Majid Nazeer

Evaluation of atmospheric correction models and Landsat surface reflectance product in an urban coastal environment

Precise atmospheric correction is important for applications where small differences in surface reflectance (SR) are significant, such as biomass estimation, crop phenology, and retrieval of water quality parameters. It also enables direct comparison between different image dates and different sensors. As a precursor to monitoring different parameters of water quality around the coastline of Hong Kong using medium-resolution sensors Landsat TM/ETM, and HJ-1A/B, this study evaluated the performance of five atmospheric correction methods. The estimated SR of the first four reflective bands of Landsat 7 ETM+ and of the identical bands of the HJ-1A/B satellites was compared with in situ multispectral radiometer (MSR) SR measurements over sand, artificial turf, grass, and water surfaces for the five atmospheric correction methods – second simulation of the satellite signal in the solar spectrum (6S), fast line-of-sight atmospheric analysis of spectral hypercubes (FLAASH), atmospheric correction (ATCOR), dark object subtraction (DOS), and the empirical line method (ELM). Among the five methods, 6S was observed to be consistently more precise for SR estimation, with significantly less difference from the in-situ-measured SR, especially over lower reflective water surfaces. Of the two image-based methods, DOS performed well over the darker surfaces of water and artificial turf, although still inferior to 6S, while ELM worked well for grass sites as compared to the DOS and equalled the good performance of 6S over the high reflective sand surfaces. The study also evaluated the new standard Landsat SR product Landsat ecosystem disturbance adaptive processing system (LEDAPS) using the in situ measured SR data for the three land surface types – sand, artificial turf, and grass. For the highly and moderately reflecting bright sand and artificial turf, LEDAPS performed poorly, while for the darker grass site it performed better, although still inferior to 6S and ELM methods. This is probably due to the variable aerosol types and atmospheric conditions of Hong Kong, as LEDAPS was mainly compiled with reference to larger continental landmass areas.

Validation of Aqua-MODIS C051 and C006 Operational Aerosol Products Using AERONET Measurements Over Pakistan

The aim of this study was to evaluate the performance of the Aqua-MODIS (MYD04) collections 5.1 (C051) and 6 (C006) operational aerosol products over Pakistan. These include the Dark Target (DT) and the Deep Blue (DB) C051/C006 aerosol optical depth (AOD) observations at 10-km spatial resolution, and which were validated using 7 years (2007-2013) AOD measurements from two AERONET stations located in Pakistans two largest cities, Lahore and Karachi. Lahore is an inland city, consisting of built-up areas and agricultural land and dominated by mainly fine aerosol particles. Karachi is an urban-coastal city with built-up areas and bare land, and mainly affected by coarse aerosol particles. The retrieval uncertainties and accuracy were evaluated using the expected error over land (EE : ±0.05 + 15%), the root-mean-square error (RMSE), the mean absolute error (MAE), and the relative mean bias (RMB). It was found that the DT C051 AOD retrievals were significantly overestimated over both AERONET sites with mean overestimation of 21% and 31% for Lahore and Karachi, respectively. On the other hand, the DB C051 retrievals were underestimated by 10% and 35% for both cities. Similar to the DT C051, the C006 AOD retrievals were overestimated over Lahore, but significantly improved over Karachi, as the mean overestimation reduced from 31% (RMB = 1.31) to 4% (RMB = 1.04) and the percentage of retrievals within the EE increased from 38.46% to 63.33%. However, the DB C006 AOD retrievals have similar errors (RMSE and MAE) and the percentage of retrievals within the EE over both cities as C051, but they have more mean underestimations. Spatio-temporal distributions showed that the DT and DB C006 were well correlated with Karachi and Lahore AERONET measurements, respectively. Therefore, these results recommend the use of the DT C006 algorithm over Karachi and the DB C006 over Lahore for qualitative regional air quality applications due to differences in land cover characteristics and aerosol types.

Combining Landsat TM/ETM+ and HJ-1 A/B CCD Sensors for Monitoring Coastal Water Quality in Hong Kong

High suspended solid (SS) concentrations in coastal waters are aesthetically undesirable, and adversely affect fisheries and coastal ecosystems. Environmental agencies usually require frequent measurements of SS over coastal regions at a spatially detailed level for water quality assessment and control. To develop a method for SS estimation in the complex coastal waters of Hong Kong, an archive of 57 Landsat Thematic Mapper (TM), Enhanced Thematic Mapper Plus (ETM+), and HJ-1 A/B Charged Couple Device (CCD) images over a 13-year period from January 2000 to December 2012 was used. Atmospherically corrected Landsat TM/ETM+ and HJ-1 A/B CCD bands 1-4 along with 240 in situ field samples of SS concentration collected within 2 h of image acquisition, were used to develop and validate regression models over a wide range of SS concentrations from 0.5-56.0 mg/L. The best representation of actual SS concentrations was given by the log-transformed combination of Band 2 (Green, 0.52-0.60 μm) and Band 3 (Red, 0.63-0.69 μm), with correlation coefficient (R) of 0.85, root-mean-square error of 2.60 mg/L and mean absolute error of 2.04 mg/L. This is attributed to the sensitivity of SS to green and red wavelengths specific to the characteristic refractive index and grain size of SS found in Hong Kong waters. This letter is considered more robust than previous studies, due to the much larger number of images and in situ samples used for model development and validation, as well as the different times of year and wide range of SS concentrations investigated.

A new approach for the estimation of phytoplankton cell counts associated with algal blooms

This study proposes a method for estimating phytoplankton cell counts associated with an algal bloom, using satellite images coincident with in situ and meteorological parameters. Satellite images from Landsat Thematic Mapper (TM), Enhanced Thematic Mapper Plus (ETM+), Operational Land Imager (OLI) and HJ-1 A/B Charge Couple Device (CCD) sensors were integrated with the meteorological observations to provide an estimate of phytoplankton cell counts. All images were atmospherically corrected using the Second Simulation of the Satellite Signal in the Solar Spectrum (6S) atmospheric correction method with a possible error of 1.2%, 2.6%, 1.4% and 2.3% for blue (450-520nm), green (520-600nm), red (630-690nm) and near infrared (NIR 760-900nm) wavelengths, respectively. Results showed that the developed Artificial Neural Network (ANN) model yields a correlation coefficient (R) of 0.95 with the in situ validation data with Sum of Squared Error (SSE) of 0.34cell/ml, Mean Relative Error (MRE) of 0.154cells/ml and a bias of -504.87. The integration of the meteorological parameters with remote sensing observations provided a promising estimation of the algal scum as compared to previous studies. The applicability of the ANN model was tested over Hong Kong as well as over Lake Kasumigaura, Japan and Lake Okeechobee, Florida USA, where algal blooms were also reported. Further, a 40-year (1975-2014) red tide occurrence map was developed and revealed that the eastern and southern waters of Hong Kong are more vulnerable to red tides. Over the 40 years, 66% of red tide incidents were associated with the Dinoflagellates group, while the remainder were associated with the Diatom group (14%) and several other minor groups (20%). The developed technology can be applied to other similar environments in an efficient and cost-saving manner.

A New MODIS C6 Dark Target and Deep Blue Merged Aerosol Product on a 3 km Spatial Grid

In Moderate Resolution Imaging Spectroradiometer (MODIS) Collection (C6) aerosol products, the Dark Target (DT) and Deep Blue (DB) algorithms provide aerosol optical depth (AOD) observations at 3 km (DT3K) and 10 km (DT10K), and at 10 km resolution (DB10K), respectively. In this study, the DB10K is resampled to 3 km grid (DB3K) using the nearest neighbor interpolation technique and merged with DT3K to generate a new DT and DB merged aerosol product (DTB3K) on a 3 km grid using Simplified Merge Scheme (SMS). The goal is to supplement DB10K with high-resolution information over dense vegetation regions where DT3K is susceptible to error. SMS is defined as “an average of the DT3K and DB3K AOD retrievals or the available one with the highest quality flag”. The DT3K and DTB3K AOD retrievals are validated from 2008 to 2012 against cloud-screened and quality-assured AOD from 19 AERONET sites located in Europe. Results show that the percentage of DTB3K retrievals within the expected error (EE = ± (0.05 + 20%)) and data counts are increased by 40% and 11%, respectively, and the root mean square error and the mean bias are decreased by 26% and 54%, respectively, compared to the DT3K retrievals. These results suggest that the DTB3K product is a robust improvement over DT3K alone, and can be used operationally for air quality and climate-related studies as a high-resolution supplement to the current MODIS product suite.

Global Validation of MODIS C6 and C6.1 Merged Aerosol Products over Diverse Vegetated Surfaces

In this study, the MODerate resolution Imaging Spectroradiometer (MODIS) Collections 6 and 6.1 merged Dark Target (DT) and Deep Blue (DB) aerosol products (DTBC6 and DTBC6.1) at 0.55 µm were validated from 2004–2014 against Aerosol Robotic Network (AERONET) Version 2 Level 2.0 AOD obtained from 68 global sites located over diverse vegetated surfaces. These surfaces were categorized by static values of monthly Normalized Difference Vegetation Index (NDVI) observations obtained for the same time period from the MODIS level-3 monthly NDVI product (MOD13A3), i.e., partially/non–vegetated (NDVIP ≤ 0.3), moderately–vegetated (0.3 0.5) surfaces. The DTBC6 and DTBC6.1 AOD products are accomplished by the NDVI criteria: (i) use the DT AOD retrievals for NDVI > 0.3, (ii) use the DB AOD retrievals for NDVI < 0.2, and (iii) use an average of the DT and DB AOD retrievals or the available one with highest quality assurance flag (DT: QAF = 3; DB: QAF ≥ 2) for 0.2 ≤ NDVI ≤ 0.3. For comparison purpose, the DTBSMS AOD retrievals were included which were accomplished using the Simplified Merge Scheme, i.e., use an average of the DTC6.1 and DBC6.1 AOD retrievals or the available one for all the NDVI values. For NDVIP surfaces, results showed that the DTBC6 and DTBC6.1 AOD retrievals performed poorly over North and South America in terms of the agreement with AERONET AOD, and over Asian region in terms of retrievals quality as the small percentage of AOD retrievals were within the expected error (EE = ± (0.05 + 0.15 × AOD). For NDVIM surfaces, retrieval errors and poor quality in DTBC6 and DTBC6.1 were observed for Asian, North American and South American sites, whereas good performance, was observed for European and African sites. For NDVID surfaces, DTBC6 does not perform well over the Asian and North American sites, although it contains retrievals only from the DT algorithm which was developed for dark surfaces. Overall, the performance of the DTBC6.1 AOD retrievals was significantly improved compared to the DTBC6, but still more improvements are required over NDVIP, NDVIM and NDVID surfaces of Asia, NDVIM and NDVID surfaces of North America, and NDVIM surfaces of South America. The performance of the DTBSMS retrievals was better than the DTBC6 and DTBC6.1 retrievals with 11–13% (31%) greater number of coincident observations, 6–9% (14–22%) greater percentage of retrievals within the EE, and 30–100% (46–100%) smaller relative mean bias compared to the DTBC6.1 (DTBC6) at a global scale.

Selection of atmospheric correction method and estimation of Chlorophyll-a (Chl-a) in coastal waters of Hong Kong

Precise atmospheric correction is important for applications where small differences in Surface Reflectance (SR) are significant, such as biomass estimation, crop phenology and retrieval of water quality parameters. As a precursor to monitor water quality parameter Chlorophyll-a (Chl-a), around the coastal waters of Hong Kong using medium resolution sensor, Landsat Thematic Mapper (TM) and Enhanced Thematic Mapper Plus (ETM+), this study evaluated the performance of five atmospheric correction methods. The estimated SR, using the five methods including, 6S (Second Simulation of the Satellite Signal in the Solar Spectrum), FLAASH (Fast Line-of-sight Atmospheric Analysis of Spectral Hypercubes), ATCOR (ATmospheric CORection), DOS (Dark Object Subtraction) and ELM (Empirical Line Method), was validated with in situ Multispectral Radiometer (MSR) SR measurements over sand, artificial turf, grass and water surfaces for the first four reflective bands of Landsat 7 ETM+ and HJ-1 A/B satellites. Among the five methods 6S was observed to be consistently more precise for SR estimation, with significantly less difference from the in situ measured SR, especially over lower reflective water surfaces. Of the two image-based methods, DOS performed well over the darker surfaces of water and artificial turf, although still inferior to 6S, while ELM worked well for grass sites and equaled the good performance of 6S over the high reflective sand surfaces. Therefore, the Landsat TM/ETM+ atmospherically corrected images along with the in situ Chl-a data from 2000 to 2012 were used to develop and validate the regression models for Chl-a concentration of 0.3 to 13.0 ug/l. The validation results showed that the most accurate Chl-a was estimated using the ratio of Band 3 (0.63-0.69 μm) and (Band 1)2 (0.45-0.52 μm) with correlation coefficient (R) 0.86, Root Mean Square Error (RMSE) of 2.70 ug/l and Mean Absolute Error (MAE) of 1.13 ug/l for coastal waters of Hong Kong.

Evaluation of Ordinary Least Square (OLS) and Geographically Weighted Regression (GWR) for Water Quality Monitoring: A Case Study for the Estimation of Salinity

Landsat-5 Thematic Mapper (TM) dataset have been used to estimate salinity in the coastal area of Hong Kong. Four adjacent Landsat TM images were used in this study, which was atmospherically corrected using the Second Simulation of the Satellite Signal in the Solar Spectrum (6S) radiative transfer code. The atmospherically corrected images were further used to develop models for salinity using Ordinary Least Square (OLS) regression and Geographically Weighted Regression (GWR) based on in situ data of October 2009. Results show that the coefficient of determination (R2) of 0.42 between the OLS estimated and in situ measured salinity is much lower than that of the GWR model, which is two times higher (R2 = 0.86). It indicates that the GWR model has more ability than the OLS regression model to predict salinity and show its spatial heterogeneity better. It was observed that the salinity was high in Deep Bay (north-western part of Hong Kong) which might be due to the industrial waste disposal, whereas the salinity was estimated to be constant (32 practical salinity units) towards the open sea.

Evaluation of Empirical and Machine Learning Algorithms for Estimation of Coastal Water Quality Parameters

Coastal waters are one of the most vulnerable resources that require effective monitoring programs. One of the key factors for effective coastal monitoring is the use of remote sensing technologies that significantly capture the spatiotemporal variability of coastal waters. Optical properties of coastal waters are strongly linked to components, such as colored dissolved organic matter (CDOM), chlorophyll-a (Chl-a), and suspended solids (SS) concentrations, which are essential for the survival of a coastal ecosystem and usually independent of each other. Thus, developing effective remote sensing models to estimate these important water components based on optical properties of coastal waters is mandatory for a successful coastal monitoring program. This study attempted to evaluate the performance of empirical predictive models (EPM) and neural networks (NN)-based algorithms to estimate Chl-a and SS concentrations, in the coastal area of Hong Kong. Remotely-sensed data over a 13-year period was used to develop regional and local models to estimate Chl-a and SS over the entire Hong Kong waters and for each water class within the study area, respectively. The accuracy of regional models derived from EPM and NN in estimating Chl-a and SS was 83%, 93%, 78%, and 97%, respectively, whereas the accuracy of local models in estimating Chl-a and SS ranged from 60–94% and 81–94%, respectively. Both the regional and local NN models exhibited a higher performance than those models derived from empirical analysis. Thus, this study suggests using machine learning methods (i.e., NN) for the more accurate and efficient routine monitoring of coastal water quality parameters (i.e., Chl-a and SS concentrations) over the complex coastal area of Hong Kong and other similar coastal environments.

Modeling of Chlorophyll-a concentration for the coastal waters of Hong Kong

In coastal waters, accurate remote sensing retrieval of Chlorophyll-a (Chl-a) is challenging. In a spatially complex urban coastal region like Hong Kong, the development of a single Chl-a estimation algorithm over whole region is unrealistic. In such case the best strategy will be to develop an individual algorithm for each water type to precisely estimate Chl-a concentration. Therefore, to define the effective water zones in the region, Fuzzy c-Means (FCM) clustering was applied to surface reflectance derived from the first four bands of the Landsat Thematic Mapper/Enhanced Thematic Mapper Plus (TM/ETM+) and HJ-1 A/B Charge Couple Device (CCD) sensors for 76 Hong Kong Environmental Protection Department (EPD) water monitoring stations. The FCM clustering results suggested the existence of five optically different water types in the region. Cluster specific algorithms were then developed for the retrieval of Chl-a concentrations using Neural Network (NN) and Regression Modeling (RM) techniques. Twenty seven Landsat TM/ETM+ (January 2000-December 2012) and thirty HJ-1 A/B CCD (September 2008-December 2012) cloud free images paired with in situ Chl-a data were used for development and validation of the NNs and RMs. The performance of the cluster specific NNs and RMs suggested that NN can efficiently estimate and map Chl-a concentrations with greater confidence as compared to band ratio algorithms developed using regression modeling. Overall, the validation results showed a correlation of 0.63 to 0.85 between the NN estimated and in situ measured Chl-a concentrations compared to a correlation of 0.26 to 0.54 between the RM estimated and in situ measured Chl-a concentrations.