An algorithm for snow and ice detection using AVHRR data An extension to the APOLLO software package

Abstract

The AVHRR (Advanced Very High Resolution Radiometer) Processing scheme Over Land, cLbud and Ocean (APOLLO) is used to extract surface and cloud parameters from satellite data. Before these parameters can be computed, it is necessary to distinguish between land and ocean surfaces and to apply algorithms for the detection of partially cloudy and cloud-filled pixels. In regions with seasonal or permanent snow and ice coverage the separation of clouds becomes much more difficult or often impossible. For this reason, and to find cloud-free and partly cloudy snow and ice pixels,- a day-time algorithm has been developed which uses all five AVHRR channels as follows: The threshold testing of the reflected part of channel-3 radiance leads to a definite distinction between snow/ice and water clouds due, to the clouds much higher reflectivity at 3.7 μm. The detection; of sea ice is based on threshold tests of visible reflectances and, in particular, of the temperature difference between channels-4 and -5. Snow is identified if a high visible reflectance is combined with a low reflectance in channel-3 and with a ratio of channel-2 to channel-1 reflectances similar to that of a cloud. The latter criterion is also mostly suitable to distinguish between snow-covered and snow-free ice areas. Some tests of this algorithm applied to AVHRR data from the 1987 Baltic Sea ice season have shown reasonable classification results with the exception of a few areas with ice clouds or with ice topped water clouds.     

基于类内拟合的遥感影像薄云雾校正方法

光学遥感观测极易受到云雾影响,降低数据质量并限制其后续应用潜力。由此,提出了一种基于类内拟合的遥感影像薄云雾校正方法。首先,利用滑动窗口逐波段地搜索局部最小值,称之为暗目标,通过拟合不同波段的暗目标样本估计出薄云雾辐射的相关性。基于此,联合云雾波段相关性与成像模型,生成不含云雾干扰的合成假彩色影像,利用K均值分类自动得到地表覆被类型。利用地类信息,进一步选取晴空区像元获取不同地类在不同波段对间的线性关系。最后,将上述两种线性关系进行联立,求解出各地表类型在不同波段上的值,从而完成影像校正。通过模拟与真实实验对方法有效性和场景适用性进行测试,并从定性目视与定量评估两方面对结果进行检验。实验结果表明：提出方法可有效去除薄云雾干扰,适用于不同地表覆被类型场景,获得高光谱保真的校正地表。     

山区可见光—近红外遥感影像浓密植被暗像元自动识别方法研究

暗目标法是目前气溶胶光学厚度遥感反演中应用最为广泛的方法,浓密植被暗像元的识别是暗目标法的基础。针对可见光—近红外影像缺少中红外波段难以有效识别浓密植被暗像元的问题,引入红波段直方图阈值法识别山区可见光—近红外影像的浓密植被暗像元。该方法利用浓密森林像元在可见光波段反射率低的特点,通过搜索红波段直方图的最小峰值自动识别浓密植被暗像元。试验中选取Landsat TM影像前4个波段利用红波段直方图阈值法识别可见光—近红外影像的浓密植被暗像元,并与在中红外波段影像和可见光—近红外影像中广泛应用的两种暗像元识别方法进行对比分析,探讨红波段直方图阈值法的有效性,最后将该方法应用于环境减灾卫星(HJ-1)CCD影像的暗像元识别和气溶胶反演。实验结果表明:红波段直方图阈值法明显优于常用的可见光—近红外影像暗像元识别方法,识别精度接近传统的中红外波段影像识别方法,相似度指数小于2和小于3的暗像元分别为83.12%和93.48%。该方法为山区可见光—近红外影像浓密植被暗像元自动识别提供了一种新的适用方法,识别结果能够满足暗目标法反演气溶胶光学厚度的要求。     

A correlation based cloud-detection algorithm combined with an examination of some split-window assumptions

Twenty-six daytime NOAA-11 AVHRR images covering the Danish waters are analysed together with 99 cloud-free bulk temperature measurements. The images are cloud screened with a cloud-detection algorithm, which in excess of the usual threshold and variation tests consists of: a min_max-routine which among suspicious pixels points out the cloud contaminated pixels, and a separation of cloud pixels from cold water pixels by means of the correlation coefficient between the channel 4 brightness temperature and the channel 4-channel 5 brightness temperature. In the cloud screened images, the channel 4-channel 5 brightness temperatures do not increase with the satellite zenith angle and do not seem to carry any significant information about the true sea surface temperature (SST). Local meteorological and oceanographic effects are believed to be more important than the channel 4-channel 5 brightness temperatures.     

Aerosol retrieval over ocean from SEVIRI for the use in GERB Earth''s radiation budget analyses

To study the effect of aerosols on the Earth's radiation budget (ERB), the Royal Meteorological Institute of Belgium (RMIB) has integrated spectral aerosol optical depth (AOD) measurements over the ocean from the Spinning Enhanced Visible and Infra-Red Scanner (SEVIRI) into its Geostationary Earth's Radiation Budget, or GERB, processing system referred to as the RGP. Aerosols affect the ERB both directly (when radiation interacts with an aerosol particle) and indirectly (when aerosols act as cloud condensation nuclei). Quantifying the indirect effect is challenging as it requires accurate aerosol retrievals in the close proximity to clouds, where aerosol retrievals may be biased due to leakages from the cloud mask (CM). The initial focus of the RGP project was on the direct effect using confidently clear scenes.

A single channel CM exploiting the SEVIRI temporal sampling was developed at the RMIB for the use in the RGP project. In this study, that single channel mask was evaluated against two multi-channel CMs, one from the Meteorological Products Extraction Facility (MPEF) at the European Organization for the Exploitation of Meteorological Satellites (EUMETSAT), and the other from the Satellite Application Facility for Supporting NoWCasting and Very Short Range Forecasting (SAFNWC), respectively. The NOAA/NESDIS Advanced Very High Resolution Radiometer (AVHRR) single channel aerosol algorithm was adjusted to SEVIRI spectral bands and consistently applied to the pixels identified as cloud-free. The aerosol products corresponding to the three CMs were compared, and the RMIB CM was found to be sufficiently accurate and conservative, for RGP applications.

Comparisons with independent AODs derived from the MODerate resolution Imaging Spectroradiometer (MODIS) onboard Terra and Aqua satellites show that the RMIB CM-based SEVIRI aerosol product compares well with its MODIS counterpart. However, a small fraction of cloud-contaminated pixels may still remain in the SEVIRI AOD imagery, chiefly within one to two SEVIRI pixels of the cloud boundary, thus limiting its use for indirect forcing studies. Also, the RMIB CM may screen high AOD non-dust aerosol events (e.g., smoke from biomass burning) as cloud. The potential of the new SEVIRI aerosol product is illustrated by generating 9 km-resolution seasonal maps of AODs and ´ÅǺngström Exponents, and by using the GERB radiative flux measurements for a preliminary quick assessment of the direct aerosol forcing.     

Physical and statistical approaches for cloud identification using Meteosat Second Generation-Spinning Enhanced Visible and Infrared Imager Data

In this paper a cloud detection algorithm applied to the MSG-SEVIRI (Meteosat Second Generation-Spinning Enhanced Visible and Infrared Imager) data is described. In order to obtain a good performance in cloud detection, physical, statistical and temporal approaches have been used. In the statistical algorithm, the spectral and textural features of the MSG-SEVIRI images have been used as input, while, in the physical tests, a set of dynamic thresholds has been used. The physical algorithm does not use real time ancillary data— such as sea surface temperature map and NWP temperature and humidity profiles. A further test is applied to that pixels having low confidence to be clear or cloudy. This test takes advantage of the best MSG-SEVIRI temporal resolution and it applies the K-Nearest Neighbour classifier to the spectral and textural features calculated in “temporal” boxes 3 × 3 pixels, defined “temporal” because their elements belong to three subsequent MSG-SEVIRI images. The MACSP (cloud MAsk Coupling of Statistical and Physical methods) algorithm has been validated against the MODIS cloud mask and compared with CPR (Cloud Profiling Radar) and SAFNWC cloud masks. The outcomes show that the MACSP detects 91.8% of the total number of the pixels used for validation against MODIS cloud mask correctly, while the SAFNWC cloud mask detects 89.2% of them correctly. In particular, the MACSP classifies as cloudy 8.8% of the pixels classified by the MODIS cloud mask as clear, while the SAFNWC cloud mask classifies as cloudy 12.1% of them. The MACSP detects 91.2% of the cloudy CPR pixels and 90.8% of the cloud-free CPR pixels, considered for comparison, correctly. On the other hand, the SAFNWC and CPR cloud masks agree in the detection of 90.7% of the cloudy pixels and of 90.2% of the cloud-free pixels.     

Cloud detection in Landsat imagery of ice sheets using shadow matching technique and automatic normalized difference snow index threshold value decision

This work presents a new algorithm designed to detect clouds in satellite visible and infrared (IR) imagery of ice sheets. The approach identifies possible cloud pixels through the use of the normalized difference snow index (NDSI). Possible cloud pixels are grown into regions and edges are determined. Possible cloud edges are then matched with possible cloud shadow regions using knowledge of the solar illumination azimuth. A scoring index quantifies the quality of each match resulting in a classified image. The best value of the NDSI threshold is shown to vary significantly, forcing the algorithm to be iterated through many threshold values. Computational efficiency is achieved by using sub-sampled images with only minor degradation in cloud-detection performance. The algorithm detects all clouds in each of eight test Landsat-7 images and makes no incorrect cloud classifications.     

Calibration of the visible and near-infrared channels of NOAA-11 and -14 AVHRRs by using reflections from molecular atmosphere and stratus cloud

A new method to determine the calibration coefficients for visible and near-infrared channels of Advanced Very High Resolution Radiometer (AVHRR) aboard NOAA satellite is presented and applied to NOAA-11 and -14 spacecrafts. The method uses the reflections from clear-sky ocean and stratus clouds. The clear-sky data analysis gives a minimum estimate of the slope coefficient (albedo per count) for a target month by using radiative transfer theory for molecular atmosphere. Cloudy-sky pixels were precisely excluded from that analysis by using multi-spectral data of AVHRR. Neighbouring pixels of cloud were also excluded to avoid three-dimensional radiative effects such as cloud shadow. On the other hand, the optical thickness (at a visible wavelength) of summer stratus clouds was retrieved from nominally calibrated reflectance of respective visible and near-infrared channels. This analysis was performed to adjust the balance between the two-channels' calibration coefficients because if the two channels were correctly calibrated, the cloud optical thickness retrieved from the two channels must be the same. Finally, the calibration coefficients were determined using iteration.     

Improving continuity of MODIS terrestrial photosynthesis products using an interpolation scheme for cloudy pixels

The Moderate Imaging Spectroradiometer (MODIS) sensors onboard the NASA Terra and Aqua satellites provide the means for frequent measurement and monitoring of the status and seasonal variability in global vegetation phenology and productivity. However, while MODIS reflectance data are often interrupted by clouds, terrestrial processes like photosynthesis are continuous, so MODIS photosynthesis data must be able to cope with cloudy pixels. We developed cloud‐correction algorithms to improve retrievals of the MODIS photosynthesis product (PSNnet) corresponding to clear sky conditions by proposing four alternative cloud‐correction algorithms, which have different levels of complexity and correct errors associated with cloudy‐pixel surface reflectance. The cloud‐correction algorithms were applied at four weather stations, two fluxtower sites and the Pacific Northwest (PNW) region of the USA to test a range of cloud climatologies. Application of the cloud‐correction algorithms increased the magnitude of both daily and annual MODIS PSNnet results. Our results indicate that the proposed cloud correction methods improve the current MODIS PSNnet product considerably at both site and regional scales and weekly to annual time steps for areas subjected to frequent cloud cover. The corrections can be applied as a post‐processing interpolation of PSNnet, and do not require reprocessing of the MOD17A2 algorithm.     

Satellite data cloud detection using deep learning supported by hyperspectral data

ABSTRACT

Deep learning methods can play an important role in satellite data cloud detection. The number and quality of training samples directly affect the accuracy of cloud detection based on deep learning. Therefore, selecting a large number of representative and high-quality training samples is a key step in cloud detection based on deep learning. For different satellite data sources, choosing sufficient and high-quality training samples has become an important factor limiting the application of deep learning in cloud detection. This paper presents a fast method for obtaining high-quality learning samples, which can be used for cloud detection of different satellite data with deep learning methods. AVIRIS (Airborne Visible Infrared Imaging Spectrometer) data, which have 224 continuous bands in the spectral range from 400–2500 nm, are used to provide cloud detection samples for different types of satellite data. Through visual interpretation, a sufficient number of cloud and clear sky pixels are selected from the AVIRIS data to construct a hyperspectral data sample library, which is used to simulate different satellite data (such as data from the Moderate Resolution Imaging Spectroradiometer (MODIS) and Landsat Operational Land Imager (OLI) satellites) as training samples. This approach avoids selecting training samples for different satellite sensors. Based on the Keras deep learning framework platform, a backpropagation (BP) neural network is employed for cloud detection from Landsat 8 OLI, National Polar-orbiting Partnership (NPP) Visible Infrared Imaging Radiometer Suite (VIIRS) and Terra MODIS data. The results are compared with cloud coverage results interpreted via artificial vision. The results demonstrate that the algorithm achieves good cloud detection results for the above data, and the overall accuracy is greater than 90%.     

MODIS图象的云检测及分析

云一直是遥感图象处理、图象分析的一大障碍.为了解决这一问题,试图探讨利用中分辨率成像光谱仪MODIS检测云的方法,该方法充分考虑到MODIS数据具有36个光谱通道,特别是红外波段细分的特点,先是基于云的波谱特性采用多光谱综合法、红外差值法及指数法来对MODIS图象上的云点进行检测,鉴于这些方法有一定的局限性,因而还运用了一种基于空间结构分析和神经网络的云自动检测算法;最后将各种方法的云检测结果进行相互映证和对照分析,结果表明,这些方法检测到的云互相吻合,说明利用MODIS图象可成功地检测云点像元.这不仅为云的去除奠定了良好基础,而且也可以提高图象识别、图象分类及图象反演的精度.     

A dynamic global cloud layer for virtual globes

We describe a technique to merge multiple environmental satellite data sets for an hourly updated, near real-time global depiction of cloud cover for virtual globe applications. A global thermal infrared composite obtained from merged geostationary- (GEO) and low-Earth-orbiting (LEO) satellite data is processed to depict clear and cloudy areas in a visually intuitive fashion. This GEO-plus-LEO imagery merging is complicated by the fact that each individual satellite observes a single ‘snapshot’ of the cloud patterns, each taken at different times, whereas the underlying clouds themselves are constantly moving and evolving. For the cloudy areas, the brightness and transparency are approximated based upon the cloud top temperature relative to the local radiometric surface temperatures (corrected for surface emissivity variations) at the time of the satellite observation. The technique clearly defines and represents mid- to high-level clouds over both land and ocean. Due to their proximity to the Earth's surface, low-level clouds such as stratocumulus and stratus clouds will be poorly represented with the current technique, since warmer temperatures in this case do not correspond to higher cloud transparency. Overcoming this problem requires the introduction of multispectral channel combinations.     

Comparison of low brightness temperatures derived from the AVHRR thermal channels with in situ measurements in Antarctica

Data from the National Oceanic and Atmospheric Administration (NOAA) satellites' Advanced Very High Resolution Radiometers (AVHRRs) represent the longest record (more than 25 years) of continuously available satellite‐based thermal measurements, and have well‐chosen spatial and spectral resolutions. As a consequence, these data are used extensively to develop cloud climatologies. However, for such applications, accurate calibration and intercalibration of both solar and thermal channels of the AVHRRs is necessary so as to homogenize the data obtained from the different AVHRR sensors. AVHRR thermal channels 4 and 5 are routinely used in threshold‐based hierarchical decision‐tree cloud detection and classification algorithms, and therefore an evaluation of the stability of these channels at low temperatures is important. In this letter, the AVHRR channel 4 and 5 brightness temperatures (BTs) are compared at five stations in Antarctica. The data for the period of June, July and August (the coldest months of every year and with minimal atmospheric influence) from 1982 to 2006 were used for the evaluations. The calibration and intercalibration of the thermal channels are found to be very robust. The root mean square errors (RMSEs) range from 2.2 to 3.4 K and the correlation coefficients from 0.84 to 0.95. No apparent artefacts or artificial jumps in the BTs are visible in the data series after changes of sensors. The BTs from the thermal channels of the AVHRRs can be used for preparing cloud climatologies, as their intercalibration is found to be consistent across different afternoon satellites.     

Four years of ocean colour remote sensing with MOS-IRS

The imaging spectrometer MOS on IRS-P3 was launched in March 1996 as the first example of a new generation of ocean colour sensors. It consists of three different spectrometers in the visible/near-infrared spectral region with 18 channels. The IRS-P3 mission is focused on the remote sensing of case 2 water, particularly the derivation of different water constituents in coastal waters. Due to the more complex spectral behaviour of case 2 water, a new methodological approach was developed which works directly with satellite measured top-of-atmosphere radiance and accounts for the correlation of the different water constituents as well as for the spectral shape.

This paper gives an overview of the mission, the scientific goals and the development and improvement of the retrieval algorithms. The potential of the algorithm is demonstrated and examples of selected European coasts are shown. Derived maps of water constituents are presented.     

Size distributions of clouds in real time from satellite imagery

Abstract

An instrument is proposed to gather global data on cloud size distributions. The instrument contains a broad-band visible/infrared radiometer to scan the clouds with a field of view comparable to the grid squares of circulation models, but a resolution within the field of view of approximately 100 m, adequate to resolve the fine structure of clouds. The image is digitized, discriminated to distinguish cloudy and clear pixels, and submitted to a dedicated parallel computer which calculates the size distribution of the clouds in real time. Only the size distribution is transmitted to Earth, so the data transmission rate from the satellite is low and the archival problems at the earth station are minimal. This paper examines several alternative definitions of the size distribution of clouds, and concludes that the procedure of ‘sizing by openings’ is most appropriate. It is then shown that the processing speed required by the image analyser can be achieved at realistic power levels with present day technology.     

Haze removal for new generation optical sensors

Haze and cloud contamination is a common problem in optical remote-sensing imagery, as it can lead to the inaccurate estimation of physical properties of the surface derived from remote sensing and reduced accuracy of land cover classification and change detection. Haze optimized transform (HOT) is a methodology applicable to radiometric compensation of additive haze effects in visible bands that exhibits a spatially complex distribution over an image. The generic approach of HOT allows for the use of older satellite imaging sensors that include at least two visible bands (e.g. Landsat Thematic Mapper (TM) and Landsat Multispectral Scanner (MSS) sensors). This study proposes modifications to extend HOT applicability to new sensors. The improvements and extended functionality adapt the method to the higher radiometric resolution specifications of newer generation sensors and use percentile-based minimum in the correction procedure to avoid causing fake minimum. Alternative filters are also evaluated to smooth raw HOT output and the cloud mask is generated as an additional output. A Landsat 8 scene of Los Angeles is used to demonstrate the improved methodology. The methodology is applicable to sensors such as QuickBird, Worldview 2/3. More than 20 additional scenes were used to evaluate the effectiveness of the methodology.     

Impact of MODIS-derived cloud cover on aerosol optical depth measured with a sun photometer: a case study from Nanjing,China

Sun photometers have been used increasingly to monitor the atmospheric environment by measuring indicators such as aerosol optical depth (AOD). However, ground-measured AOD results are subject to the presence of clouds in the air. When cloud cover is not extensive, it is still possible to use sun photometry to determine AOD, even though accuracy is reduced by cloud contamination. This research aims to detect cloud cover from Moderate Resolution Imaging Spectroradiometer (MODIS) data and then assess its impact on in situ-measured AOD. Normalized difference cloud index (NDCI) and linear spectral unmixing (LSU) were used to detect cloud cover from MODIS data. AOD at the time of acquisition of MODIS data was measured on the ground by sun photometry within 20 min of satellite overpasses (10 min before and 10 min after). Correlation analysis of NDCI- and LSU-derived cloud cover with in situ-measured AOD data demonstrates that LSU has a higher correlation coefficient with AOD than with NDCI. At 550 nm, a unit of cloud cover (e.g. 1%) raises ground-observed AOD by 0.0157. The findings of this study can be used to modify ground-derived AOD results to improve their reliability.     

Technical note. Using dual thresholds for cloud and mosaicking

A method of cloud removal is presented utilizing data in the visible and near-infrared bands (0.5-1.1 μm) from a pair of HCMM (Heat Capacity Mapping Mission) satellite images of the South Island of New Zealand. The method uses two registered images, acquired at different times with different cloud distributions. The major improvement over other cloud detection techniques is the simultaneous use of a dual threshold to produce better cloud removal in the resulting mosaic picture from each original image. A previously developed mosaic method used by the EPIC image processing system at the Division of Information Technology, Department of Scientific and Industrial Research, New Zealand, to remove cloud is also used for comparison. The results show that both methods can be used to remove cloud but the appropriate method depends on the purpose of cloud removal.     

基于MODIS数据的积雪监测

通过对遥感卫星资料中云和雪的光谱特征的分析,提出利用中分辨率成像光谱仪(MODIS)红外、可见光谱段数据进行云、雪监测和分离的方法;并提供监测实例来说明利用MODIS数据可进行积雪监测。     

卫星信道三状态Markov模型设计与仿真

为研究大气空间环境对卫星通信的影响,将卫星通信的天气状况分为晴朗天气、云雾天气、雨雪天气3种情况,分析不同天气状况下的卫星信道传播特性并建立相应的Rice分布模型、C.Loo分布模型和Suzuki分布模型,引入三状态Markov过程实现因天气变化造成的信道状态的转换,构成卫星信道三状态Markov模型并对该模型的统计特性进行分析。最后,采用正弦波叠加法并利用实测气象卫星云图数据对该卫星信道三状态Markov模型进行仿真实验。仿真结果表明,所建模型可以用来模拟实际的卫星信道传播特性。