Smoke plumes are the first things seen from space when wildfires occur. Thus, fire smoke detection is important for early fire detection. Deep Learning (DL) models have been used to detect fire smoke in satellite imagery for fire detection. However, previous DL-based research only considered lower spatial resolution sensors (e.g., Moderate-Resolution Imaging Spectroradiometer (MODIS)) and only used the visible (i.e., red, green, blue (RGB)) bands. To contribute towards solutions for early fire smoke detection, we constructed a six-band imagery dataset from Landsat 5 Thematic Mapper (TM) and Landsat 8 Operational Land Imager (OLI) with a 30-metre spatial resolution. The dataset consists of 1836 images in three classes, namely “Smoke”, “Clear”, and “Other_aerosol”. To prepare for potential on-board-of-small-satellite detection, we designed a lightweight Convolutional Neural Network (CNN) model named “Variant Input Bands for Smoke Detection (VIB_SD)”, which achieved competitive accuracy with the state-of-the-art model SAFA, with less than 2% of its number of parameters. We further investigated the impact of using additional Infra-Red (IR) bands on the accuracy of fire smoke detection with VIB_SD by training it with five different band combinations. The results demonstrated that adding the Near-Infra-Red (NIR) band improved prediction accuracy compared with only using the visible bands. Adding both Short-Wave Infra-Red (SWIR) bands can further improve the model performance compared with adding only one SWIR band. The case study showed that the model trained with multispectral bands could effectively detect fire smoke mixed with cloud over small geographic extents.

The accuracy of surface reflectance estimation for satellite sensors using radiance-based calibrations can depend significantly on the choice of solar spectral irradiance (or solar spectrum) model used for atmospheric correction. Selecting an accurate solar spectrum model is also important for radiance-based sensor calibration and estimation of atmospheric parameters from irradiance observations. Previous research showed that Landsat 8 could be used to evaluate the quality of solar spectrum models. This paper applies the analysis using five previously evaluated and three more recent solar spectrum models using both Landsat 8 (OLI) and Landsat 9 (OLI2). The study was further extended down to 10 nm resolution and a wavelength range from Ultraviolet A (UVA) to shortwave infrared (SWIR) (370–2480 nm) using inversion of field irradiance measurements. The results using OLI and OLI2 as well as the inversion of irradiance measurements were that the more recent Chance and Kurucz (SA2010), Meftah (SOLAR-ISS) and Coddington (TSIS-1) models performed better than all of the previous models. The results were illustrated by simulating dark and bright surface reflectance signatures obtained by atmospheric correction with the different solar spectrum models. The results showed that if the SA2010 model is assumed to be the “true” solar irradiance, using the TSIS-1 or the SOLAR-ISS model will not significantly change the estimated ground reflectance. The other models differ (some to a large extent) in varying wavelength areas.

Rangeland ecosystems in the western United States are vulnerable to climate change, fire, and anthropogenic disturbances, yet classification of rangeland areas remains difficult due to frequently sparse vegetation canopies that increase the influence of soils and senesced vegetation, the overall abundance of senesced vegetation, heterogeneity of life forms, and limited ground-based data. The Rangeland Condition Monitoring Assessment and Projection (RCMAP) project provides fractional vegetation cover maps across western North America using Landsat imagery and artificial intelligence from 1985 to 2023 at yearly time-steps. The objectives of this case study are to apply hyperspectral data from several new data streams, including Sentinel Synthetic Aperture Radar (SAR) and Earth Surface Mineral Dust Source Investigation (EMIT), to the RCMAP model. We run a series of five tests (Landsat-base model, base + SAR, base + EMIT, base + SAR + EMIT, and base + Landsat NEXT [LNEXT] synthesized from EMIT) over a difficult-to-classify region centered in southwest Montana, USA. Our testing results indicate a clear accuracy benefit of adding SAR and EMIT data to the RCMAP model, with a 7.5% and 29% relative increase in independent accuracy (R²), respectively. The ability of SAR data to observe vegetation height allows for more accurate classification of vegetation types, whereas EMIT’s continuous characterization of the spectral response boosts discriminatory power relative to multispectral data. Our spectral profile analysis reveals the enhanced classification power with EMIT is related to both the improved spectral resolution and representation of the entire domain as compared to legacy Landsat. One key finding is that legacy Landsat bands largely miss portions of the electromagnetic spectrum where separation among important rangeland targets exists, namely in the 900–1250 nm and 1500–1780 nm range. Synthesized LNEXT data include these gaps, but the reduced spectral resolution compared to EMIT results in an intermediate 18% increase in accuracy relative to the base run. Here, we show the promise of enhanced classification accuracy using EMIT data, and to a smaller extent, SAR.

Detecting fire smoke is crucial for the timely identification of early wildfires using satellite imagery. However, the spatial and spectral similarity of fire smoke to other confounding aerosols, such as clouds and haze, often confuse even the most advanced deep-learning (DL) models. Nonetheless, these aerosols also present distinct spectral characteristics in some specific bands, and such spectral patterns are useful for distinguishing the aerosols more accurately. Early research tried to derive various threshold values from the reflectance and brightness temperature in specific spectral bands to differentiate smoke and cloud pixels. However, such threshold values were determined based on domain knowledge and are hard to generalise. In addition, such threshold values were manually derived from specific combinations of bands to infer spectral patterns, making them difficult to employ in deep-learning models. In this paper, we introduce a DL module called input amplification (InAmp) which is designed to enable DL models to learn class-specific spectral patterns automatically from multi-spectral satellite imagery and improve the fire smoke detection accuracy. InAmp can be conveniently integrated with different DL architectures. We evaluate the effectiveness of the InAmp module on different Convolutional neural network (CNN) architectures using two satellite imagery datasets: USTC_SmokeRS, derived from Moderate Resolution Imaging Spectroradiometer (MODIS) with three spectral bands; and Landsat_Smk, derived from Landsat 5/8 with six spectral bands. Our experimental results demonstrate that the InAmp module improves the fire smoke detection accuracy of the CNN models. Additionally, we visualise the spectral patterns extracted by the InAmp module using test imagery and demonstrate that the InAmp module can effectively extract class-specific spectral patterns.