Forecast of NDVI in coniferous areas using temporal ARIMA analysis and climatic data at a regional scale

Important issues such as the prediction of drought, fire risk and forest disease are based on analysis of forest vegetation response. A method of forecasting the short-term response of forest vegetation on the basis of an autoregressive integrated moving average (ARIMA) analysis was designed in this study. We used 10-day maximum value composite (MVC) bands of the Normalized Difference Vegetation Index (NDVI) obtained from National Oceanic and Atmospheric Administration (NOAA) Advanced Very High Resolution Radiometer (AVHRR) data from 1993 to 1997. Using the theory of stochastic processes (Box–Jenkins), the MVC-NDVI series was analysed and a seasonal ARIMA (SARIMA) model was developed for forecasting NDVI in the following 10-day periods. The SARIMA model identified a moving-average regular term with a 10-day lag and an autoregressive 37 10-day period seasonal term with a one-season (1-year) component. The study also demonstrated a slight relationship between the NDVI and the precipitation level in some species of conifers by using climatic time series and the analysis of dynamic models and allowed us to elaborate an image of the immediate future NDVI for the study area (Castile and Leon, Spain).     

Determination of seasonal and interannual variation in New Zealand pasture growth from NOAA-7 data

The normalized vegetation index (NVI) has been calculated from afternoon overpasses of NOAA-7 for two important farming regions of New Zealand, approximately 1000 k² in area, for the period from October 1981 through June 1984. The uniform nature of the terrain and farming practices in these areas make them ideal targets for remote sensing from satellites with limited spatial resolution. The frequency of useful data coverage has been increased by sampling within cloud-free parts of a partly cloudy target area and also by deriving an empirical correction for off-nadir view angles. Daily area-mean rainfall and soil moisture were estimated for both regions and monthly area-mean pasture growth for one of them. The NVI was found to reflect the varying rainfall and soil moisture on time scales of one week or more during the growing season and between years. A correlation of 0.81 was found between NVI and pasture growth on a monthly mean basis. These results suggest that operational satellite monitoring of these and other areas would provide valuable assistance in agricultural management and forward planning.     

Combining seasonal ARIMA models with computational intelligence techniques for time series forecasting

Seasonal autoregressive integrated moving average (SARIMA) models form one of the most popular and widely used seasonal time series models over the past three decades. However, in several researches it has been argued that they have two basic limitations that detract from their popularity for seasonal time series forecasting tasks. SARIMA models assume that future values of a time series have a linear relationship with current and past values as well as with white noise; therefore, approximations by SARIMA models may not be adequate for complex nonlinear problems. In addition, SARIMA models require a large amount of historical data to produce desired results. However, in real situations, due to uncertainty resulting from the integral environment and rapid development of new technology, future situations must be forecasted using small data sets over a short span of time. Using hybrid models or combining several models has become a common practice to overcome the limitations of single models and improve forecasting accuracy. In this paper, a new hybrid model, which combines the seasonal autoregressive integrated moving average (SARIMA) and computational intelligence techniques such as artificial neural networks and fuzzy models for seasonal time series forecasting is proposed. In the proposed model, these two techniques are applied to simultaneously overcome the linear and data limitations of SARIMA models and yield more accurate results. Empirical results of forecasting two well-known seasonal time series data sets indicate that the proposed model exhibits effectively improved forecasting accuracy, so that it can be used as an appropriate seasonal time series model.     

The collection 5 MODIS burned area product — Global evaluation by comparison with the MODIS active fire product

The results of the first consecutive 12 months of the NASA Moderate Resolution Imaging Spectroradiometer (MODIS) global burned area product are presented. Total annual and monthly area burned statistics and missing data statistics are reported at global and continental scale and with respect to different land cover classes. Globally the total area burned labeled by the MODIS burned area product is 3.66 × 10⁶ km² for July 2001 to June 2002 while the MODIS active fire product detected for the same period a total of 2.78 × 10⁶ km², i.e., 24% less than the area labeled by the burned area product. A spatio-temporal correlation analysis of the two MODIS fire products stratified globally for pre-fire leaf area index (LAI) and percent tree cover ranges indicate that for low percent tree cover and LAI, the MODIS burned area product defines a greater proportion of the landscape as burned than the active fire product; and with increasing tree cover (> 60%) and LAI (> 5) the MODIS active fire product defines a relatively greater proportion. This pattern is generally observed in product comparisons stratified with respect to land cover. Globally, the burned area product reports a smaller amount of area burned than the active fire product in croplands and evergreen forest and deciduous needleleaf forest classes, comparable areas for mixed and deciduous broadleaf forest classes, and a greater amount of area burned for the non-forest classes. The reasons for these product differences are discussed in terms of environmental spatio-temporal fire characteristics and remote sensing factors, and highlight the planning needs for MODIS burned area product validation.     

模糊神经网络与SARIMA结合的时间序列预测模型

模糊神经网络和SARIMA模型分别对非线性和线性时间序列有很好的预测能力,但在实际应用中大多数序列并非稳定、单纯线性或非线性的。为了提高预测精度,提出了一种基于T-S模糊神经网络与SARIMA结合的时间序列预测模型。针对悉尼航班乘客收入数据给出了三种混合模型,并与模糊神经网络、支持向量机、SARIMA和BP神经网络四种单独模型进行比较。实验结果表明,从预测精度和参数选择方面来看,所给模型是有效的。     

A hybrid SARIMA and support vector machines in forecasting the production values of the machinery industry in Taiwan

This paper proposes a hybrid methodology that exploits the unique strength of the seasonal autoregressive integrated moving average (SARIMA) model and the support vector machines (SVM) model in forecasting seasonal time series. The seasonal time series data of Taiwan’s machinery industry production values were used to examine the forecasting accuracy of the proposed hybrid model. The forecasting performance was compared among three models, i.e., the hybrid model, SARIMA models and the SVM models, respectively. Among these methods, the normalized mean square error (NMSE) and the mean absolute percentage error (MAPE) of the hybrid model were the lowest. The hybrid model was also able to forecast certain significant turning points of the test time series.     

Autoregressive forecast of monthly total ozone concentration: A neurocomputing approach

The present study endeavors to generate autoregressive neural network (AR-NN) models to forecast the monthly total ozone concentration over Kolkata (22°34′, 88°22′), India. The issues associated with the applicability of neural network to geophysical processes are discussed. The autocorrelation structure of the monthly total ozone time series is investigated, and stationarity of the time series is established through the periodogram. From various autoregressive moving average (ARMA) and autoregressive models fit to the time series, the autoregressive model of order 10 is identified as the best. Subsequently, 10 autoregressive neural network (AR-NN) models are generated; the 10th order autoregressive neural network model with extensive input variable selection performs the best among all the competitive models in forecasting the monthly total ozone concentration over the study zone.     

A study of the seasonal dynamics of grassland growth rates in Inner Mongolia based on AVHRR data and a light‐use efficiency model

Growth rate data for different pastures could provide important reference data for developing rotation grazing plans, for hay production, and for forage replenishment. Based on AVHRR NDVI data and a light‐use efficiency (LUE) model, we estimated absorbed photosynthetically active radiation and LUE (ε) by integrating air and soil temperature, precipitation and total solar radiation time series data from 1986 to 1999, and calculated the absolute growth rate (AGR) and cumulative absolute growth rate (CAGR) of aboveground biomass in each growing season in China's Inner Mongolia region. AGR and CAGR estimated by the LUE model were validated using monthly growth data obtained for the vegetation in desert steppe, typical steppe, and meadow steppe ecosystems from 1986 to 1995. The LUE model provided sufficiently good simulation accuracy that its use should permit improved livestock feed management in the study area. From 1986 to 1999, average CAGR of steppe vegetation during the growing season increased quickly in June and July, reached a maximum in July and August, and declined in September. In 1999, AGR reflected the pattern of seasonal vegetation dynamics during the growing season.     

A MODIS direct broadcast algorithm for mapping wildfire burned area in the western United States

Improved wildland fire emission inventory methods are needed to support air quality forecasting and guide the development of air shed management strategies. Air quality forecasting requires dynamic fire emission estimates that are generated in a timely manner to support real-time operations. In the regulatory and planning realm, emission inventories are essential for quantitatively assessing the contribution of wildfire to air pollution. The development of wildland fire emission inventories depends on burned area as a critical input. This study presents a Moderate Resolution Imaging Spectroradiometer (MODIS) - direct broadcast (DB) burned area mapping algorithm designed to support air quality forecasting and emission inventory development. The algorithm combines active fire locations and single satellite scene burn scar detections to provide a rapid yet robust mapping of burned area. Using the U.S. Forest Service Fire Sciences Laboratory (FiSL) MODIS-DB receiving station in Missoula, Montana, the algorithm provided daily measurements of burned area for wildfire events in the western U.S. in 2006 and 2007. We evaluated the algorithm's fire detection rate and burned area mapping using fire perimeter data and burn scar information derived from high resolution satellite imagery. The FiSL MODIS-DB system detected 87% of all reference fires > 4 km², and 93% of all reference fires > 10 km². The burned area was highly correlated (R² = 0.93) with a high resolution imagery reference burn scar dataset, but exhibited a large over estimation of burned area (56%). The reference burn scar dataset was used to calibrate the algorithm response and quantify the uncertainty in the burned area measurement at the fire incident level. An objective, empirical error based approach was employed to quantify the uncertainty of our burned area measurement and provide a metric that is meaningful in context of remotely sensed burned area and emission inventories. The algorithm uncertainty is ± 36% for fires 50 km² in size, improving to ± 31% at a fire size of 100 km². Fires in this size range account for a substantial portion of burned area in the western U.S. (77% of burned area is due to fires > 50 km², and 66% results from fires > 100 km²). The dominance of these large wildfires in burned area, duration, and emissions makes these events a significant concern of air quality forecasters and regulators. With daily coverage at 1-km² spatial resolution, and a quantified measurement uncertainty, the burned area mapping algorithm presented in this paper is well suited for the development of wildfire emission inventories. Furthermore, the algorithm's DB implementation enables time sensitive burned area mapping to support operational air quality forecasting.     

Generating vegetation leaf area index Earth system data record from multiple sensors. Part 2: Implementation,analysis and validation

The evaluation of a new global monthly leaf area index (LAI) data set for the period July 1981 to December 2006 derived from AVHRR Normalized Difference Vegetation Index (NDVI) data is described. The physically based algorithm is detailed in the first of the two part series. Here, the implementation, production and evaluation of the data set are described. The data set is evaluated both by direct comparisons to ground data and indirectly through inter-comparisons with similar data sets. This indirect validation showed satisfactory agreement with existing LAI products, importantly MODIS, at a range of spatial scales, and significant correlations with key climate variables in areas where temperature and precipitation limit plant growth. The data set successfully reproduced well-documented spatio-temporal trends and inter-annual variations in vegetation activity in the northern latitudes and semi-arid tropics. Comparison with plot scale field measurements over homogeneous vegetation patches indicated a 7% underestimation when all major vegetation types are taken into account. The error in mean values obtained from distributions of AVHRR LAI and high-resolution field LAI maps for different biomes is within 0.5 LAI for six out of the ten selected sites. These validation exercises though limited by the amount of field data, and thus less than comprehensive, indicated satisfactory agreement between the LAI product and field measurements. Overall, the inter-comparison with short-term LAI data sets, evaluation of long term trends with known variations in climate variables, and validation with field measurements together build confidence in the utility of this new 26 year LAI record for long term vegetation monitoring and modeling studies.     

1999年锡林郭勒草地AVHRR-NDVI时空变化研究

采用地面样地和遥感数据,分析了锡林郭勒草原4种草地1999年AVHRR-NDVI的时空变化。NDVI时间序列显示,草甸草地和典型草地NDVI在5月份开始返青,但荒漠草地和沙地草地NDVI在5月份为全年最低。6月份草地生长最快,以4月平均NDVI作为各草地的基准,草甸草地全年NDVI最大时的增加值为0.42,其中4～6月份的NDVI增加值为0.34,占81%;典型草地全年NDVI最大时的增加值为0.30,其中4～6月增加值为0.22,占73%;沙地草地全年NDVI最大时的增加值0.27,其中4～6月增加值为0.10,占37%。NDVI空间分布呈现明显的东西过渡特点。从5月开始,NDVI自东向西增长,其推进特点是南北两侧快,中部缓慢。8月草地植被达到全盛时,NDVI=0.1的等值线呈现由西向东的“楔型”。5～8月之间,NDVI=0.1等值线由东向西移动3个经度。由于植被指数对低覆盖植被比较敏感,该等值线的形态和位置可能是气候变化的一个指标。     

Calibration of NOAA-7 AVHRR,GOES-5, and GOES-6 VISSR/VAS solar channels

The NOAA-7, GOES-5, and GOES-6 VISSR/VAS solar channels have been calibrated for the periods from October 1983 through January 1985 (NOAA-7, GOES-6) and from October 1983 through July 1984 (GOES-5). Space and the White Sands National Monument area in Mexico, whose reflectance properties are well known, are used as calibration targets. The shortwave reflected terrestrial radiance that is measured at satellite altitude is computed using a fairly accurate radiative transfer model which accounts for multiple scattering and bidirectional effects (Tanré et al., 1979). The relevant atmospheric characteristics are estimated from climatological data (ozone amount, aerosol size-frequency distribution, and refractive index) and observations at the nearest meteorological sites (water vapor amount, visibility). The approach produces accuracies of 8–13% depending on the channel considered. For both types of instruments, no drift in the solar channels in detected during the 15-month period. The gain changes, about 15% of the mean values, are largely attributed to inhomogeneities of the ground target (shading effects due to the presence of dunes). No systematic effect of the normalization procedure applied by NOAA to the raw VISSR/VAS data is detected. There is some evidence that the GOES-5 solar channels gradually deteriorated from March 1984 until the satellite failure in July 1984. Comparisons between gains determined in orbit and those before launch show that the NOAA-7 solar channels read higher by about 15%. The disparities, however, cannot be explained by model errors and must have occurred before the time period analyzed here.     

Satellite remote sensing of total dry matter production in the Senegalese Sahel

Nine predominately cloud-free NOAA-7 advanced very high resolution radiometer images were obtained during a 3-month period during the 1981 rainy season in the Sahel of Senegal. The 0.55–0.68- and 0.725–1.10-μm channels were used to form the normalized difference green leaf density vegetation index and the 11.5–12.5-μm channel was used as a cloud mask for each of the nine images. Changes in the normalized difference values among the various dates were closely associated with precipitation events. Six of the images spanning an 8-week period were used to generate a cumulative integrated index. Ground biomass samplings in the 30,000 km² study area were used to assign total dry biomass classes to the cumulative index.     

Satellite remote sensing of total herbaceous biomass production in the senegalese sahel: 1980–1984

NOAA-6 and NOAA-7 1-km and 4-km advanced very high resolution radiometer data were obtained at frequent intervals during the 1980, 1981, 1982, 1983, and 1984 rainy or growing seasons in the Sahel zone of northern Senegal. Above-ground herbaceous biomass clippings, visual estimates, and hand-held radiometer measurements of herbaceous vegetation were made during and at the conclusion of the rainy seasons for 4 of the 5 years. The satellite data were compared to sampled above-ground biomass data and the integral of the satellite data over time was compared to end-of-growing-season above-ground total dry biomass. A strong correlation between the integrated NOAA-7 satellite data and end-of-season above-ground dry biomass was found for ground samples collected over a 3-year period. The satellite data documented the highly variable precipitation regime in the Senegalese Sahel both within years and among years and suggest a direct method of monitoring Sahelian total herbaceous biomass production in areas where the percentage cover of woody species is less than 10%. Predicted average total dry biomass production was 1093 kg/ha for 1981, 536 kg/ha for 1982, 178 kg/ha in 1983, and 55 kg/ha in 1984 for the $\text{～ 30,000 km}{}^2$ study area.     

基于长短期神经网络的遥测参数趋势预测

时间序列分析的主要目的是根据已有的历史数据对未来进行预测。传统的时间序列预测主要依靠基于模型的方法，比如季节性差分整合移动平均自回归模型（SARIMA）和指数平滑法（EXP）等。此类方法的参数选择严重依赖于专家经验，适用性并不广泛。针对周期性遥测参数，采用长短期记忆网络（LSTM），学习长时序依赖关系并给出多步预测值。试验通过将预测问题转化为监督学习问题建立半实时仿真环境，并重点研究了观测窗口、预测窗口、网络结构等对性能指标的影响。对比LSTM、SARIMA、EXP，结果表明LSTM具备优异的线性拟合能力和良好的非线性关系映射能力。LSTM预测方法摆脱了传统方法受制于专家经验和模型精度低等问题，为开展实时遥测参数预测奠定了基础。     

Univariate approach to the monthly total ozone time series over Kolkata,India: autoregressive integrated moving average (ARIMA) and autoregressive neural network (AR-NN) models

This study reports univariate modelling methodologies applied to the monthly total ozone concentration (TOC) over Kolkata (22°32′, 88°20′), India, derived from the measurements made by the Earth Probe Total Ozone Mapping Spectrometer (EP/TOMS). The univariate models have been generated in two forms, namely autoregressive integrated moving average (ARIMA) and autoregressive neural network (AR-NN). Three ARIMA models in the forms of ARIMA(1,1,1), ARIMA(0,1,1) and ARIMA(0,2,2) and 11 autoregressive neural network models, AR-NN(n), have been generated for a time series. Goodness of fit of the models to the time series of monthly TOC has been assessed using prediction error, Pearson correlation coefficient and Willmott's indices. After rigorous skill assessment, the ARIMA (0,2,2) has been identified as the best predictive model for the time series under study.     

Combining seasonal time series ARIMA method and neural networks with genetic algorithms for predicting the production value of the mechanical industry in Taiwan

Supplying industrial firms with an accurate method of forecasting the production value of the mechanical industry to facilitate decision makers in precise planning is highly desirable. Numerous methods, including the autoregressive integrated-moving average (ARIMA) model and artificial neural networks can make accurate forecasts based on historical data. The seasonal ARIMA (SARIMA) model and artificial neural networks can also handle data involving trends and seasonality. Although neural networks can make predictions, deciding the most appropriate input data, network structure and learning parameters are difficult. Therefore, this article presents a hybrid forecasting method that combines the SARIMA model and neural networks with genetic algorithms. Analytical results generated by the SARIMA model are inputted as the input data of a neural network. Subsequently, the number of neurons in the hidden layer and the number of learning parameters of the neural network architecture are globally optimized using genetic algorithms. This model is subsequently adopted to forecast seasonal time series data of the production value of the mechanical industry in Taiwan. The results presented here provide a valuable reference for decision makers in industry.     

Land surface air temperature mapping using TOVS and AVHRR

Surface air temperature is an important variable in land surface hydrological studies. This paper evaluates the ability of satellites to map air temperature across large land surface areas. Algorithms recently have been developed that derive surface air temperature using observations from the TOVS (TIROS Operational Vertical Sounder) suite of instruments and also from the AVHRR (Advanced Very High Resolution Radiometer), which have flown on the NOAA operational sun synchronous satellites TIROS-N NOAA-14. In this study we evaluate TOVS soundings from NOAA-10 (nominal local time of overpass 7:30 a.m./p.m.) and data from AVHRR aboard NOAA-9 (nominal local time 2:30 a.m./p.m.). Instantaneous estimates from the AVHRR and TOVS were compared with the hourly ground observations collected from 26 meteorological stations in the Red River-Arkansas River basin for a 3-month period from May to July 1987. Detailed comparisons between the satellite and ground estimates of surface air temperatures are reported and the feasibility of estimating the diurnal variation is explored. The comparisons are interpreted in the geographical context, i.e. land cover and topography, and in the seasonal context, i.e. early and midsummer. The results show that the average bias over the 3-month period compared with ground-based observations is approximately 2°C or less for the three times of day with TOVS having lower biases than AVHRR. Knowledge of these error estimates will greatly benefit use of satellite data in hydrological modelling.     

An active-fire based burned area mapping algorithm for the MODIS sensor

We present an automated method for mapping burned areas using 500-m Moderate Resolution Imaging Spectroradiometer (MODIS) imagery coupled with 1-km MODIS active fire observations. The algorithm applies dynamic thresholds to composite imagery generated from a burn-sensitive vegetation index and a measure of temporal texture. Cumulative active fire maps are used to guide the selection of burned and unburned training samples. An accuracy assessment for three geographically diverse regions (central Siberia, the western United States, and southern Africa) was performed using high resolution burned area maps derived from Landsat imagery. Mapped burned areas were accurate to within approximately 10% in all regions except the high-tree-cover sub-region of southern Africa, where the MODIS burn maps underestimated the area burned by 41%. We estimate the minimum detectable burn size for reliable detection by our algorithm to be on the order of 120 ha.     

Global land monitoring from AVHRR: potential and limitations

Global Vegetation Index ( GVI) time series of visible, near-IR and thermal IR Advanced Very High Resolution Radiometer (AVHRR)weekly composite data with a 015^° spatial resolution collected from NOAA-9 and -11 satellites have been used to develop a prototype global land monitoring system. The system is based on standardized anomalies of the Normalized Difference Vegetation Index (NDVI) and channel 4 brightness temperature ( T₄ )for the period April 1985-September 1994. Processing included: post-launch updated calibration; cloud screening; filling in the cloud induced data gaps by monthly averaging and spatial interpolation; suppressing residual noise by smoothing; calculating 5-year monthly means and standard deviations of NDVI and T₄and their standardized anomalies. The derived anomalies show potential for detecting and interpreting the seasonal cycle and statistically significant interannual variability. Yet, discontinuities and residua! trends can be traced in time series of NDVI and T4. Discontinuities result from the switch from NOAA-9 to NOAA-11 in 1988, and the Mount Pinatubo eruption in 1991. Trends are a combined effect of satellite orbit drift and a possible persistent error in post-launch calibration of solar channels. The orbit drift affects the solar and thermal IR channels through systematic variation of illumination geometry and diurnal heating/cooling of the surface and atmosphere, respectively. Examples are given to illustrate the magnitude of these effects, which reduce the ability to monitor small year-to-year surface changes. The present system yields more accurate results in geographic regions, where atmospheric, angular and diurnal variability effects have a lesser impact on the derived anomalies, i.e. over vegetated areas outside the tropics during local summers. For global-scale monitoring, angular, atmospheric and diurnal variability corrections must be incorporated in the present system.