizing an allometric model based on the field tree dendrometric data. Later based on the correlation among the field data and the variables (i.e., Normalized Difference Vegetation Index, NDVI and Crown Height Model, CHM) extracted from the available remote sensing data, the carbon stock mapping and validation had been done in a GIS environment. The calibrated NDVI and CHM had been used to compute possible carbon stock in either case of the WV-3 image and LiDAR data, respectively. A comparative discussion has been introduced to bring out the issues, especially for the developing countries, where WV-3 data could be a better solution over the hardly available LiDAR data. This study could assist city planners in understanding and deciding the applicability of remote sensing data sources based on their availability and the level of expediency, ensuring a sustainable urban green management system.

Detecting the oak lace bug infestation in oak forests using MODIS and meteorological data

The oak lace bug (Corythucha arcuata, Say 1832) is a new invasive sap-sucking species in the European oak forests that was first recorded in Central Europe in 2013. It invaded the region from Southeastern Europe, spreads rapidly, and shows no signs of receding after establishment. In this study, focusing on the oak forests in the transboundary area of Hungary and Croatia, we applied two novel methods for detecting and assessing the impact of the oak lace bug (OLB) during the period 2000–2019 based on MODIS NDVI measured at 250 m spatial and 8-day temporal resolution. The first detection method is based purely on NDVI and has the potential to be used in near real-time detection. The second one, based on the residual Z-score of the NDVI models using daily meteorological and soil water content data as independent variables, aims at improved OLB damage assessment by decoupling the effects of the OLB from those caused by the environmental drivers. The presented detection methods had 61.1% to 93.8% agreement with the in situ data, with a better agreement in forests with high oak share. The overall share of the false-positive OLB detections for the strictest method of model residuals was 1.8%. The results confirmed a strong and year-to-year persistent NDVI decrease (down to -14.5% in pure oak forests) during the late summer which can be attributed to the OLB. The origin of the infestation in the study area was identified to be near a resting station on the major highway from Southeastern to Western Europe, corroborating the assumptions that the OLB spread was primarily facilitated by the transport system. The detected speed of the OLB radial spread in the first 3 years of infestation was under 6 km y-1, but since then it increased to above 50 km y-1.

Improving the monitoring of corn phenology in large agricultural areas using remote sensing data series

Aim of study: Mexico's large irrigation areas demand non-structural actions to improve the irrigation service, such as monitoring crop phenology; however, its application has been limited by the large volumes of field information generated, diversity of crop management and climatic variability. The objective of this study was to generate and validate a methodology to monitor corn (Zea mays L.) phenology from the historical relationship of the vegetation indexes (VIs), EVI and NDVI, with the phenological development (PD) of corn grown in large irrigation zones.Area of study: Irrigation District (ID) 075 “Valle del Fuerte”, northern Sinaloa, Mexico.Material and methods: We used a database of 20 years of climate, field crop growth and crop phenology data, and Landsat satellite images. A methodology was proposed on a large scale supported with GIS and remote sensing data series.Main results: The methodology was validated in 19 plots with an acceptable correlation between observed PD and estimated PD for the two VIs, with slightly better values for EVI than for NDVI. NDVI and EVI models agreed with experimental PD observations in 92.1% of the farms used to validate the methodology, in 2.5% only the NDVI model coincided with the real, in 3.1% only the EVI model coincided, and in 2.3% both models disagreed with observation, generated a stage out of phase with respect to the real phenological stage.Research highlights: is possible to generalize the methodology applied to large irrigation zones with remote sensing data and GIS.

Analysis on dynamic change of vegetation coverage in coastal wetland of Yellow River delta

How to efficiently use multi-source remote sensing images to extract the coastal wetland vegetation landscape information of the Yellow River Delta, and provide scientific basis for wetland protection, utilization and ecological restoration, is of great significance. In this paper, the vegetation coverage of Landsat TM/OLI remote sensing images from 1984 to 2018 is dynamically monitored, and the remote sensing images are analyzed using NDVI and pixel dichotomous models. The conclusions are as follows: From 1984 to 2018, the coverage of vegetation in the northern part of the Yellow River Delta is relatively low, and the coverage of vegetation in the south due to agricultural development is high; Over the past 30 years, the vegetation coverage area of the Yellow River Delta has shown an overall upward trend, and the area of low vegetation coverage areas has declined significantly, decreasing by 19.64% from 1984 to 2018. For the study of the Yellow River Delta coastal wetland, the methods used include the NDVI and pixel dichotomy model, and the results obtained are consistent with the facts, which can help the local ecological protection or soil and water conservation.

Monitoring Flue-Cured Tobacco Leaf Chlorophyll Content under Different Light Qualities by Hyperspectral Reflectance

Rapid assessment of foliar chlorophyll content in tobacco is critical for assessment of growth and precise management to improve quality and yield while minimizing adverse environmental impact. Our objective is to develop a precise agricultural practice predicting tobacco-leaf chlorophyll-a content. Reflectance experiments have been conducted on flue-cured tobacco over 3 consecutive years under different light quality. Leaf hyperspectral reflectance and chlorophyll-a content data have been collected at 15-day intervals from 30 days after transplant until harvesting. We identified the central band that is sensitive to tobacco-leaf chlorophyll-a content and the optimum wavelength combinations for establishing new spectral indices (simple ratio index, RVI; normalized difference vegetation index, NDVI; and simple difference vegetation index, DVI). We then established linear and BackPropagation (BP) neural network models to estimate chlorophyll-a content. The central bands for leaf chlorophyll-a content are concentrated in the visible range (410 - 680 nm) in combination with the shortwave infrared range (1900 - 2400 nm). The optimum spectral range for the spectral band combinations RVI, NDVI, and DVI are 440 and 470 nm, 440 and 470 nm, and 440 and 460 nm, respectively. The linear RVI, NDVI, and DVI models, SMLR model and the BP neural network model have respective R2 values of 0.76, 0.77, 0.69, 0.78 and 0.86, and root mean square error values of 0.63, 1.60, 1.59, 2.04 and 0.05 mg chlorophyll-a/g (fresh weight), respectively. Our results identified chlorophyll-a sensitive spectral regions and new indices facilitate a rapid, non-destructive field estimation of leaf chlorophyll-a content for tobacco.

Can vegetation optical depth reflect changes in leaf water potential during soil moisture dry-down events?

Plant water use strategy is one of the key factors to predict drought impact on vegetation and land-atmosphere fluxes. Vegetation optical depth (VOD) based on microwave radiative transfer inversion has recently been used to assess plant water use strategy. However, VOD is sensitive to both total aboveground biomass (AGB) and leaf water content, with only the latter being a proxy of leaf water potential whose diurnal variation can be used to characterize vegetation iso/anisohydricity. In this study, by using a network of soil water measurements (used as a proxy for predawn leaf water potential), satellite retrieved normalized difference vegetation index (NDVI, as a proxy for AGB), and two satellite VOD products from the Advanced Microwave Scanning Radiometer-Earth Observing System (AMSR-E) sensor, we compare three linear models and one machine learning model to investigate to what extent can VOD be used to represent leaf water potential changes during soil moisture dry-down periods. Linear models with both NDVI and leaf water potential, on average, can explain 33% and 51% of VOD variations of each product respectively. Models using only NDVI explain 27% and 46% of the VOD variance, compared to less than 10% by models using leaf water potential only. With the NDVI and leaf water potential (full) model, leaf water potential contributes around 17% of the VOD variance, which is smaller than NDVI (33%). The machine learning model has overall better performance than the linear models, and also highlight the dominant contribution of AGB to VOD signals. After the AGB contribution to VOD is eliminated by normalizing daytime VOD with nighttime VOD, the residuals carry the information of diurnal variations of leaf water potential and calculations from both VOD datasets are consistent with each other (r = 0.42±0.17, P < 0.01 for 88 out of 94 sites). The response of VODdaytimeVODnighttime to soil water potential can also be used as a new metric for ecosystem iso/anisohydricity. Our study demonstrates that a large proportion of variations in VOD are caused by AGB for temperate ecosystems, and higher accuracy VOD products with additional root-zone soil water potential are needed for ecosystem iso/anisohydricity estimations.

MACHINE LEARNING APPROACH FOR KHARIF RICE YIELD PREDICTION INTEGRATING MULTI-TEMPORAL VEGETATION INDICES AND WEATHER AND NON-WEATHER VARIABLES

Abstract. The development of kharif rice yield prediction models was attempted through Machine Learning approaches such as Artificial Neural Network and Random Forest for the 42 blocks covering 13,141 sq km upland rainfed area of Purulia and Bankura district, West Bengal. Models were developed integrating monthly NDVI with weather and non-weather variables at block-level for the period 2006 to 2015. The model correlation obtained was 0.702 with MSE 0.01. Though the weather variables vs NDVI models are quite satisfactory, NDVI vs kharif rice yield models however, show relatively less correlation, about 0.6 revealing the requirement of varied additional farmer-controlled inputs. Development of NDVI vs crop yield models for different crop growth stages or fortnightly over a larger data set with selective adding of weather and non-weather variables to NDVI would be the most appropriate.