This paper presents a method for detecting spacecraft anomalies using Long Short-Term Memory (LSTM) networks and nonparametric dynamic thresholding. The authors demonstrate the effectiveness of LSTMs in overcoming challenges in spacecraft telemetry anomaly detection, including interpretability, scale, precision, and complexity. They also propose a complementary unsupervised and nonparametric anomaly thresholding approach, along with strategies for mitigating false positives. The study uses expert-labeled telemetry anomaly data from the Soil Moisture Active Passive (SMAP) satellite and the Mars Science Laboratory (MSL) rover, Curiosity. The authors evaluate the performance of their approach using real-world data from Incident Surprise, Anomaly (ISA) reports. They also highlight key milestones, improvements, and observations from an early implementation of the system for the SMAP mission. The paper discusses the limitations of current anomaly detection methods, which often rely on predefined thresholds and manual analysis. These methods are limited by the need for expert knowledge and the inability to handle complex, high-dimensional data. The authors propose a novel approach that uses LSTMs to predict telemetry data and a nonparametric thresholding method to detect anomalies. The study shows that the LSTM-based approach achieves high prediction performance while maintaining interpretability. The nonparametric thresholding method is used to evaluate residuals and address issues related to diversity, non-stationarity, and noise in data streams. The authors also discuss strategies for mitigating false positives, including pruning anomalous sequences and using historical anomaly data to improve system performance. The results show that the LSTM-based approach with nonparametric thresholding achieves high precision and recall in detecting spacecraft anomalies. The method is particularly effective in detecting contextual anomalies, which require temporal context and are harder to detect. The authors also compare their approach to a parametric thresholding method and find that the nonparametric approach performs better in terms of precision and recall. The paper concludes that the proposed method is a promising approach for detecting spacecraft anomalies and has the potential to be implemented in a wide range of applications involving multivariate time series data. The authors also make public a large real-world, expert-labeled set of anomalous spacecraft telemetry data and offer open-source implementations of the methodologies presented in this paper.This paper presents a method for detecting spacecraft anomalies using Long Short-Term Memory (LSTM) networks and nonparametric dynamic thresholding. The authors demonstrate the effectiveness of LSTMs in overcoming challenges in spacecraft telemetry anomaly detection, including interpretability, scale, precision, and complexity. They also propose a complementary unsupervised and nonparametric anomaly thresholding approach, along with strategies for mitigating false positives. The study uses expert-labeled telemetry anomaly data from the Soil Moisture Active Passive (SMAP) satellite and the Mars Science Laboratory (MSL) rover, Curiosity. The authors evaluate the performance of their approach using real-world data from Incident Surprise, Anomaly (ISA) reports. They also highlight key milestones, improvements, and observations from an early implementation of the system for the SMAP mission. The paper discusses the limitations of current anomaly detection methods, which often rely on predefined thresholds and manual analysis. These methods are limited by the need for expert knowledge and the inability to handle complex, high-dimensional data. The authors propose a novel approach that uses LSTMs to predict telemetry data and a nonparametric thresholding method to detect anomalies. The study shows that the LSTM-based approach achieves high prediction performance while maintaining interpretability. The nonparametric thresholding method is used to evaluate residuals and address issues related to diversity, non-stationarity, and noise in data streams. The authors also discuss strategies for mitigating false positives, including pruning anomalous sequences and using historical anomaly data to improve system performance. The results show that the LSTM-based approach with nonparametric thresholding achieves high precision and recall in detecting spacecraft anomalies. The method is particularly effective in detecting contextual anomalies, which require temporal context and are harder to detect. The authors also compare their approach to a parametric thresholding method and find that the nonparametric approach performs better in terms of precision and recall. The paper concludes that the proposed method is a promising approach for detecting spacecraft anomalies and has the potential to be implemented in a wide range of applications involving multivariate time series data. The authors also make public a large real-world, expert-labeled set of anomalous spacecraft telemetry data and offer open-source implementations of the methodologies presented in this paper.