Yingxin Gu

A five-year analysis of MODIS NDVI and NDWI for grassland drought assessment over the central Great Plains of the United States

Received 18 December 2006; revised 16 February 2007; accepted 28 February 2007; published 27 March 2007. [1] A five-year (2001–2005) history of moderate resolution imaging spectroradiometer (MODIS) normalized difference vegetation index (NDVI) and normalized difference water index (NDWI) data was analyzed for grassland drought assessment within the central United States, specifically for the Flint Hills of Kansas and Oklahoma. Initial results show strong relationships among NDVI, NDWI, and drought conditions. During the summer over the Tallgrass Prairie National Preserve, the average NDVI and NDWI were consistently lower (NDVI 0.6 and NDWI>0.4). NDWI values exhibited a quicker response to drought conditions than NDVI. Analysis revealed that combining information from visible, near infrared, and short wave infrared channels improved sensitivity to drought severity. The proposed normalized difference drought index (NDDI) had a stronger response to summer drought conditions than a simple difference between NDVI and NDWI, and is therefore a more sensitive indicator of drought in grasslands than NDVI alone.Citation: Gu, Y., J. F. Brown, J. P. Verdin, and B. Wardlow (2007), A five-year analysis of MODIS NDVI and NDWI for grassland drought assessment over the central Great Plains of the United States, Geophys. Res. Lett., 34, L06407, doi:10.1029/ 2006GL029127.

Phenological Classification of the United States: A Geographic Framework for Extending Multi-Sensor Time-Series Data

This study introduces a new geographic framework, phenological classification, for the conterminous United States based on Moderate Resolution Imaging Spectroradiometer (MODIS) Normalized Difference Vegetation Index (NDVI) time-series data and a digital elevation model. The resulting pheno-class map is comprised of 40 pheno-classes, each having unique phenological and topographic characteristics. Cross-comparison of the pheno-classes with the 2001 National Land Cover Database indicates that the new map contains additional phenological and climate information. The pheno-class framework may be a suitable basis for the development of an Advanced Very High Resolution Radiometer (AVHRR)-MODIS NDVI translation algorithm and for various biogeographic studies.

Detecting Ecosystem Performance Anomalies for Land Management in the Upper Colorado River Basin Using Satellite Observations, Climate Data, and Ecosystem Models

US Geological Survey Earth Resources Observation and Science Center, 47914 252nd Street, Sioux Falls, SD 57198, USA; E-Mail: wylie@usgs.gov * Author to whom correspondence should be addressed; E-Mail: ygu@usgs.gov; Tel.: +1-605-594-6576; Fax: +1-605-594-6529. Received: 26 June 2010; in revised form: 22 July 2010 / Accepted: 28 July 2010 / Published: 29 July 2010 Abstract: This study identifies areas with ecosystem performance anomalies (EPA) within the Upper Colorado River Basin (UCRB) during 2005–2007 using satellite observations, climate data, and ecosystem models. The final EPA maps with 250-m spatial resolution were categorized as normal performance, underperformance, and overperformance (observed performance relative to weather-based predictions) at the 90% level of confidence. The EPA maps were validated using “percentage of bare soil” ground observations. The validation results at locations with comparable site potential showed that regions identified as persistently underperforming (overperforming) tended to have a higher (lower) percentage of bare soil, suggesting that our preliminary EPA maps are reliable and agree with ground-based observations. The 3-year (2005–2007) persistent EPA map from this study provides the first quantitative evaluation of ecosystem performance anomalies within the UCRB and will help the Bureau of Land Management (BLM) identify potentially degraded lands. Results from this study can be used as a prototype by BLM and other land managers for making optimal land management decisions. Keywords: satellite remote sensing; MODIS NDVI; ecosystem performance; ecosystem performance anomalies; ecosystem models; climate data; land management

Identifying grasslands suitable for cellulosic feedstock crops in the Greater Platte River Basin: dynamic modeling of ecosystem performance with 250 m eMODIS

This study dynamically monitors ecosystem performance (EP) to identify grasslands potentially suitable for cellulosic feedstock crops (e.g., switchgrass) within the Greater Platte River Basin (GPRB). We computed grassland site potential and EP anomalies using 9‐year (2000–2008) time series of 250 m expedited moderate resolution imaging spectroradiometer Normalized Difference Vegetation Index data, geophysical and biophysical data, weather and climate data, and EP models. We hypothesize that areas with fairly consistent high grassland productivity (i.e., high grassland site potential) in fair to good range condition (i.e., persistent ecosystem overperformance or normal performance, indicating a lack of severe ecological disturbance) are potentially suitable for cellulosic feedstock crop development. Unproductive (i.e., low grassland site potential) or degraded grasslands (i.e., persistent ecosystem underperformance with poor range condition) are not appropriate for cellulosic feedstock development. Grassland pixels with high or moderate ecosystem site potential and with more than 7 years ecosystem normal performance or overperformance during 2000–2008 are identified as possible regions for future cellulosic feedstock crop development (ca. 68 000 km2 within the GPRB, mostly in the eastern areas). Long‐term climate conditions, elevation, soil organic carbon, and yearly seasonal precipitation and temperature are important performance variables to determine the suitable areas in this study. The final map delineating the suitable areas within the GPRB provides a new monitoring and modeling approach that can contribute to decision support tools to help land managers and decision makers make optimal land use decisions regarding cellulosic feedstock crop development and sustainability.

Monitoring the status of forests and rangelands in the Western United States using ecosystem performance anomalies

The effects of land management and disturbance on ecosystem performance (i.e. biomass production) are often confounded by those of weather and site potential. The current study overcomes this issue by calculating the difference between actual and expected ecosystem performance (EEP) to generate ecosystem performance anomalies (EPA). This study aims to delineate and quantify average EPA from 2000–2009 within the Greater Platte and Upper Colorado River Basins, USA. Moderate Resolution Imaging Spectroradiometer (MODIS) normalized difference vegetation index (NDVI) images averaged over the growing season (GSN) served as a proxy of actual ecosystem performance. Yearly EEP was determined with rule-based piecewise regression tree models of abiotic data (climate, soils, elevation, etc.), independently created for each land cover. EPA were calculated as the residuals of the EEP to GSN relationship, and characterized as normal performing, underperforming, and overperforming at the 90% confidence level. Validation revealed that EPA values were related to biomass production (R 2 = 0.56, P = 0.02) and likely to the proportion of biomass removed by livestock in the Nebraska Sandhills. Overall, 60.6% of the study area was (normal) performing near its EEP, 3.0% was severely underperforming, 5.0% was highly overperforming, and the remainder was slightly underperforming or overperforming. Generally, disturbances such as fires, floods, and insect damage, in addition to high grazing intensity, result in a negative EPA. Conversely, mature stands and appropriate management often result in positive EPA values. This method provides information critical to land managers to evaluate the appropriateness of previous management practices and restoration efforts and quantify disturbance impacts. Results are at a scale sufficient for many of the large management units of the region and for locating areas needing further investigation. Applications of EPA data to monitoring invasive species, grazing impacts, and vulnerability to plant community shifts have been suggested by land management professionals.

Mapping carbon flux uncertainty and selecting optimal locations for future flux towers in the Great Plains

Flux tower networks (e.g., AmeriFlux, Agriflux) provide continuous observations of ecosystem exchanges of carbon (e.g., net ecosystem exchange), water vapor (e.g., evapotranspiration), and energy between terrestrial ecosystems and the atmosphere. The long-term time series of flux tower data are essential for studying and understanding terrestrial carbon cycles, ecosystem services, and climate changes. Currently, there are 13 flux towers located within the Great Plains (GP). The towers are sparsely distributed and do not adequately represent the varieties of vegetation cover types, climate conditions, and geophysical and biophysical conditions in the GP. This study assessed how well the available flux towers represent the environmental conditions or “ecological envelopes” across the GP and identified optimal locations for future flux towers in the GP. Regression-based remote sensing and weather-driven net ecosystem production (NEP) models derived from different extrapolation ranges (10 and 50%) were used to identify areas where ecological conditions were poorly represented by the flux tower sites and years previously used for mapping grassland fluxes. The optimal lands suitable for future flux towers within the GP were mapped. Results from this study provide information to optimize the usefulness of future flux towers in the GP and serve as a proxy for the uncertainty of the NEP map.

An Optimal Sample Data Usage Strategy to Minimize Overfitting and Underfitting Effects in Regression Tree Models Based on Remotely-Sensed Data

Regression tree models have been widely used for remote sensing-based ecosystem mapping. Improper use of the sample data (model training and testing data) may cause overfitting and underfitting effects in the model. The goal of this study is to develop an optimal sampling data usage strategy for any dataset and identify an appropriate number of rules in the regression tree model that will improve its accuracy and robustness. Landsat 8 data and Moderate-Resolution Imaging Spectroradiometer-scaled Normalized Difference Vegetation Index (NDVI) were used to develop regression tree models. A Python procedure was designed to generate random replications of model parameter options across a range of model development data sizes and rule number constraints. The mean absolute difference (MAD) between the predicted and actual NDVI (scaled NDVI, value from 0–200) and its variability across the different randomized replications were calculated to assess the accuracy and stability of the models. In our case study, a six-rule regression tree model developed from 80% of the sample data had the lowest MAD (MADtraining = 2.5 and MADtesting = 2.4), which was suggested as the optimal model. This study demonstrates how the training data and rule number selections impact model accuracy and provides important guidance for future remote-sensing-based ecosystem modeling.

Downscaling 250-m MODIS Growing Season NDVI Based on Multiple-Date Landsat Images and Data Mining Approaches

The satellite-derived growing season time-integrated Normalized Difference Vegetation Index (GSN) has been used as a proxy for vegetation biomass productivity. The 250-m GSN data estimated from the Moderate Resolution Imaging Spectroradiometer (MODIS) sensors have been used for terrestrial ecosystem modeling and monitoring. High temporal resolution with a wide range of wavelengths make the MODIS land surface products robust and reliable. The long-term 30-m Landsat data provide spatial detailed information for characterizing human-scale processes and have been used for land cover and land change studies. The main goal of this study is to combine 250-m MODIS GSN and 30-m Landsat observations to generate a quality-improved high spatial resolution (30-m) GSN database. A rule-based piecewise regression GSN model based on MODIS and Landsat data was developed. Results show a strong correlation between predicted GSN and actual GSN (r = 0.97, average error = 0.026). The most important Landsat variables in the GSN model are Normalized Difference Vegetation Indices (NDVIs) in May and August. The derived MODIS-Landsat-based 30-m GSN map provides biophysical information for moderate-scale ecological features. This multiple sensor study retains the detailed seasonal dynamic information captured by MODIS and leverages the high-resolution information from Landsat, which will be useful for regional ecosystem studies.

Projecting future grassland productivity to assess the sustainability of potential biofuel feedstock areas in the Greater Platte River Basin

This study projects future (e.g., 2050 and 2099) grassland productivities in the Greater Platte River Basin (GPRB) using ecosystem performance (EP, a surrogate for measuring ecosystem productivity) models and future climate projections. The EP models developed from a previous study were based on the satellite vegetation index, site geophysical and biophysical features, and weather and climate drivers. The future climate data used in this study were derived from the National Center for Atmospheric Research Community Climate System Model 3.0 ‘SRES A1B’ (a ‘middle’ emissions path). The main objective of this study is to assess the future sustainability of the potential biofuel feedstock areas identified in a previous study. Results show that the potential biofuel feedstock areas (the more mesic eastern part of the GPRB) will remain productive (i.e., aboveground grassland biomass productivity >2750 kg ha−1 year−1) with a slight increasing trend in the future. The spatially averaged EPs for these areas are 3519, 3432, 3557, 3605, 3752, and 3583 kg ha−1 year−1 for current site potential (2000–2008 average), 2020, 2030, 2040, 2050, and 2099, respectively. Therefore, the identified potential biofuel feedstock areas will likely continue to be sustainable for future biofuel development. On the other hand, grasslands identified as having no biofuel potential in the drier western part of the GPRB would be expected to stay unproductive in the future (spatially averaged EPs are 1822, 1691, 1896, 2306, 1994, and 2169 kg ha−1 year−1 for site potential, 2020, 2030, 2040, 2050, and 2099). These areas should continue to be unsuitable for biofuel feedstock development in the future. These future grassland productivity estimation maps can help land managers to understand and adapt to the expected changes in future EP in the GPRB and to assess the future sustainability and feasibility of potential biofuel feedstock areas.

Mapping marginal croplands suitable for cellulosic feedstock crops in the Great Plains, United States

Growing cellulosic feedstock crops (e.g., switchgrass) for biofuel is more environmentally sustainable than corn‐based ethanol. Specifically, this practice can reduce soil erosion and water quality impairment from pesticides and fertilizer, improve ecosystem services and sustainability (e.g., serve as carbon sinks), and minimize impacts on global food supplies. The main goal of this study was to identify high‐risk marginal croplands that are potentially suitable for growing cellulosic feedstock crops (e.g., switchgrass) in the US Great Plains (GP). Satellite‐derived growing season Normalized Difference Vegetation Index, a switchgrass biomass productivity map obtained from a previous study, US Geological Survey (USGS) irrigation and crop masks, and US Department of Agriculture (USDA) crop indemnity maps for the GP were used in this study. Our hypothesis was that croplands with relatively low crop yield but high productivity potential for switchgrass may be suitable for converting to switchgrass. Areas with relatively low crop indemnity (crop indemnity <