UPDATES AND IMPROVEMENTS ON ACCURACY TO CALIBRATE CUBESAT RADIOMETER CONSTELLATIONS

L. Apudo¹, M. Aksoy¹, J. Bradburn¹

¹University at Albany

CubeSat-based radiometer constellations are increasingly being used for science missions as they provide unprecedented spatial coverage, shorter revisit times in a cost-effective way. One of the biggest challenges in using CubeSat constellations, however, is the decrease in accuracy and increase in uncertainty in the calibrated measurements due to sensitivity to ambient conditions and infrequent external calibration. In an effort to overcome this challenge, a software-based framework called “Adaptive Calibration of CubeSat Radiometer Constellations (ACCURACy)” has been developed [1].

ACCURACy has been built to intercalibrate homogeneous constellations of CubeSat radiometers in real time. The framework utilizes the relationship between the gain of the individual radiometers in the constellation and their telemetry data such as physical temperature measurements recorded by thermistors placed in various parts of the CubeSats. Telemetry data are used to identify and cluster constellation radiometers in similar states into time-adaptive groups. Calibration data are shared within each such group to determine cluster-level gains, which are used to calibrate all radiometers in the clusters.

ACCURACy runs behind a graphical user interface (GUI) that has been developed using MATLAB’s App Builder and uses a linear least squares regression approach to calibrate measured brightness temperatures in each cluster. On the other hand, we are currently building the next version of the GUI which uses a convolutional neural network (CNN) calibration model instead of linear regression. A CNN calibration algorithm has been trained and evaluated using synthetic and real radiometer data which include steady-state and power cycling induced transient measurements. The performance of the CNN approach was compared to that of a linear least squares regression-based calibration method, and significant improvements have been noted. Such CNN models will be incorporated into the new version of ACCURACy. Furthermore, in the interest of generalizing the framework, we are developing methodologies for expanding the ACCURACy framework to intercalibrate heterogeneous radiometer constellations where the relationship between telemetry data and radiometer gain characteristics may be different for each constellation member.

In this presentation, we will present the updates in the ACCURACy framework regarding the integration of CNN calibration algorithms and capabilities to calibrate heterogeneous constellations. The performance of ACCURACy will be compared to the state-of-the-art constellation intercalibration methods to highlight the improvements in accuracy and precision.

[1] Bradburn, J., Aksoy, M., Apudo, L., Vukolov, V., Ashley, H., & VanAllen, D. (2025). ACCURACy: A Novel Calibration Framework for CubeSat Radiometer Constellations. Remote Sensing, 17(3), 486.