Classification Of Satellite Data Research Articles

Coupling UAV and satellite data for tree species identification to map the distribution of Caspian poplar

ContextMapping the distribution of species, especially those that are endemic and endangered like certain tree species, is a vital step in the effective planning and execution of conservation programs and monitoring efforts. This task gains even more significance as it directly contributes to forest conservation by highlighting the importance of species diversity.ObjectivesOur study objective was to assess the detection accuracy of a specific tree using different remote sensing sources and approaches.MethodsInitially, individual trees were identified and classified using a canopy height model derived from UAV data. Next, we carried out the classification of satellite data within the Google Earth Engine. Lastly, we scaled the UAV-RGB dataset to match the spatial resolution of Sentinel-2, which was then employed to train random forest models using the multispectral data from Sentinel-2.ResultsFor the UAV data, we achieved overall accuracies of 56% for automatically delineated tree crowns and 83% for manually delineated ones. Regarding the second approach using Sentinel-2 data, the classification in the Noor forest yielded an overall accuracy of 74% and a Kappa coefficient of 0.57, while in the Safrabasteh forest, the accuracy was 80% with a Kappa of 0.61. In the third approach, our findings indicate an improvement compared to the second approach, with the overall accuracy and Kappa coefficient of the classification rising to 82% and 0.68, respectively.ConclusionsIn this study, it was found that according to the purpose and available facilities, satellite and UAV data can be successfully used to identify a specific tree species.

Land Use and Land Cover Change Detection Using the Random Forest Approach: The Case of The Upper Blue Nile River Basin, Ethiopia.

Monitoring land use change dynamics is critical for tackling food security, climate change, and biodiversity loss on a global scale. This study is designed to classify land use and land cover in the upper Blue Nile River Basin (BNRB) using a random forest (RF) algorithm. The Landsat images for Landsat 45, Landsat 7, and Landsat 8 are used for classification purposes. The study area is classified into seven land use/land cover classes: cultivated lands, bare lands, built-ups, forests, grazing lands, shrublands, and waterbodies. The accuracy of classified images is 83%, 85%, and 91% using the Kappa index of agreements. From 1983 to 2022 periods, cultivated lands and built-up areas increased by 47541 and 1777 km2, respectively, at the expense of grazing lands, shrublands, and forests. Furthermore, the area of water bodies has increased by 662 km2 due to the construction of small and large-scale irrigation and hydroelectric power generation dams. The main factors that determine agricultural land expansion are related to population growth. Therefore, land use and land cover change detection using a random forest is an important technique for multispectral satellite data classification to understand the optimal use of natural resources, conservation practices, and decision-making for sustainable development.

Decadal Remote Sensing Analysis of Seagrass Changes in Palu Bay, Central Sulawesi

Seagrass meadows provide a variety of material, non-material and regulatory coastal ecosystem service; however, as the distribution of seagrass beds changes over time due to both anthropogenic activities and natural factors, it is important to monitor changes in seagrass condition. Seagrass meadows in Palu Bay are threatened by activities such as coastal development and land reclamation. Additionally, the bay was hit by a significant tsunami in 2018, which could have impacted ecosystems in the bay, including seagrass meadows. The aim of this study was to detect changes in seagrass extent and distribution over a 10-year period from 2012 to 2022 and changes in land use over approximately a decade (2010 and 2021) through the use of remote sensing technology. Changes in eagrass meadow areal extent were analyzed using data from a 2012 Landsat 7 Satellite Data Acquisition and a 2022 Landsat 8 Satellite Data Acquisition. Water column correction was implemented using the Lyzenga Algorithm. The results showed a significant decrease in the area of seagrass meadows around the coastal area of Palu Bay. Seagrass meadows in 2012 and 2022 covered 127.08 Ha and 87.79 Ha, respectively, indicating a decrease in extent of 43.29 Ha. As the accuracy of the satellite data classification results was 80%, the results are considered acceptable. Anthropogenic activities (mainly mining and construction related) are strongly suspected as the main drivers of this decline, while earthquake and tsunami events likely aggravated the degradation of coastal ecosystems in Palu Bay, including seagrass meadows.

Utilization of Remote Sensing Images for Compiling The Characteristics Map of Indramayu Beach, West Java

The beach has an important role in the development and economy of a region. Therefore, it is important to understand the condition and characteristics of the coast so that it can be managed and developed wisely. One way to understand the condition of the beach is through making a map of the characteristics of the beach. A coastal characteristic map is a graphical representation of the conditions and characteristics of the geology, geography, and hydrology of a coast. Making a coastal characteristic map with remote sensing requires several steps, namely: 1). satellite data collection: Pre-processing satellite data; 2). classification of satellite data; 3). satellite data analysis; and finally 4). making a coastal characteristic map. The area that is the fabric material is the beach of Indramayu Regency, West Java. Based on the analysis and interpretation of 2015 Landsat 8 (LDCM) image data, 2020 Sentinel 2 Image data, BATNAS, and DEMNAS data, it is known that the characteristics of the Indramayu beach are that it has geo-quaternary building materials which include: (a) coastal bunds, ( b) swamp deposits, (c) shallow marine deposits, (d) sand deposits, (e) flood plains, (f) river deposits, and (g) floodplain deposits. The morphology includes: (a) beach sand plains, (b) concave coastal swamps, (c) old beach bunds, (d) young beach bunds, (e) alluvial plains, (f) flood plains, and (g) fan plains delta. Finally, it has a distribution of surface sediments on the seabed which includes: (a) sand, (b) silt, gravelly sand, (d) slightly gravelly silt, (e) slightly gravelly sandy mud, and (f) gravelly mud.

Multi-Temporal Detection of Agricultural Land Losses Using Remote Sensing and Gis Techniques, Shanderman, Iran

Abstract Over the last decades, north of Iran underwent remarkable land use/cover changes due to socio-economic and environmental factors. This study, focused on agricultural land changes for the period of 1990–2020 at Shanderman, Iran, employed Landsat 5 TM, and Landsat 8 OLI/TIRS images. A supervised maximum likelihood classification technique was utilized for the purposes of satellite data classification to four classes: agricultural land, forest, grassland, and built-up area. Results of land change modeller showed that, during the last three decades, agricultural land, grassland and forest decreased by 42.81%, 35.50%, and 4.05%, respectively, while built-up area increased by 361.23%. Most of the losses in agriculture areas occurred in 1990–2011 (44.64%). The predominant losses in 2011–2020 belonged to the forestland (12.47%), making them approx. 3.44 times higher than in 1990–2011. The results highlight the need for serious attention to the deforestation phenomenon, which leads to the conversion of forest into agricultural and built-up areas.

Conception of the Sflavar Approach: Classification of Satellite Data

Unsupervised classification is the search for homogeneous groups in a dataset. This problem is therefore of great complexity and the use of approximation algorithms is inevitable. In this regard, we will call upon a system resulting from a vision of collective intelligence inspired by the behavior of frogs as individuals seeking food, collaborating with each other to accomplish tasks that they cannot perform individually. As a measure of optimization we will make a change by integrating other variation operators. This will be applied to a LANDSAT5 TM (Thematic Mapper) satellite image of the ORAN region.

Tracing land use and land cover change in peri-urban Delhi, India, over 1973–2017 period

Land use and land cover changes over 1973-2017 period in peripheral Delhi were mapped based on digital classification of satellite data and their driving forces ascertained. Urban area expanded and agricultural area diminished at annual rates of 38.6% and 2.1%, respectively, during the 1973-2017 period. Urban expansion occurred more in scrub and sparse vegetation areas than in cultivated lands or ponds. Loss of cultivated land happened mostly due to abandonment of cropping and tree planting in farmhouses developed by the urban elites. Improvement in the state of forests in terms of their expansion as well as densification offsets their loss due to urbanisation, encroachment and logging. The increment in the green cover was due to strict enforcement of compensatory afforestation/forest conservation law, growing demand of ecotourism, emergence of tree-clad farmhouses and increased environmental awareness and surveillance. This research will help in comprehending policies favouring sustainable urban development.

Advantage of Combining OBIA and Classifier Ensemble Method for Very High-Resolution Satellite Imagery Classification

Accurate and timely collection of urban land use and land cover information is crucial for many aspects of urban development and environment protection. Very high-resolution (VHR) remote sensing images have made it possible to detect and distinguish detailed information on the ground. While abundant texture information and limited spectral channels of VHR images will lead to the increase of intraclass variance and the decrease of the interclass variance. Substantial studies on pixel-based classification algorithms revealed that there were some limitations on land cover information extraction with VHR remote sensing imagery when applying the conventional pixel-based classifiers. Aiming at evaluating the advantages of classifier ensemble strategies and object-based image analysis (OBIA) method for VHR satellite data classification under complex urban area, we present an approach-integrated multiscale segmentation OBIA and a mature classifier ensemble method named random forest. The framework was tested on Chinese GaoFen-1 (GF-1), and GF-2 VHR remotely sensed data over the central business district (CBD) of Zhengzhou metropolitan. Process flow of the proposed framework including data fusion, multiscale image segmentation, best optimal segmentation scale evaluation, multivariance texture feature extraction, random forest ensemble learning classifier construction, accuracy assessment, and time consumption. Advantages of the proposed framework were compared and discussed with several mature state-of-art machine learning algorithms such as the k -nearest neighbor (KNN), support vector machine (SVM), and decision tree classifier (DTC). Experimental results showed that the OA of the proposed method is up to 99.29% and 98.98% for the GF-1 dataset and GF-2 dataset, respectively. And the OA is increased by 26.89%, 11.79%, 11.89%, and 4.26% compared with the traditional machine learning algorithms such as the decision tree classifier (DTC), support vector machine (SVM), k -nearest neighbor (KNN), and random forest (RF) on the test of the GF-1 dataset; OA increased by 32.31%, 13.48%, 9.77%, and 7.72% for the GF-2 dataset. In terms of time consuming, by rough statistic, OBIA-RF spends 223.55 s, SVM spends 403.57 s, KNN spends 86.93 s, and DT spends 0.61 s on average of the GF-1 and GF-2 datasets. Taking the account classification accuracy and running time, the proposed method has good ability of generalization and robustness for complex urban surface classification with high-resolution remotely sensed data.

Detecting the development stages of natural forests in northern Iran with different algorithms and high-resolution data from GeoEye-1.

Nowadays, taking advantage of multispectral sensors with the high spatial and spectral resolution and using a variety of plant indices and remote sensing have provided the possibility of more accurate analysis and classification of satellite data in the identification of natural phenomena. Nowadays, obtaining information about the structure of forests through remote sensing data to manage renewable resources is of interest to managers and researchers. This study produced the development maps of natural forests in northern Iran by taking advantage of GeoEye-1 data, training samples, and various algorithms through the pixel-based, object-based, and model-based methods. The classification's ultimate accuracy was calculated by each of the above methods with the overall accuracy parameters and kappa coefficient. By examining the accuracy of map classification resulting from different methods, the maximum accuracy (78%) in object-based method was estimated based on the segmentation of NDVI and the maximum likelihood algorithm. Meanwhile, some other classification methods showed much less accuracy. The results showed that the algorithms following the structural patterns for pixel distribution classification provided a higher accuracy. Also, these results showed the high potential of high-resolution data from GeoEye-1 in the production of forest development maps and the effect of choosing the appropriate algorithm in the production of higher accuracy maps.

Classification of Australian Waterbodies across a Wide Range of Optical Water Types

Baseline determination and operational continental scale monitoring of water quality are required for reporting on marine and inland water progress to Sustainable Development Goals (SDG). This study aims to improve our knowledge of the optical complexity of Australian waters. A workflow was developed to cluster the modelled spectral response of a range of in situ bio-optical observations collected in Australian coastal and continental waters into distinct optical water types (OWTs). Following clustering and merging, most of the modelled spectra and modelled specific inherent optical properties (SIOP) sets were clustered in 11 OWTs, ranging from clear blue coastal waters to very turbid inland lakes. The resulting OWTs were used to classify Sentinel-2 MSI surface reflectance observations extracted over relatively permanent water bodies in three drainage regions in Eastern Australia. The satellite data classification demonstrated clear limnological and seasonal differences in water types within and between the drainage divisions congruent with general limnological, topographical, and climatological factors. Locations of unclassified observations can be used to inform where in situ bio-optical data acquisition may be targeted to capture a more comprehensive characterization of all Australian waters. This can contribute to global initiatives like the SDGs and increases the diversity of natural water in global databases.

Shelterbelt Agroforestry Systems Inventory and Removal Analyzed by Object-based Classification of Satellite Data in Saskatchewan, Canada

Shelterbelt agroforestry systems inventory is challenging given their narrow linear feature and extensive distribution. The objective of this study was to evaluate the capability of Sentinel-2A Multispectral Instrument (MSI) and Sentinel-1B Synthetic Aperture Radar (SAR) imagery in (i) delineating shelterbelt tree rows on managed agricultural land in Saskatchewan, Canada, and (ii) detecting shelterbelt removal during the period 2008–2016. Contrast split segmentation for the normalized difference vegetation index and Gaussian filter (line filter) data sets from SAR were used to delineate feature borders. Several feature variables from the spectral bands of MSI were used as inputs for an object-based classification using the random forest classifier. A resulting land cover map, including the linear features of existing shelterbelts, was created with an overall accuracy of 80% and kappa value of 0.69. Shelterbelt change detection analysis using the land cover map (2016) and a legacy shelterbelt inventory map (2008) showed that 354 km of shelterbelts were removed within the study area (1,400 km2), accounting for 29.8% of the total shelterbelt length present in 2008. Our results demonstrated that the combination of Sentinel SAR and MSI imagery can provide sufficient information for mapping future shelterbelt planting, as well as allow the detection of shelterbelt removal.

Classification of Satellite Images in Assessing Urban Land Use Change Using Scale Optimization in Object-Oriented Processes (A Case Study: Ardabil City, Iran)

By using satellite imagery, the recognition and evaluation of various phenomena and extraction of information necessary for the planning of land resources or other purposes are easily accomplished. The purpose of this study is to compare the efficiency of seven commonly used methods of monitored classification of satellite data to evaluate land use changes using TM and OLI Landsat, IRS, Spot5 and Quick Bird bands as well as different color combinations of these images to detect agricultural land, residential areas and aquatic areas using object-oriented processing. Digital processing of satellite images was carried out in 1998 and 2016 using advanced methods. Training samples were extracted in five user classes by eCognition software using segmentation scale optimization, different color combinations and coefficients of shape and compression. The appropriate segmentation scale for arable land, human complications and the blue areas were, respectively, 50, 8 and 10. Then each image was classified separately using seven methods and extracted samples, and efficiency of each classification method was obtained by calculating two general health and Kappa coefficients. The results show that the accuracy of each classification method and the neural network with a total accuracy of 94.475 and Kappa coefficient of 92.095 were selected as the most accurate classification method. These results show that the sampling of educational samples with proper precision of the classes in the images and dependency probability of each satellite images pixel can be useful in classifying group available in helpful area.

Monitoring spatial LULC changes and its growth prediction based on statistical models and earth observation datasets of Gautam Budh Nagar, Uttar Pradesh, India

It is well known and witnessed the fact that in recent years the growth of urbanization and increasing urban population in the cities, particularly in developing countries, are the primary concern for urban planners and other environmental professionals. The present study deals with multi-temporal satellite data along with statistical models to map and monitor the LULC change patterns and prediction of urban expansion in the upcoming years for one of the important cities of Ganga alluvial Plain. With the help of our study, we also tried to portray the impact of urban sprawl on the natural environment. The long-term LULC and urban spatial change modelling was carried out using Landsat satellite data from 2001 to 2016. The assessment of the outcome showed that increase in urban built-up areas favoured a substantial decline in the agricultural land and rural built-up areas, from 2001 to 2016. Shannon’s entropy index was also used to measure the spatial growth patterns over the period of time in the study area based on the land-use change statistics. Prediction of the future land-use growth of the study area for 2019, 2022 and 2031 was carried out using artificial neural network method through Quantum GIS software. Results of the simulation model revealed that 14.7% of urban built-up areas will increase by 2019, 15.7% by 2022 and 18.68% by 2031. The observation received from the present study based on the long-term classification of satellite data, statistical methods and field survey indicates that the predicted LULC map of the area will be precious information for policy and decision-makers for sustainable urban development and natural resource management in the area for food and water security.

Review of machine learning methods for Big satellite Data classification

With the appearance of free access to Big satellite data, the development of machine learning methods based on geospatial data, in particular satellite data, is becoming more and more relevant. In this paper, we consider and analyze the peculiarities of the basic machine learning methods and results of their application to the tasks of land cover classification based on high resolution satellite data. Special attention is paid to deep architectures, in particular, convolutional neural networks, which nowadays are the most powerful and precise method for visual pattern recognizing. We determine the main advantages of the deep learning methods over the traditional approaches to the classification tasks, that have been used over the last decades and based on expert knowledge to the features extraction from the input data.

Sugarcane crop identification from LISS IV data using ISODATA, MLC, and indices based decision tree approach

Image classification is one of the crucial techniques in detecting the crops from remotely sensed data. Crop identification and discrimination provide an important basis for many agricultural applications with various purposes, such as cropping pattern analysis, acreage estimation, and yield estimation. Accurate and faster estimation of crop area is very essential for projecting yearly agriculture production for deciding agriculture policies. Remote sensing is a technique that allows mapping of large areas in a fast and economical way. In many applications of remote sensing, a user is often interested in identifying the specific crop only while other classes may be of no interest. Indian Remote Sensing Satellite (IRS-P6) LISS IV sensor image of spatial resolution 5.8 m has been used to identify the sugarcane crop for the Chhapar village of Muzaffarnagar District, India. Classification of satellite data is one of the primary steps for information extraction for crop land identification. In recent years, decision tree approach to image analysis has been developed for the assessment and improvement of traditional statistically based image classification. In this study, ISODATA, MLC, and vegetation indices based decision tree approaches are used for classifying LISS IV imagery. The 11 vegetation index images have been generated for decision tree classification. All the three methods are compared and it is found that the best performance is given by the decision tree method. Vegetation indices based decision tree method for sugarcane classification, the user’s accuracy, producer’s accuracy, overall accuracy, and kappa coefficient were found 88.17, 86.59, and 87.93% and 0.86 respectively.

Satellite Parametric Description to Ontology Concepts and Semantic Classification of Satellite Data

Current satellite data retrieval systems retrieves data using latitude, longitude, date, time and sensor parameters like wind, cloud etc. To achieve concept based satellite data retrieval like Storm, Hurricane, Overcast and Frost etc., requires ontological concept descriptions using satellite observation parameters and concept based classification of satellite data. The current research work has designed and implemented a two phase methodology to achieve this. The phase 1 defines ontology concepts through satellite observation parameters and phase 2 describes ontology concept based satellite data classification. The efficiency of the methodology is been tested by taking the Kalpana satellite data from MOSDAC and weather ontology. This achieves concept based retrieval of satellite data, application interoperability and strengthen the ontologies. The current methodology is implemented and results in concept based satellite data classification, storage and retrieval.

Neural Network Approach for Automatic Landuse Classification of Satellite Images: One-Against-Rest and Multi-Class Classifiers

Neural Network (ANN) is an important Artificial Intelligence (AI) and Machine Learning (ML) method used in various remote sensing applications such as image classification, pattern recognition etc.One of important remote sensing applications is the landuse classification i.e. classification of satellite data into various landuse classes such as forest, waterbody, snowcover etc. Landuse classification from satellite data can take place in manual, semi-automatic or automatic mode. Automatic landuse classification is necessary to reduce manual efforts,which can be achieved by making use of machine learning techniques. This paper uses neural network approach for automatic landuse classification from satellite data by providing two classification approaches using multi layer perceptron (MLP) namely one against rest classification (OARC) and multi class classification (MCC), and then provides the comparison between these two approaches. KeywordsNeural Network (ANN), Multi Layer Perceptron (MLP), Error Back Propagation (EBP), Landuse Classification, One-Against-Rest Classification (ORAC), Multi-Class Classification (MCC), Landuse Classification, Remote Sensing

Assessing Spatiotemporal Agreement between Multi-Temporal Built-up Land Layers and Integrated Cadastral and Building Data

GIScience 2016 Short Paper Proceedings Assessing Spatiotemporal Agreement between Multi- Temporal Built-up Land Layers and Integrated Cadastral and Building Data Johannes H. Uhl 1 , Stefan Leyk 1 , Aneta J. Florczyk 2 , Martino Pesaresi 2 , Deborah Balk 3 University of Colorado Boulder, Department of Geography, Boulder, CO 80309, U.S.A. Email: {johannes.uhl; stefan.leyk}@colorado.edu European Commission – Joint Research Centre (JRC), Institute for the Protection and Security of the Citizen (IPSC), Global Security and Crisis Management Unit, 21027 Ispra, Italy Email: {martino.pesaresi; aneta.florczyk}@jrc.ec.europa.eu City University of New York, Institute for Demographic Research and Baruch College, New York, NY 10010, U.S.A. Email: deborah.balk@baruch.cuny.edu Abstract There is an increasing availability of multi-temporal land use and built-up land datasets. However, little research has been done regarding the spatiotemporal uncertainty of these data products. In this work we present an approach that has the potential to be applicable for spatiotemporal evaluation of the novel Global Human Settlement Layer (GHSL) created by automatic classification of global collections of Landsat data recorded in the epochs 1975, 1990, 2000, and 2014. The proposed approach produces the reference data by integrating publicly available parcel and building data and derives agreement statistics with the GHSL. 1. Introduction The Global Human Settlement Layer (GHSL) project aims to assess the human presence in the planet by estimating the amount of built-up area based on remote sensing data and census data (Pesaresi et al. 2015). In Pesaresi et al. (2013) the GHSL information production workflow was tested for sensors ranging between 0.5 and 10m spatial resolution, which usually perform very well in detection of built-up areas but are typically constrained regarding data access, redistribution rights, and are typically not available consistently for long periods of time. For these reasons, these data are difficult to use for spatiotemporally uniform and systematic extraction of built-up areas on global, national or even regional level. Therefore, the GHSL workflow was tested with global collections of publicly available Landsat imagery collected in the past 40 years (Pesaresi et al. 2016). The Landsat GHSL dataset is available as seamless global mosaic at high spatial resolution (approx. 38m) and for various points in time (1975, 1990, 2000, 2014, see Figure 1a). GHSL data offers promising opportunities for population projections, disaster management and risk assessment (Freire et al. 2015; Freire et al. 2016), as well as for analysing and modelling urban dynamics and land use change. Before such novel data products can be made available to the research community, an extensive quality assessment is required. However, such assessments are difficult due to the lack of reliable reference data particularly for earlier time periods and in less developed regions. In this paper we present and discuss a novel approach to evaluate multi-temporal spatial data on built-up land such as GHSL or developed land cover classes using publicly available parcel (cadastral) data integrated with building footprints in the U.S. The aim of this study is to establish a protocol for future testing the accuracy of fine-scale multi-temporal products derived from automatic classification of satellite data featuring the presence of built-up areas. The selection of the test areas discussed here are driven and constrained by the availability of the reference data and consequently the results derived are

Satellite imagery and airborne geophysics for geologic mapping of the Edembo area, Eastern Hoggar (Algerian Sahara)

Satellite imagery combined with airborne geophysical data and field observations were employed for new geologic mapping of the Edembo area in the Eastern Hoggar (Tuareg Shield, Sahara). Multi-spectral band fusion, filtering, and transformation techniques, i.e., band combination, band-rationing and principal component analysis of ETM+ and ASTER data are used for better spectral discrimination of the different rocks units. A thematic map assessed by field data and available geologic information is compiled by supervised classification of satellite data with high overall accuracy (>90%). The automated extraction technique efficiently aided the detection of the structural lineaments, i.e., faults, shear zones, and joints. Airborne magnetic and Gamma-ray spectrometry data showed the pervasiveness of the large structures beneath the Paleozoic sedimentary cover and aeolian sands. The aeroradiometric K-range is used for discrimination of the high-K granitoids of Djanet from the peralumineous granites of Edembo, and to verify the Silurian sediments with their high K-bearing minerals. The new geological map is considered to be a high resolution improvement on all pre-existing maps of this hardly accessible area in the Tuareg Shield. Integration of the airborne geophysical and space-borne imagery data can hence provide a rapid means of geologically mapping areas hitherto poorly known or difficult to access.

Determining vertical leakage from the Great Artesian Basin, Australia, through up-scaling field estimates of phreatic evapotranspiration

Summary Understanding the water balance of large groundwater systems is fundamental for the sustainable management of the resource. The vertical leakage (i.e. discharge to upper aquifers or the unconfined water table) component of the Great Artesian Basin (GAB) is an example of a poorly constrained but large component of the water balance of Australia’s largest groundwater resource. Field estimates of phreatic evapotranspiration (ET) were made at discharge zones along the southwestern margin of the GAB using eddy covariance station and micro-lysimeter measurements, and inversion of chloride/isotope soil profile measurements. The field estimates were assigned to three major landforms associated with areas of increasingly higher evaporative discharge and progressively decreasing depths to the water table. These landforms were mapped using remote sensing and digital elevation data, with characteristically higher soil moisture, salt precipitation, and lower surface temperature compared to areas distal to discharge zones. Based on the field measurements, broad ranges of phreatic ET (0.5–10, 10–100 and 100–300 mm y −1 ) were assigned to the major land-types. The higher phreatic ET discharge zones mapped by supervised classification of satellite data are 8–28% of the total regional vertical leakage component estimated by numerical modelling of the GAB. In comparison, the higher discharge zones estimated by landform mapping are 73–251% of the total vertical leakage component estimated by modelling. The mapped distribution of the high discharge areas has important implications for modelling of the GAB. In the western sub-basin, most of the estimated recharge can be accounted for by phreatic ET in the high discharge zones located around the Basin margins, implying that vertical leakage rates distal to the margins are very low and that discharge may exceed current recharge. In contrast, the results for the eastern sub-basin suggest that vertical leakage rates around the South Australian portion of the Basin margin are low and that more of the vertical leakage component in the eastern sub-basin is occurring distal to the Basin margins. Consequently, the pathways for vertical leakage in the eastern sub-basin are likely to be more complex than for the western sub-basin.