Response of seasonal vegetation development to climatic variations in eastern central Asia

Meteorological records show that central Asia has experienced one of the strongest warming signals in the world over the last 30 years. The objective of this study was to examine the seasonal vegetation response to the recent climatic variation on the Mongolian steppes, the third largest grassland in the world. The onset date of green-up for central Asia was estimated using time-series analysis of advanced very high resolution radiometer (AVHRR) normalized difference vegetation index (NDVI) biweekly composite data collected between January 1982 and December 1991. Monthly precipitation and mean temperature data (1982-1990) were acquired from 19 meteorological stations throughout the grasslands of the eastern Mongolian steppes in China. Our results showed that while the taiga forest north of the Mongolian steppes (>50°N) experienced an earlier onset of green-up during the study period, a later onset was observed at the eastern and northern edges of the Gobi Desert (40°N-50°N). Responses of different vegetation types to climatic variability appeared to vary with vegetation characteristics and spring soil moisture availability of specific sites. Plant stress caused by drought was the most significant contributor to later vegetation green-up as observed from satellite imagery over the desert steppe. Areas with greater seasonal soil moisture greened up earlier in the growing season. Our results suggested that water budget limitations determine the pattern of vegetation responses to atmospheric warming.     

Assessing vegetation response to drought in the northern Great Plains using vegetation and drought indices

The Normalized Difference Vegetation Index (NDVI) derived from the Advanced Very High Resolution Radiometer (AVHRR) has been widely used to monitor moisture-related vegetation condition. The relationship between vegetation vigor and moisture availability, however, is complex and has not been adequately studied with satellite sensor data. To better understand this relationship, an analysis was conducted on time series of monthly NDVI (1989-2000) during the growing season in the north and central U.S. Great Plains. The NDVI was correlated to the Standardized Precipitation Index (SPI), a multiple-time scale meteorological-drought index based on precipitation. The 3-month SPI was found to have the best correlation with the NDVI, indicating lag and cumulative effects of precipitation on vegetation, but the correlation between NDVI and SPI varies significantly between months. The highest correlations occurred during the middle of the growing season, and lower correlations were noted at the beginning and end of the growing season in most of the area. A regression model with seasonal dummy variables reveals that the relationship between the NDVI and SPI is significant in both grasslands and croplands, if this seasonal effect is taken into account. Spatially, the best NDVI-SPI relationship occurred in areas with low soil water-holding capacity. Our most important finding is that NDVI is an effective indicator of vegetation-moisture condition, but seasonal timing should be taken into consideration when monitoring drought with the NDVI.     

Differences in Landsat-based trend analyses in drylands due to the choice of vegetation estimate

Drylands cover about 41% of the globe's surface and provide important ecosystem services, but land use and climate change exert considerable pressure on these ecosystems. Both of these drivers frequently result in gradual vegetation change and landscape-scale trend analysis based on yearly vegetation estimates can capture such changes. Such trend analyses based on high-resolution time series of satellite imagery have so far not widely been used and existing studies in drylands relied on different vegetation measures. Spectral mixture analysis (SMA) has been chosen due to its superiority to simpler vegetation estimates in quantifying vegetation cover in single-date studies, however SMA can be challenging to implement for large areas. Here, we quantify the trade-off involved when using simple vegetation estimates instead of SMA fractions for subsequent trend analyses. We calculated NDVI, SAVI and Tasseled Cap Greenness, as well as SMA green vegetation fractions for a time series of Landsat images from 1984-2005 for a study region in Crete. Linear trend analysis showed that trend coefficients and the spatial patterns of trends were similar across all vegetation estimates and the entire study region, especially for areas where vegetation changed gradually. On average, trends based on simple measures differed less than 5% from SMA-based trends with decreasing similarity in trend results from Tasseled Cap Greenness to SAVI and NDVI. Vegetation estimates differed markedly in their response to disturbance events such as fires. Trend analyses based on qualitative measures can easily be applied across very large areas and using multi-sensor time series based on high-resolution data. While the subtle differences between vegetation estimates may still be important for some applications, the robustness of trend analyses regarding the choice of vegetation estimate bears considerable promise to reconstruct fine-scale vegetation dynamics and land use histories and to assess climate change impacts on the world's drylands.     

Sensitivity of vegetation indices to substrate brightness in hyper-arid environment: The Makhtesh Ramon Crater (Israel) case study

The influence of soil background on most vegetation indices (VIs) derived from remotely sensed imagery is a well known phenomenon, and has generated interest in the development of indices that would be less sensitive to this influence. Several such indices have been developed thus far. This paper focuses on testing and comparing the sensitivity of seven intensively used, Landsat Thematic Mapper (TM) derived, VIs (NDVI, SAVI, MSAVI, PVI, WDVI, SAVI 2 and TSAVI) to bare surface variation with almost no vegetation signal. The study was conducted on a terrain composed of a high variety of bare surface materials of which basalt and gypsum are two extremely dark and bright substrates respectively. It was found that SAVI and MSAVI respond to bare surface material very similarly. Such close similarity was also found between PVI and WDVI, and between SAVI 2 and TSAVI. NDVI tends to be overestimated on dark surfaces, while SAVI, PVI and TSAVI show more sensitivity to bright surfaces. Comparison between DeltaVI (the difference between pairs of VIs) and the brightness of the different surface materials showed a high correlation in each case, which underlines the fact that the response of different VIs to bare surface variation is mainly related to the surface brightness.     

Hyperspectral vegetation indices and novel algorithms for predicting green LAI of crop canopies: Modeling and validation in the context of precision agriculture

A growing number of studies have focused on evaluating spectral indices in terms of their sensitivity to vegetation biophysical parameters, as well as to external factors affecting canopy reflectance. In this context, leaf and canopy radiative transfer models are valuable for modeling and understanding the behavior of such indices. In the present work, PROSPECT and SAILH models have been used to simulate a wide range of crop canopy reflectances in an attempt to study the sensitivity of a set of vegetation indices to green leaf area index (LAI), and to modify some of them in order to enhance their responsivity to LAI variations. The aim of the paper was to present a method for minimizing the effect of leaf chlorophyll content on the prediction of green LAI, and to develop new algorithms that adequately predict the LAI of crop canopies. Analyses based on both simulated and real hyperspectral data were carried out to compare performances of existing vegetation indices (Normalized Difference Vegetation Index [NDVI], Renormalized Difference Vegetation Index [RDVI], Modified Simple Ratio [MSR], Soil-Adjusted Vegetation Index [SAVI], Soil and Atmospherically Resistant Vegetation Index [SARVI], MSAVI, Triangular Vegetation Index [TVI], and Modified Chlorophyll Absorption Ratio Index [MCARI]) and to design new ones (MTVI1, MCARI1, MTVI2, and MCARI2) that are both less sensitive to chlorophyll content variations and linearly related to green LAI. Thorough analyses showed that the above existing vegetation indices were either sensitive to chlorophyll concentration changes or affected by saturation at high LAI levels. Conversely, two of the spectral indices developed as a part of this study, a modified triangular vegetation index (MTVI2) and a modified chlorophyll absorption ratio index (MCARI2), proved to be the best predictors of green LAI. Related predictive algorithms were tested on CASI (Compact Airborne Spectrographic Imager) hyperspectral images and, then, validated using ground truth measurements. The latter were collected simultaneously with image acquisition for different crop types (soybean, corn, and wheat), at different growth stages, and under various fertilization treatments. Prediction power analysis of proposed algorithms based on MCARI2 and MTVI2 resulted in agreements between modeled and ground measurement of non-destructive LAI, with coefficients of determination (r²) being 0.98 for soybean, 0.89 for corn, and 0.74 for wheat. The corresponding RMSE for LAI were estimated at 0.28, 0.46, and 0.85, respectively.     

A detection problem: Sensitivity and uncertainty analysis of a land surface temperature approach to detecting dynamics of water use by groundwater-dependent vegetation

Sustainable management of groundwater-dependent vegetation (GDV) requires the accurate identification of GDVs, characterisation of their water use dynamics and an understanding of associated errors. This paper presents sensitivity and uncertainty analyses of one GDV mapping method which uses temperature differences between time-series of modelled and observed land surface temperature (LST) to detect groundwater use by vegetation in a subtropical woodland. Uncertainty in modelled LST was quantified using the Jacobian method with error variances obtained from literature. Groundwater use was inferred where modelled and observed LST were significantly different using a Student's t-test. Modelled LST was most sensitive to low-range wind speeds (<1.5 m s⁻¹), low-range vegetation height (<=0.5 m), and low-range leaf area index (<=0.5 m² m⁻²), limiting the detectability of groundwater use by vegetation under such conditions. The model-data approach was well-suited to detection of GDV because model-data errors were lowest for climatic conditions conducive to groundwater use.     

The vegetation cycle in West Africa from AVHRR-NDVI data: Horizons of predictability versus spatial scales

The predictability of the vegetation cycle is analyzed as a function of the spatial scale over West Africa during the period 1982-2004. The NDVI-AVHRR satellite data time series are spatially aggregated over windows covering a range of sizes from 8 × 8 km² to 1024 × 1024 km². The times series are then embedded in a low-dimensional pseudo-phase space using a system of time delayed coordinates. The correlation dimension (D_c) and entropy of the underlying vegetation dynamics, as well as the noise level (σ) are extracted from a nonlinear analysis of the time series. The horizon of predictability (H_P) of the vegetation cycle defined as the time interval required for an n% RMS error on the vegetation state to double (i.e. reach 2n% RMS) is estimated from the entropy production. Compared to full resolution, the intermediate scales of aggregation (in the range of 64 × 64 km² to 256 × 256 km²) provide times series with a slightly improved signal to noise ratio, longer horizon of predictability (about 2 to 5 decades) and preserve the most salient spatial patterns of the vegetation cycle. Insights on the best aggregation scale and on the expected vegetation cycle predictability over West Africa are provided by a set of maps of the correlation dimension (D_c), the horizon of predictability (H_P) and the level of noise (σ).     

An application of MODIS data to snow cover monitoring in a pastoral area: A case study in Northern Xinjiang, China

Snow is an important land cover on the earth's surface. It is characterized by its changing nature. Monitoring snow cover extent plays a significant role in dynamic studies and prevention of snow-caused disasters in pastoral areas. Using NASA EOS Terra/MODIS snow cover products and in situ observation data during the four snow seasons from November 1 to March 31 of year 2001 to 2005 in northern Xinjiang area, the accuracy of MODIS snow cover mapping algorithm under varied snow depth and land cover types was analyzed. The overall accuracy of MODIS daily snow cover mapping algorithm in clear sky condition is high at 98.5%; snow agreement reaches 98.2%, and ranges from 77.8% to 100% over the 4-year period for individual sites. Snow depth (SD) is one of the major factors affecting the accuracy of MODIS snow cover maps. MODIS does not identify any snow for SD less than 0.5 cm. The overall accuracy increases with snow depth if SD is equal to or greater than 3 cm, and decreases for SD below 3 cm. Land cover has an important influence in the accuracy of MODIS snow cover maps. The use of MOD10A1 snow cover products is severely affected by cloud cover. The 8-day composite products of MOD10A2 can effectively minimize the effect of cloud cover in most cases. Cloud cover in excess of 10% occurs on 99% of the MOD10A1 products and 14.7% of the MOD10A2 products analyzed during the four snow seasons. User-defined multiple day composite images based on MOD10A1, with flexibilities of selecting composite period, starting and ending date and composite sequence of MOD10A1 products, have an advantage in effectively monitoring snow cover extent for regional snow-caused disasters in pastoral areas.     

A scalable approach to mapping annual land cover at 250 m using MODIS time series data: A case study in the Dry Chaco ecoregion of South America

Land use and land cover (LULC) maps from remote sensing are vital for monitoring, understanding and predicting the effects of complex human-nature interactions that span local, regional and global scales. We present a method to map annual LULC at a regional spatial scale with source data and processing techniques that permit scaling to broader spatial and temporal scales, while maintaining a consistent classification scheme and accuracy. Using the Dry Chaco ecoregion in Argentina, Bolivia and Paraguay as a test site, we derived a suite of predictor variables from 2001 to 2007 from the MODIS 250 m vegetation index product (MOD13Q1). These variables included: annual statistics of red, near infrared, and enhanced vegetation index (EVI), phenological metrics derived from EVI time series data, and slope and elevation. For reference data, we visually interpreted percent cover of eight classes at locations with high-resolution QuickBird imagery in Google Earth. An adjustable majority cover threshold was used to assign samples to a dominant class. When compared to field data, we found this imagery to have georeferencing error < 5% the length of a MODIS pixel, while most class interpretation error was related to confusion between agriculture and herbaceous vegetation. We used the Random Forests classifier to identify the best sets of predictor variables and percent cover thresholds for discriminating our LULC classes. The best variable set included all predictor variables and a cover threshold of 80%. This optimal Random Forests was used to map LULC for each year between 2001 and 2007, followed by a per-pixel, 3-year temporal filter to remove disallowed LULC transitions. Our sequence of maps had an overall accuracy of 79.3%, producer accuracy from 51.4% (plantation) to 95.8% (woody vegetation), and user accuracy from 58.9% (herbaceous vegetation) to 100.0% (water). We attributed map class confusion to limited spectral information, sub-pixel spectral mixing, georeferencing error and human error in interpreting reference samples. We used our maps to assess woody vegetation change in the Dry Chaco from 2002 to 2006, which was characterized by rapid deforestation related to soybean and planted pasture expansion. This method can be easily applied to other regions or continents to produce spatially and temporally consistent information on annual LULC.