In this paper, we propose a novel hybrid method that combines image processing techniques with deep learning to tackle the problem of detecting date palm trees from high-resolution satellite imagery. The image processing component of the method is based on simple operations such as combining spectral bands, filtering, and peak detection and is used to produce an initial set of detections. Then, a simple convolutional neural network (CNN) trained with a manually labeled dataset is used to refine the initial set of detections. The method was tested on a dataset containing high-resolution 4-band WorldView-3 imagery of date palm tree farms of varying degrees of complexity in the Al-Ahsa region in Saudi Arabia. Comparisons of the proposed method with a Mask R-CNN show that the proposed method can improve the F1-Score on average by 7.8%, especially in dense and unorganized date palm tree farms commonly found in the Al-Ahsa region. Finally, to demonstrate the applicability of the method, the proposed method was used to estimate the number of palm trees within Al-Ahsa Oasis, the largest self-contained oasis in the world.

Since the first spacecraft were deployed for earth observation (EO) in the early 1960s, technological change in EO systems has significantly improved remote sensing capabilities. To fully understand the pace of past change and develop empirical baselines for future projections, it is important to quantitatively characterize performance of key instruments used in EO missions. This study presents trends in performance in EO instruments using empirical data from spacecraft and instruments deployed from 1959-2022. Parameterized models are derived to quantitatively determine temporal progression in key figures-of-merit. Pareto frontiers of mass and resolution of sounders, SAR, and radiometers are also determined. The results show a consistent shift in Pareto frontiers of radiometers across successive 15-year periods spanning 1972 to 2025.

There are few frameworks to quantitatively determine value of remote sensing data to inform decisions related to sustainable development. In most cases, value remains vaguely defined and qualitatively discussed. Here, we develop an approach in which decisions, that can be supported by remote sensing data, are systematically structured, and monetized value of possible outcomes are used as a basis for quantifying value of remote sensing. It is shown that a canonical set of decision structures and application of a Bayesian approach leads to a set of analytic expressions that can be used to build insights of the decision problem and to identify driving factors that affect value of remote sensing. The method is demonstrated for water quality monitoring, forest management, and solar power infrastructure inspection. The specific case studied for water quality showed that remote sensing provides value if likelihood of a HAB is 30% or more (even if data quality is not high). In forest management (with specific restoration costs, mangrove loss costs etc.), remote sensing offers value beyond a threshold of cost of field surveys. For solar farm inspection, remote sensing is most valuable for low anomaly rates and high panel replacement costs.

Earth observation (EO) activities from space have rapidly advanced with a five-fold increase in average annual launches during 2011–2021 as compared to the previous decade. With increasing interest in public and private sectors for charting out strategies and roadmaps for future missions, it is important to track technology trends. Here, a quantitative assessment of key advancement in EO capabilities is presented along with projections for 2030. The analysis shows that, if the trends continue as-is, the number of countries owning (and engaging with) EO spacecraft will expand to 64 by 2030 which will be a 42% increase from 2020 (of 45 countries). Additionally, minimum spatial resolution of imagers on scientific and commercial EO spacecraft will fall below 10cm, and below 30cm for Synthetic Aperture Radars.

In this paper, we demonstrate a self-supervised deep learning pipeline that can effectively learn to detect vehicles without using any pre-labeled training data. The pipeline uses a morphological vehicle detection algorithm to automatically generate training sets for a convolutional neural network (CNN). We tested this methodology on a mixed-use urban neighborhood in Riyadh, Saudi Arabia using 0.31-meter multispectral Worldview-3 satellite imagery with eight bands in the visible and near-infrared wavelengths. This method leverages the class imbalance inherent to many vehicle detection problems by generating a balanced training sample from a high-precision, low-recall morphological model to train a neural network to identify general vehicle characteristics. This approach is built on broadly applicable image processing methods and, with appropriate adjustments, might be adapted to high-resolution from various satellite or aerial sources.

High resolution, high frequency, multi-spectral remotely-sensed Earth Observation (EO) data is increasingly available. Challenges of ensuring radiometric data quality and consistency of data products, however, have not been widely recognized. As new products and applications rapidly emerge, and decisions linked to public safety, well-being, and environmental assessments are at stake, the issue of EO data quality and quantification of errors and their impacts has gained increasing salience. This paper presents a generalizable decision-theoretic framework for quantifying the value of data quality (partly through improved calibration), and demonstrates its application for a water quality monitoring case. Results from the particular case of freshwater monitoring show that after 30% or more probability of occurrence of a harmful algal bloom, remote sensing is useful to have regardless of the data quality level. However, factors such as cost of remote sensing data (in this case, comparing internal processing vs. external processing of tasked satellite imagery) will affect the threshold at which remote sensing value is positive regardless of data quality level.

Due to recently growing adoption, residential photovoltaic (PV) installations are becoming a key contribution to renewable energy production. In order to efficiently track the inherently decentralized deployment of these PV systems, this study leverages widely available high-resolution satellite imagery, along with the demonstrated ability of convolutional neural networks (CNNs) in image classification tasks. In particular, this effort presents development of a custom-trained CNN that operates on images of residential areas received as part of a larger image processing pipeline. This custom-trained CNN is shown to achieve comparable accuracy to the state-of-the-art achieved via transfer learning, but with reduced computational burden and required image resolution. Using imagery from two cities in California, this approach achieves PV classification with a precision of 91.9% and recall of 92.4%. This accuracy is sufficient to inform actionable insights from satellite imagery regarding the political, social, and economic factors affecting PV deployment.

Earth observation systems are providing a growing amount of high resolution and multi-spectral data including agricultural crop yield predictions, local weather forecasts, wild fire tracking, and water quality monitoring. With the advent of multiple elements and sophisticated processing and modeling, there are also increasing avenues for introduction of errors. It is important to quantify and characterize these errors in the remotely sensed data as it gets increasingly used to inform vital decisions. Here, a data value-chain approach is presented for conceptualizing introduction and propagation of errors in remote sensing data acquisition, processing, and decisions. A detailed case of calibration errors and their propagation to higher level data products is then discussed. An NDVI analysis is used as an example, and it is shown (for the selected region), that a 3% error in reflectance ratio of red and NIR bands translates into a difference of 60 % for category 5 NDVI classification (representing high vegetation). This amplification of errors along the data value chain is caused by nonlinear operators such as the application of quotients to differences of reflectance values as well as the use of classifiers based on fixed color thresholding.

The Inclusive Wealth Index has been proposed as a suitable measurement of sustainable growth and development of nations. This index consists of natural, produced, and human capital. A systematic and verifiable assessment of natural capital across countries has proven to be challenging. This paper proposes that natural capital be evaluated in a consistent manner through the use of earth observation satellites. 96 and 110 instruments were identified to observe forest and agricultural resources, respectively, thus a strong capability exists for these natural capital categories. However, current remote sensing capabilities can only go so far. Only 20 instruments were identified that can potentially be used for fish stock assessment. Although 72 instruments were identified to have mineral observation capabilities, data post processing is needed to accurately assess mineral stocks. 35 instruments were identified to assist in fossil fuel observation, but given that supply data exists for fossil fuels, this capability gap is not considered critical. By examining these capability gaps, scientists can direct their attention to payloads that will enable remote sensing technologies to assist in the measurement of the Inclusive Wealth Index.

New earth observation missions are being implemented and conceptualized with architectures employing multiple spacecraft and novel observing strategies that combine distributed, multi-platform systems. These new types of missions are enabling earth observation with multi-angular, multi-spectral data acquisition at high resolution and high revisit frequencies. The architectural choices have grown exponentially in such distributed systems, and there is a need for quantitative measures, that go beyond cost estimation, and assess value (or scientific return) over a mission's operational life-cycle. Here, we formulate a novel metric, Net Architecture Value (NAV), that can be used in early stage conceptual mission design and architecture trade studies. We propose that useful data of adequate quality, obtained over regions of interest, acquired by a system over its lifetime can be used as a proxy measure of value. We demonstrate the application of this approach for a distributed space mission.

The availability of high-resolution remote sensing data has created opportunities to gain new information about urban infrastructure. Here, high resolution ortho-imagery with three bands and maximum likelihood classification of images is used to obtain information about roads and pavement (i.e. transportation related) infrastructure. The cities of Phoenix, Arizona (for the years 2004, 2006, 2008, 2012) and Seattle, Washington (for 2002, 2005, 2009) were analyzed. We discuss the implications and utility of this (and other infrastructure) data for constructing metrics for urban growth and environmental sustainability.

In Earth Science and Telecommunications from space we are now transitioning from phase 1 (single monolithic satellites in GEO or LEO) to phase 2 which consists of distributed ensembles of LEO and GEO satellites. This however is not the end game. The third phase will be the manufacturing and assembly of satellites directly in space, allowing significantly larger apertures and orders of magnitude improvement in spatial resolution. The key challenges and opportunities of this paradigm shift are summarized and quantified.

While there is growing interest in implementing future NASA Earth Science missions as Distributed Spacecraft Missions (DSMs), there are currently very few tools available to help in the design of DSMs. The objective of our project is to provide a framework that facilitates DSM Pre-Phase A investigations and optimizes DSM designs with respect to a-priori Science goals. Our Trade-space Analysis Tool for Constellations (TAT-C) enables the investigation of questions such as: “Which type of constellations should be chosen? How many spacecraft should be included in the constellation? Which design has the best cost/risk value?”. This paper provides a description of the TAT-C tool and its components.

Gross primary productivity is an excellent metric of how much forests act as carbon dioxide sinks but currently have up to 40% uncertainty in their global estimates. A large proportion of the uncertainty has been attributed to artifacts in the sun-sensor geometry of monolithic spacecrafts leading to insufficient sampling of the bi-directional reflectance of vegetation. This paper proposes to use small satellite clusters with spectrometers as a new measurement solution to improve angular sampling locally and scale up measurements globally. Initial observing system simulations with four satellites launched as secondary payloads via the ISS and operating in different imaging modes show error estimates of less than 12% when compared to dense airborne measurements, a 50% improvement to the worst case error produced by corresponding monoliths.

A modern and efficient information and communication technology (ICT) infrastructure is essential for managing the challenges in the complex urban systems development. The ICT infrastructure is a complex system consisting of many subsystems and interconnections, which makes the process of planning, designing, and maintaining a comprehensive ICT infrastructure expensive and difficult. Most approaches used for the ICT infrastructure modeling focus typically on a single ICT system, for example, a wireless network. This paper presents a systems modeling approach based on integrating different subsystems and their characteristics into a single model, applying system decomposition, establishing the logical relations between system components, and defining relevant key performance indicators. It is shown that this systems modeling approach facilitates holistic planning, design, and evaluation of the complex ICT infrastructure for a sustainable city. This is demonstrated in the form of a two-scenario Masdar city case study. The case study exhibits the practicality of the derived ICT model and the feasibility of the results.

When computer-based tools are used for analysis of complex systems, the design of user interactions and interfaces becomes an essential part of development that determines the overall quality. The objective of this study is to investigate the processes and results of user interactions with integrated analysis tools to synthesize design implications for future tool development. In this study, two space exploration logistics tools are compared in a controlled user experiment. Through a comparative usability analysis, this study evaluated user performance and perception to provide design implications for future integrated analysis tools. For a comprehensive evaluation, multiple methods were used for data collection, including observation, questionnaire and interview. In addition to a result-oriented performance analysis, a process-oriented approach was used for analyzing patterns in user behaviors and errors. Results are presented with reference to the related features embedded in the interfaces of the two tools. Based on the comparative results, synthesized design insights for hierarchical structure, model transparency, automation, and visualization and feedback are discussed for integrated analysis tools in general.

We propose a medication reminder system which supports users' medication compliance. Poor compliance of medication causes significant problems including worsening disease and increase of healthcare costs. One common reason for poor compliance is patients' forgetfulness. We identified four types of failures caused by forgetfulness which requires the system to consider various users' behaviors other than users' consumption of medication, while existing medication reminder systems consider only users' consumption of medication. The architecture of the system is designed to work with various sensors and actuators, which enable the system to consider vast variety of users' behaviors. We describe the design process of the system initiated by the identifying failures. One example of implementation is also shown including the development of sensors and actuators.