In CentralAfrica, where climatic variability is low [23,41,42] and annual rainfall is spatially homogeneous, any modification in dry season length and intensity could have dramatic consequences on vegetationstructure and greenness . Some slight climatic differences might be more important than previously thought. In the study area, differences in veg- etation phenology and traits have been evidenced and these can be linked not only to differences in soil properties, but also to slight differences in climate variables which may have been overlooked. The distribution of semi-deciduous and disturbed vegetation is in line with the Sangha River Inter- val, a region that has probably experienced more impact of past climate changes than elsewhere in the study area. In fact, the Sangha sub-river basin of the Congo River has regu- larly received less precipitation between 1950 and 1980 than the Oubangui and Central Congo sub-basin bordering it . Although new and more detailed data are necessary to confirm this, we can tentatively conclude the relative fragility of this area in the face of climate changes, notably precipitation decrease and/or dry season increase.
Remote Sens. 2019, 11, 1656 2 of 16
observation data, where the NDVI is widely used as an indicator of ecosystem greenness [ 22 , 23 ] due to its linking of photosynthesis with carbon uptake in main terrestrial ecosystems [ 24 – 28 ]. A range of satellite products (MODIS/Terra or MODIS/Aqua, https://modis.gsfc.nasa.gov/ ; SPOT/VGT, http://www.spot-vegetation.com/ ; and NOAA/AVHRR, https://www.ncdc.noaa.gov/ ) have been studied to increase understanding of changes in ecosystem greenness at the global scale; however, impacts of landscape heterogeneity and spatial scale on the relationship between greenness and ecosystem productivity are unclear [ 29 , 30 ] because research tends to be based on local scale, single-site studies. Field-based measurements of the NDVI are time consuming, particularly in highly heterogeneous landscapes, and field sampling protocols tend to be based on the Validation of European Remote Sensing Instruments (VALERI [ 31 ], http://w3.avignon.inra.fr/valeri/ ) protocol for the estimation of NDVI, leaf area index (LAI), or vegetation cover. The VALERI method was developed for various spatial scales, from small study plots to the size of a pixel, of coarse-resolution satellite data. The VALERI database is available on request for a large number of sites worldwide at the http://w3.avignon.inra.fr/valeri/fic_htm/database/main.phpwebpage . Geostatistical techniques and pixel-based classification algorithms offer a valuable alternative for an accurate and precise classification of landscape [ 32 , 33 ]. GPP and NDVI field measurements are scaled up using in situ FLUXNET [ 34 ] radiation data [ 35 , 36 ], and subsequent regression analysis is often used to define transfer functions that allow the estimation of GPP from NDVI data. However, the uncertainties related to the scale mismatch between the in situ point measurements, flux footprint, and the pixel resolution remain a critical issue.
Remote Sens. 2019, 11, x FOR PEER REVIEW 4 of 18
Figure 2. Flowchart indicating the steps followed to perform the analyses presented in this study.
Acronyms: Gross primary production (GPP), remotely sensed vegetation indices (RSVI).
We downloaded all images for Europe for 2000–2016 from a MODIS (Moderate Resolution Imaging Spectroradiometer) composite processed by the MAIAC (Multi-Angle Implementation of Atmospheric Correction) [25–27] data set with a resolution of 1 km. The coarse resolution of the MAIAC dataset is not a problem when monitoring data coming from eddy covariance towers because land-use area around the towers is normally homogeneous . We used raw datafrom the MAIAC data set, because this data set has less atmospheric noise and detects clouds better than previous data sets [25–27]. We therefore did not further filter our data. We calculated NDVI, EVI , NIR v as defined by  and , and CCI  using the central pixel from all records from the flux towers (Aqua + Terra images) by combining reflectance bands as:
1.1. Remote sensing and large-area LULC mapping
Over the past decade, remotely sensed datafrom satellite- based sensors have proven useful for large-area LULC char- acterization due to their synoptic and repeat coverage. Con- siderable progress has been made classifying LULC patterns at the state (Eve & Merchant, 1998) and national (Craig, 2001; Homer et al., 2004; Vogelmann et al., 2001) levels using multi- spectral, medium resolution datafrom the Landsat Thematic Mapper (TM) and Enhanced Thematic Mapper (ETM+) as a primary input. Similar advances in LULC classification have also been made at national (Loveland et al., 1991; Lu et al., 2003) to global (DeFries et al., 1998; DeFries and Townshend, 1994; Hansen et al., 2000; Loveland and Belward, 1997; Love- land et al., 2000) scales using multi-temporal, coarse resolu- tion data (1 and 8 km) from the Advanced Very High Resolu- tion Radiometer (AVHRR). However, few of these mapping efforts have classified detailed, crop-related LULC patterns (Craig, 2001), particularly at the annual time step required to reflect common agricultural LULC changes. The develop- ment of a regional-scale crop mapping and monitoring pro- tocol is challenging because it requires remotely sensed data that have wide geographic coverage, high temporal reso- lution, adequate spatial resolution relative to the grain of the landscape (i.e., typical field size), and minimal cost. Re- motely sensed datafrom traditional sources such as Landsat and AVHRR have some of these characteristics, but are lim-
This study represents substantial progress toward deve- loping a system for prediction of fire occurrence risk based on temporal trends in fire density and drought. Temporal trends were measured by a satellite fuel dry- ness index, dead ratio (DR), in different types of vegeta- tion and regions in Mexico, at a national scale, with a monthly temporal resolution, for the period 2003 to 2014. DR trends varied by vegetation type and region, with drier fuel conditions measured in the most arid type of fuels and regions. Furthermore, significant rela- tionships were found relating monthly fire density and DR for the analyzed vegetation types and regions in the period of study. In addition, we obtained preliminary seasonal autoregressive integrated moving average models for prediction of monthly DR values that might be incorporated into future fire risk forecast operational tools. While these initial results suggest that there is potential for the indices utilized to capture the variations in temporal trends of fuel dryness and their impact on fire occurrence in the country, a longer time period of monitoring will be required for improving our under- standing of long-term climatic effects, such as El Niño and La Niña impact, on drought and associated fire risk in the country.
The coastal region of Sabah, Malaysia is mostly affected by harmful algal blooms (HABs) that often cause massive fish kills, and sometimes human casualties. Lack of a well-agreed, transparent and reproducible method, aperiodic nature and limited (technical) ability to monitor HAB at large regional scale, have all led to reinforced methodological improvement for bloom prediction, scientific management of coastal water resources, and regulatory attention. MODerate Resolution Imaging Spectroradiometer (MODIS), one of the well validated ocean colour sensors, enables acquiring high spectral resolution images, with short revisit time, freely accessible, and bay-wide coverage. Yet, the relative efficiency of MODIS-derived Chl-a (Chlorophyll-a), ABI (Algal Bloom Index), and nFLH (normalized Fluorescence Line Height) have not been compared in coastal regions of Malaysia. Fifteen MODIS Level 2 images acquired between 2005 and 2013 were considered as time series data that matched HAB events mentioned in previous studies. As historical real time in-situ data collection is often difficult (inaccessible), and thus unavailable; this project had to validate results obtained from literature, assuming that in-situ, would indicate HAB location at least during MODIS acquisition dates. Variations of HAB affected areas with temporal and spatial scales derived from bloom indices are shown in colour maps. Reliability of bloom information was measured by subjectively comparing HAB results provided by indices, and previously published in-situ results. ABI outperformed Chl-a and nFLH indices based on comparisons in both normal and HAB conditions occurring in the coastal waters of Sabah and Sarawak. The configuration and reliability retrieved fromMODIS-ABI allowed their application in different likely tropical region as automated HAB monitoring systems and coastal water management programmes.
Ground-based measured climate data are either scarce or low quality in Ethiopia, particularly in the DRB (Conway 2000; Tena et al. 2016). As a result of this, we used qual- ity-controlled satellite-derived datasets for this study. The 4-km gridded monthly minimum and maximum tem- perature dataset, which was constructed by the Enhanced National Climate Time-series Service (ENACTS) ini- tiative was obtained from the National Meteorological Agency (NMA) of Ethiopia for the 1983 to 2015 period. The ENACTS derived temperature dataset was produced by merging the Moderate Resolution Imaging Spectro- radiometer (MODIS) satellite land surface temperature (LST) with over 300 temperature stations data over Ethi- opia (Tufa et al. 2018). Similarly, Climate Hazards Group InfraRed Precipitations with Stations (CHIRPS) version 2 derived rainfall product (http://dx.doi.org/10.15780 / G2RP4 Q) was used for the rainfall trend analysis.
Most of the above works, however, only consider inte- grating multiple sources of data in parallel fashion, ig- noring hierarchical, or vertical structure of multi-omics data. Furthermore, only few machine learning algo- rithms, including SSL, deals with networks of vertical structure. The purpose of the paper is to develop a semi-supervised learning algorithm for multiple layered networks that utilize matrix separation and graph inte- gration method in vertical fashion. For biological sys- tems, however, vast number of components in each layers and countless unknown relations between differ- ent layers cause issues of computational complexity and sparseness for analyzing with multi-layered networks. To alleviate the problems, we propose an efficient matrix inversion algorithm composed with Nyström method  and Woodbury formula . The remainder of the paper is organized as the following. In Methods, we discuss graph based semi-supervised learning for multiple-layered networks. In Experiments and Results and Discussion, we present experimental results of the proposed algorithm that was applied to disease co-occurrence prediction problem on two layered net- work of symptom and disease.
Volcanic activity can be of many different types, but is mainly split into two categories: effusive and explosive (Francis and Oppenheimer, 2004). The main difference between the two styles arises from the presence or absence of magma fragmentation. The important parameter driving fragmentation is the gas content. Most explosive events occur when the dissolved gas exsolves as it rises in a viscous magma that does not let gas escape, building pressure and resulting in the fragmentation of magma into small particles (tephra) in a sudden manner. By contrast, in volcanoes with less viscous magma, gas can escape easily and lava can be extruded or flow passively in lava flows or domes, producing effusive eruptions with no fragmentation. However, the range of possible activity is actually continuous between these two eruptive styles with no clear separation (Francis and Oppenheimer, 2004). Here, we define effusive activity following Parfitt and Wilson (2008) “an effusive eruption is an eruption in which lava flows away from a vent as a coherent liquid”. Lava flows are often basaltic, i.e., low silica content, with low viscosity and high temperature flowing in a sustained manner for hours to days and up to months. We focused on basaltic activity because it is the most abundant, i.e., more than half of volcanoes are formed with basaltic magma composition (Walker, 2000), basaltic flows can travel for greater lengths than more silica-rich magmas and are found in all tectonic settings. A large majority of effusive activity occurs underwater, unseen by most at mid-oceanic ridges (three quarters of the Earth’s magma output is emitted at divergent plate boundaries; Simkin and Siebert, 1994), and only visible at the surface in Iceland and East Africa. Therefore, effusive eruptions occurring at the surface give us the opportunity to observe and study lava flow emplacement to understand their dynamics.
The NDVI is the most widely used of the vegetation indices for classification applications (Lloyd 1990, Myneni et al. 1995, Li and Moon 2004). For example, using NDVI data, Running et al. (1995) present a decision tree classification based on the permanence of above-ground biomass, longevity of leaves, and leaf type. Critically, the use of a vegetation index can yield a classification that is more accurate than one derived from the data used in its calculation (Anderson et al. 1993, Nemani et al. 1993, Hirata et al. 2001). This feature, together with the ready availability of NDVI data in major archives (e.g. Smith et al. 1997), has led to the NDVI being used widely as a discriminating variable in image classification. Indeed, the NDVI derived from a variety of different sensors has been used to classify vegetated terrain at scales ranging from the local to global (Benedetti et al. 1994, Lobo et al. 1997, Hansen et al. 2000, Han et al. 2004). The NDVI is, for example, at the core of major global land cover mapping programmes (e.g. Loveland et al. 2000). Typically, the data used are a time series of NDVI images that provide a measure of phenological variability in space and time, which can facilitate inter-class discrimination (Tucker et al. 1985, DeFries and Townsend 1994). However, the use < of the NDVI to classify vegetation has some major limitations, such as a relative
were recently developed as a result of signature analysis of the reflectance spectra of two deciduous species (maple and chestnut) (Gitelson and Merzlyak, 1994a, 1994b; Gitelson et al., 1996) and of tobacco plant (Lichtenthaler et al., 1996). These indices allow for chlorophyll estimation in dark- green to yellow leaves within a wide range of pigment variation. It was reported also that the index R760/R695 is a sensitive indicator of plant stress. Vegetation water content (VWC) is an important indicator in agricultural and forestry applications. The VWC could possibly provide information for agriculture that can be used to infer water stress for irrigation decisions, aid in yield estimation and in the assessment of drought conditions. At a whole-plant level, soil drought and leaf water deficit lead to a progressive suppression of photosynthtesis, and is associated with alterations in carbon and nitrogen assimilation (Zlatev and Lidon, 2012). The principle application to forestry is determining fire susceptibility (Pyne et al., 1996). The VWC is also used in retrieving soil moisture from microwave remote sensing observations. Ceccato et al. (2002 a, b) found that the short wave infra red (SWIR) channel was critical to estimating VWC and the near Infra red (NIR) channel was needed to account for
Abstract The advancement in satellite technology in terms of spatial, temporal, spectral and radio- metric resolutions leads, successfully, to more speciﬁc and intensiﬁed research on agriculture. Auto- matic assessment of spatio-temporal cropping pattern and extent at multi-scale (community level, regional level and global level) has been a challenge to researchers. This study aims to develop a semi-automated approach using Indian Remote Sensing (IRS) satellite data and associated vegeta- tion indices to extract annual cropping pattern in Muzaffarpur district of Bihar, India at a ﬁne scale (1:50,000). Three vegetation indices (VIs) – NDVI, EVI2 and NDSBVI, were calculated using three seasonal (Kharif, Rabi and Zaid) IRS Resourcesat 2 LISS-III images. Threshold reference values for vegetation and non-vegetation thematic classes were extracted based on 40 training samples over each of the seasonal VI. Using these estimated value range a decision tree was established to classify three seasonal VI stack images which reveals seven different cropping patterns and plantation. In addition, a digitised reference map was also generated frommulti-seasonal LISS-III images to check the accuracy of the semi-automatically extracted VI based classiﬁed image. The overall accuracies of 86.08%, 83.1% and 83.3% were achieved between reference map and NDVI, EVI2 and NDSBVI, respectively. Plantation was successfully identiﬁed in all cases with 96% (NDVI), 95% (EVI2) and 91% (NDSBVI) accuracy.
Abstract. High-latitude treeless ecosystems represent spa- tially highly heterogeneous landscapes with small net car- bon fluxes and a short growing season. Reliable observa- tions and process understanding are critical for projections of the carbon balance of the climate-sensitive tundra. Space- borne remote sensing is the only tool to obtain spatially con- tinuous and temporally resolved information on vegetationgreenness and activity in remote circumpolar areas. How- ever, confounding effects from persistent clouds, low sun elevation angles, numerous lakes, widespread surface inun- dation, and the sparseness of the vegetation render it highly challenging. Here, we conduct an extensive analysis of the timing of peak vegetation productivity as shown by satel- lite observations of complementary indicators of plant green- ness and photosynthesis. We choose to focus on productivity during the peak of the growing season, as it importantly af- fects the total annual carbon uptake. The suite of indicators are as follows: (1) MODIS-based vegetation indices (VIs) as proxies for the fraction of incident photosynthetically active radiation (PAR) that is absorbed (fPAR), (2) VIs combined with estimates of PAR as a proxy of the total absorbed radia- tion (APAR), (3) sun-induced chlorophyll fluorescence (SIF) serving as a proxy for photosynthesis, (4) vegetation optical depth (VOD), indicative of total water content and (5) empir- ically upscaled modelled gross primary productivity (GPP). Averaged over the pan-Arctic we find a clear order of the annual peak as APAR 5 GPP < SIF < VIs/VOD. SIF as an
Pattern recognition is an important artificial intelligence task with practical applications in many fields such as medical and species distribution. Such application involves overlapping data points which are demonstrated in the multi- label dataset. Hence, there is a need for a recognition algorithm that can separate the overlapping data points in order to recognize the correct pattern. Existing recognition methods suffer from sensitivity to noise and overlapping points as they could not recognize a pattern when there is a shift in the position of the data points. Furthermore, the methods do not implicate temporal information in the process of recognition, which leads to low quality of data clustering. In this study, an improved pattern recognition method based on Hierarchical Temporal Memory (HTM) is proposed to solve the overlapping in data points of multi- label dataset. The imHTM (Improved HTM) method includes improvement in two of its components; feature extraction and data clustering. The first improvement is realized as T S-Layer Neocognitron algorithm which solves the
As traditional fieldwork is time-consuming and costly, re- mote sensing methods have been utilized as cost-effective ap- proaches to detect vegetation changes at large spatial scales. Monitoring landscapes through satellite-derived vegetation indices (VIs), such as the normalized difference vegetation index (NDVI) and enhanced vegetation index (EVI), is a suc- cessful method for assessing vegetation conditions and phe- nology (Glenn et al., 2008; Zucca et al., 2015). Previous stud- ies suggested that time series satellite data can reliably detect the phenology, forage quantity, and quality of grassland areas using the VIs derived from Advanced Very High Resolution Radiometer (AVHRR) data and Moderate Resolution Imag- ing Spectroradiometer (MODIS) satellite images (Kawamura et al., 2003, 2005a, b, c). Significant delays in the vegeta- tion green-up during 1982–1991 in the desert steppe and in part of the typical steppe of Inner Mongolia were detected using AVHRR NDVI (Yu et al., 2003). Moreover, although the signs of the trends in the vegetation green-up dates de- tected by various methods were broadly consistent spatially and for different vegetation types, large differences occurred in the magnitudes of the observed trends. The large variance obtained using different methods is particularly apparent for arid and semiarid vegetation types (Cong et al., 2012, 2013; Zhao et al., 2012).
The network is defined by an adjacency list consisting of pairs of numbers: the node identifier of the starting node of the edge and the node identifier of the ending node of the link. Thus the network will incorporate in a simple adjacency list the temporal sequencing of the contacts and the interaction between the properties of the type of dis- ease being modelled and the types of locations on which the cattle are held. The construction of the network, as described above, ensures that all the links are real routes through which infection can pass from one location via an intermediary location to a third location. This is in con- trast to networks where the nodes are locations and the edges are movements. These lose the temporal sequence information available in the source movement data. For the two networks (with 7-day infectious period, denoted by "the 7-day infection network", and with 14- day infectious period, denoted by "the 14-day infection network", as described above), the following standard network parameters  have been calculated within Ora- cle, by processing Oracle output in MS Excel, and with our own routines (the Contagion library ):
A partir dos dados de campo, foram extraídas assinaturas temporais representativas das formações florestais, savânicas e campestres. Para assegurar que a extração dessas assinaturas seria realizada com base em pixels puros do MODIS, isto é, com nível mínimo de mistura espectral, as coordenadas geográficas de campo foram sobrepostas nas imagens ortorretificadas do Landsat-8 e também em uma grade de células de 250 metros x 250 metros, coincidente com os pixels do MODIS. O software utilizado foi o ArcGIS 10.1. Procurou-se selecionar ao menos quatro pixels para cada ponto amostral, os quais foram convertidos para um único valor por meio de média aritmética simples. Entretanto, em alguns casos, em função do tamanho da área visitada em campo, somente foi possível a seleção de um pixel puro. A extração de séries temporais de NDVI e EVI dos pixels selecionados foi feita por meio do software ENVI 4.8. As séries temporais de EVI e de NDVI de cada um dos 66 pontos de campo correspondentes às três formações vegetais naturais do Cerrado foram então convertidas em séries representativas por formação e por município. Assim, foram obtidas séries temporais de formação florestal de Jataí e SMA, formação savânica de todos os quatro municípios e formação campestre de Mateiros. Essas séries temporais, por apresentarem ruídos causados, por exemplo, pela presença de nuvens durante a passagem do satélite, foram suavizadas pelo filtro denominado logística dupla, disponível no programa TIMESAT (Jönsson e Eklundh, 2002, 2004). Esse software, de domínio público, disponibiliza três métodos de suavização: logística dupla, Savitzky-Golay e gaussiano assimétrico. Testes intensivos realizados por Borges et al. (2014) indicaram melhor desempenho da logística dupla para dados do MODIS EVI do Cerrado, mais especificamente, do oeste da Bahia. A logística dupla corresponde a uma função harmônica e polinomial. Sua formulação matemática é dada pela Equação 3, onde o parâmetro x1 determina a posição do ponto de inflexão esquerdo, enquanto x2 determina a taxa de variação. O parâmetro x3 determina a posição do ponto de inflexão do lado direito, enquanto x4 fornece a taxa de variação nesse ponto. De acordo com Jönsson e Eklundh (2004), esse filtro garante uma forma suave às séries temporais.
Data centers must consider both computational and network availability needs and environmental goals. As these apparently conflicting goals are considered, decision makers will need to think critically about what strategies to implement in addition to their regular thinking about how best to implement them. For example, data center cooling system design engineers and facilities managers may already be accustomed to searching for chiller equipment that uses the fewest possible kW per ton to produce chilled water. Clearly, doing this will improve cooling efficiency. However, these facilities experts may not think to ask why the chilled water needs to be at the 42°F (6°C) commonly specified for an office building instead of, say, 49°F (9°C). From a whole-systems perspective, standard efficiency chillers at 49°F (9°C) would deliver cooling more efficiently than premium efficiency chillers at a lower set point, because latent dehumidification in the data center is eliminated. Similar initiatives can be employed to reduce the volume of outside air