Tanvir Islam

Machine Learning Techniques for Downscaling SMOS Satellite Soil Moisture Using MODIS Land Surface Temperature for Hydrological Application

AbstractMany hydrologic phenomena and applications such as drought, flood, irrigation management and scheduling needs high resolution satellite soil moisture data at a local/regional scale. Downscaling is a very important process to convert a coarse domain satellite data to a finer spatial resolution. Three artificial intelligence techniques along with the generalized linear model (GLM) are used to improve the spatial resolution of Soil Moisture and Ocean Salinity (SMOS) derived soil moisture, which is currently available at a very coarse scale of ~40 Km. Artificial neural network (ANN), support vector machine, relevance vector machine and generalized linear models are chosen for this study to integrate the Moderate Resolution Imaging Spectroradiometer (MODIS) Land Surface Temperature (LST) with the SMOS derived soil moisture. Soil moisture deficit (SMD) derived from a hydrological model called PDM (Probability Distribution Model) is used for the downscaling performance evaluation. The statistical evaluation has also been made with the day-time and night-time MODIS LST differences with the mean day and night-time PDM SMD data for the selection of effective MODIS products. The accuracy and robustness of all the downscaling algorithms are discussed in terms of their assumptions and applicability. The statistical performance indices such as R2, %Bias and RMSE indicates that the ANN (R2  = 0.751, %Bias = −0.628 and RMSE = 0.011), RVM (R2  = 0.691, %Bias = 1.009 and RMSE = 0.013), SVM (R2  = 0.698, %Bias = 2.370 and RMSE = 0.013) and GLM (R2  = 0.698, %Bias = 1.009 and RMSE = 0.013) algorithms on the whole are relatively more skillful to downscale the variability of the soil moisture in comparison to the non-downscaled data (R2  = 0.418 and RMSE = 0.017) with the outperformance of ANN algorithm. The other attempts related to growing and non-growing seasons have been used in this study to reveal that season based downscaling is even better than continuous time series with fairly high performance statistics.

Data fusion techniques for improving soil moisture deficit using SMOS satellite and WRF-NOAH Land surface model

Microwave remote sensing and mesoscale weather models have high potential to monitor global hydrological processes. The latest satellite soil moisture dedicated mission SMOS and WRF-NOAH Land Surface Model (WRF-NOAH LSM) provide a flow of coarse resolution soil moisture data, which may be useful data sources for hydrological applications. In this study, four data fusion techniques: Linear Weighted Algorithm (LWA), Multiple Linear Regression (MLR), Kalman Filter (KF) and Artificial Neural Network (ANN) are evaluated for Soil Moisture Deficit (SMD) estimation using the SMOS and WRF-NOAH LSM derived soil moisture. The first method (and most simplest) utilizes a series of simple combinations between SMOS and WRF-NOAH LSM soil moisture products, while the second uses a predictor equation generally formed by dependent variables (Probability Distributed Model based SMD) and independent predictors (SMOS and WRF-NOAH LSM). The third and fourth techniques are based on rigorous calibration and validation and need proper optimisation for the final outputs backboned by strong non-linear statistical analysis. The performances of all the techniques are validated against the probability distributed model based soil moisture deficit as benchmark; estimated using the ground based observed datasets. The observed high Nash Sutcliffe Efficiencies between the fused datasets with Probability Distribution Model clearly demonstrate an improved performance from the individual products. However, the overall analysis indicates a higher capability of ANN and KF for data fusion than the LWA or MLR approach. These techniques serve as one of the first demonstrations that there is hydrological relevant information in the coarse resolution SMOS satellite and WRF-NOAH LSM data, which could be used for hydrological applications.

Error Correction Modelling of Wind Speed Through Hydro-Meteorological Parameters and Mesoscale Model: A Hybrid Approach

Accurate estimation of wind speed is essential for many hydrological applications. One way to generate wind velocity is from the fifth generation PENN/NCAR MM5 mesoscale model. However, there is a problem in using wind speed data in hydrological processes due to large errors obtained from the mesoscale model MM5. The theme of this article has been focused on hybridization of MM5 with four mathematical models (two regression models- the multiple linear regression (MLR) and the nonlinear regression (NLR), and two artificial intelligence models – the artificial neural network (ANN) and the support vector machines (SVMs)) in such a way so that the properly modelled schemes reduce the wind speed errors with the information from other MM5 derived hydro-meteorological parameters. The forward selection method was employed as an input variable selection procedure to examine the model generalization errors. The input variables of this statistical analysis include wind speed, temperature, relative humidity, pressure, solar radiation and rainfall from the MM5. The proposed conjunction structure was calibrated and validated at the Brue catchment, Southwest of England. The study results show that relatively simple models like MLR are useful tools for positively altering the wind speed time series obtaining from the MM5 model. The SVM based hybrid scheme could make a better robust modelling framework capable of capturing the non-linear nature than that of the ANN based scheme. Although the proposed hybrid schemes are applied on error correction modelling in this study, there are further scopes for application in a wide range of areas in conjunction with any higher end models.

The ASTER Global Emissivity Dataset (ASTER GED): Mapping Earth's emissivity at 100 meter spatial scale

Thermal infrared (TIR) data, acquired by instruments on several NASA satellite platforms, are primarily used to estimate the surface temperature/emissivity of the Earths land surface. One such instrument launched on NASAs Terra satellite in 1999 is the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER), which has a spatial resolution of 90 m. Using ASTER data, NASA/Jet Propulsion Laboratory recently released the most detailed emissivity map of the Earth termed the ASTER Global Emissivity Dataset (ASTER GED) that was acquired by processing millions of cloud free ASTER scenes from 2000 to 2008. The ASTER GEDv3 provides an average emissivity at ~100 m and ~1 km, while GEDv4 provides a monthly emissivity from 2000 to 2015 at ~5 km spatial resolution in the wavelength range between 8 and 12 µm. Validation with lab spectra from four desert sites resulted in an average absolute band error of ~1%, compared to current heritage MODIS products that had average absolute errors of 2.4% (Collection 4) and 4.6% (Collection 5).

Predicting Spatial and Decadal LULC Changes Through Cellular Automata Markov Chain Models Using Earth Observation Datasets and Geo-information

Remote sensing and GIS are important tools for studying land use/land cover (LULC) change and integrating the associated driving factors for deriving useful outputs. This study is based on utilization of Earth observation datasets over the highly urbanized Allahabad district in India. Allahabad district has experienced intense change in LULC in the last few decades. To monitor the changes, advanced techniques in remote sensing and GIS, such as Cellular Automata (CA)-Markov Chain Model (CAMCM) were used to identify the spatial and temporal changes that have occurred in LULC in this area. Two images, 1990 and 2000, were used for calibration and optimization of the Markovian algorithm, while 2010 was used for validating the predictions of CA-Markov using the ground based land cover image. After validating the model, plausible future LULC changes for 2020 were predicted using the CAMCM. Analysis of the LULC pattern maps, achieved through classification of multi-temporal satellite datasets, indicated that the socio-economic and biophysical factors have greatly influenced the growth of agricultural lands and settlements in the area. The two urbanization indicators calculated in this study viz. Land Consumption Ratio (LCR) and Land Absorption Coefficient (LAC) were also used, which indicated a drastic change in the area in terms of urbanization. The predicted LULC scenario for year 2020 provides useful inputs to the LULC planners for effective and pragmatic management of the district and a direction for an effective land use policy making. Further suggestions for an effective policy making are also provided which can be used by government officials to protect this important land resource.

Appraisal of NLDAS-2 Multi-Model Simulated Soil Moistures for Hydrological Modelling

Soil moisture is a key variable in hydrological modelling, which could be estimated by land surface modelling. However the previous studies have focused on evaluating these soil moisture estimates by using point-based measurements, and there is a lack of attention for their appraisal over basin scales particularly for hydrological applications. In this study, we carry out for the first time, a detailed evaluation of five sources of soil moisture products (NLDAS-2 multi-model simulated soil moistures: Noah, VIC, Mosaic and SAC; and a ground observation), against a widely used hydrological model Xinanjiang (XAJ) as a benchmark at a U.S. basin. Generally speaking, all products have good agreements with the hydrological soil moisture simulation, with superior performance obtained from the SAC model and the VIC model. Furthermore, the results indicate that the in-situ measurements in deeper soil layer are still usable for hydrological applications. Nevertheless further improvement is still required on the definition of land surface model layer thicknesses and the related data fusion with the remotely sensed soil moisture. The potential usage of the NLDAS-2 soil moisture datasets in real-time flood forecasting is discussed.

Remote sensing applications in environmental research

Introduction.- Remote sensing-based determination of conifer needle flushing phenology over boreal-dominant regions.- Information System for Integrated Watershed Management using Remote Sensing and GIS.- Sensitivity Exploration of SimSphere Land Surface Model Towards its Use for Operational Products Development from Earth Observation Data.- Remote estimation of land surface temperature for different LULC features of a moist deciduous tropical forest region.- Geospatial Strategy for Estimation of Soil Organic Carbon in Tropical Wildlife Reserve.- A Comparative Assessment Between the Application of Fuzzy Unordered Rules Induction Algorithm and J48 Decision Tree Models in spatial Prediction of Shallow Landslides at Lang Son City, Vietnam.

Non-parametric rain/no rain screening method for satellite-borne passive microwave radiometers at 19–85 GHz channels with the Random Forests algorithm

This paper presents a novel non-parametric pattern recognition method to screen rain/no rain status for satellite-borne passive microwave radiometers in the 19–85 GHz channels. The method is based on randomized decision trees with bootstrap aggregation (Random Forests (RF) algorithm). It relies on pragmatic associations between the input features using Tropical Rainfall Measuring Mission (TRMM) Microwave Imager (TMI) calibrated brightness temperatures and precipitation radar (PR) rain/no rain information as targets. Both these instruments are carried on board the TRMM satellite. In order to develop the method, first, the 10 most significant input features are selected by using feature importance criteria through out-of-bag (OOB) statistics from a total of 17 input features. The input features include the brightness temperatures, as well as some computed signatures – polarization differences (PD), polarization-corrected temperatures (PCT), and scattering indices (SI) at in the 19–85 GHz channels. The feature selection is carried out for different types of surface terrain (ocean, land, and coast), and the selected features are then used for final RF algorithm development. During the dichotomous statistical assessment of the method against the PR rain/no rain status as ‘truth’, the presented method produced reasonable threat scores of 0.50, 0.43, and 0.39, respectively, over ocean, land, and coast surface terrains. Furthermore, the results are compared with the dichotomous scores derived by the Goddard profiling algorithm (GPROF) and, remarkably, the RF-based method corroborated better statistical scores than that of the GPROF. The presented method does not rely on any a priori information and is applicable to other passive microwave radiometers at similar frequencies.

Landscape transform and spatial metrics for mapping spatiotemporal land cover dynamics using Earth Observation data-sets

Abstract Analysis of Earth observation (EO) data, often combined with geographical information systems (GIS), allows monitoring of land cover dynamics over different ecosystems, including protected or conservation sites. The aim of this study is to use contemporary technologies such as EO and GIS in synergy with fragmentation analysis, to quantify the changes in the landscape of the Rajaji National Park (RNP) during the period of 19 years (1990–2009). Several statistics such as principal component analysis (PCA) and spatial metrics are used to understand the results. PCA analysis has produced two principal components (PC) and explained 84.1% of the total variance, first component (PC1) accounted for the 57.8% of the total variance while the second component (PC2) has accounted for the 26.3% of the total variance calculated from the core area metrics, distance metrics and shape metrics. Our results suggested that notable changes happened in the RNP landscape, evidencing the requirement of taking appropriate measures to conserve this natural ecosystem.

WRF Dynamical Downscaling and Bias Correction Schemes for NCEP Estimated Hydro-Meteorological Variables

Rainfall and Reference Evapotranspiration (ETo) are the most fundamental and significant variables in hydrological modelling. However, these variables are generally not available over ungauged catchments. ETo estimation usually needs measurements of weather variables such as wind speed, air temperature, solar radiation and dew point. After the development of reanalysis global datasets such as the National Centre for Environmental Prediction (NCEP) and high performance modelling framework Weather Research and Forecasting (WRF) model, it is now possible to estimate the rainfall and ETo for any coordinates. In this study, the WRF modelling system was employed to downscale the global NCEP reanalysis datasets over the Brue catchment, England, U.K. After downscaling, two statistical bias correction schemes were used, the first was based on sophisticated computing algorithms i.e., Relevance Vector Machine (RVM), while the second was based on the more simple Generalized Linear Model (GLM). The statistical performance indices for bias correction such as %Bias, index of agreement (d), Root Mean Square Error (RMSE), and Correlation (r) indicated that the RVM model, on the whole, displayed a more accomplished bias correction of the variability of rainfall and ETo in comparison to the GLM. The study provides important information on the performance of WRF derived hydro-meteorological variables using NCEP global reanalysis datasets and statistical bias correction schemes which can be used in numerous hydro-meteorological applications.