Paul Doraiswamy

Demonstration of the accuracy of improved-resolution hyperspectral imagery

Numerous researchers have demonstrated the accuracy and utility of improved spatial resolution multispectral imagery by sharpening it with higher spatial resolution panchromatic imagery. A much more limited number of researchers have sharpened hyperspectral imagery with panchromatic imagery. In this research we have developed an algorithm that spatially sharpens specific ranges of hyperspectral bands with spectrally correlated multispectral bands of a higher spatial resolution to improve the spatial resolution of the hyperspectral imagery while maintaining or improving its spectral fidelity. Preliminary validation of the algorithm has been conducted using a 7m AVIRIS scene of the Maryland Eastern Shore containing corn, soybean, and wheat fields. This data was used to simulate 28m HSI and 7m MSI that were used in the sharpening process. Initial analysis has verified the spectral accuracy of the sharpened data. In the next phase of the study, airborne spectral data from two different sensors will be used in the sharpening process with the results used as input for USDA/ARS crop yield and stress models.

Crop Rotation Changes in Iowa Due to Ethanol Production

Iowa is the largest corn acreage state with 14.2 million acres which is an increase of 1.6 million acres from 2006. Two sources of data were used for this study. First, NASS calculates the number of acres planted each year for each county. This data is useful to determine general trends; however it will not show where the crops were planted within a county. Secondly, National Agricultural Statistics Service (NASS) has created a land cover classification image each year from either Landsat TM or Advanced Wide Field Sensor (AWiFS) imagery from 2001 through 2007. These images show the location of where each crop is being planted and when combined with each other can be used to determine crop rotations. By analyzing the NASS state acreage reports from 2001-2007, it does not appear that new areas are being put into cultivation. The total area used in corn and soybean production has remained the same throughout this time period. Thus increases in corn acreages have come from decreases in soybean acreages. Based on the NASS land cover classification, the area for the state in a continuous corn rotation has increased from 15.2% in 2001-2002 to 20.9% in 2006-2007. Areas in continuous soybean rotation have diminished from 8.1% in 2001-2002 to 4.3% in 2006-2007. There are thirty three counties which have consistently high accuracies for the 2001-2007 time period. These counties also show that the increase in corn acreages have come from soybean acreages, not from increasing overall agricultural areas.

Application of MODIS-derived parameters for regional yield assessment

Monitoring regional agricultural crop condition has traditionally been accomplished using NOAA AVHRR (1 km) data. New methods are developed for assessing crop yields by retrieving biophysical parameters from remotely sensed imagery and integrating with crop simulation models. The MODIS imagery with its 250 m resolution and a potentially daily coverage offers an opportunity for operational applications. The objective of this research was to assess the potential application of MODIS data for operational crop condition and yield estimates. A field study was conducted during the 2000 crop season in McLean county Illinois (IL), USA. Twenty corn and soybean fields were monitored with measurements for crop reflectance, Leaf area Index (LAI) and other crop growth parameters. A radiative transfer model was used to independently develop the LAI from the MODIS 250-m data. Crop growth parameters retrieved from the imagery were integrated in a crop yield simulation model. The magnitude and spatial variability of estimated LAI and the NASA product was partly due to differences in the classification of crop type and the pixel resolutions. A comparison with the NASA derived MODIS vegetation parameters and independently derived parameters are presented.

Evaluation of MODIS data for assessment of regional spring wheat yield in Kazakhstan

The new MODIS satellite imagery offers an opportunity for operational applications. The 1-km AVHRR imagery is not suitable for retrieval of field level parameter and the Landsat data is not frequent enough for monitoring changes in crop parameters during the critical crop growth periods. The objective of this research is to evaluate the application of MODIS data for crop yields assessment in the spring wheat belt of north Kazakhstan. Field measurements during the season included crop reflectance, leaf area index (LAI), biomass and final yields. A radiative transfer model was used to develop the LAI parameters from the MODIS 250-m data. The magnitude and spatial variability of crop yield estimates were compared with ground-based yield samples. Spring wheat crop yields were simulated using a crop model with LAI generated from MODIS imagery. Yield results were within the limits reported by the Kazakhstan Ministry of Agriculture.

Estimation of weed infestation in spring crops using MODIS data

Spring crops are a monoculture in Northern Kazakhstan. Mid-resolution satellite data of red and near infrared spectral bands from Terras MODIS satellite can be used for assessment of regional land use. The objective of this research was the analysis of weed infestation in spring grain crops. Weed infestation was classified into three categories, namely: minimum, average and maximum. The analysis of spectral characteristics of fields during the 2002 vegetative period permitted development of criteria to distinguish all three classes of weed infestation. It turned out, that the degree of weed population appeared to be closely linked with a number of uncultivated fallow fields. So the land use classification map was used to define the fallow fields in each farm in the oblasts. The estimation of weed population and its influence on crop productivity was derived from these analyses. The study was conducted for about 10 million hectare of spring crops in five oblasts.

MODIS applications for mapping regional crop yields

Monitoring crop condition and yields at regional scales using traditional operational NOAA AVHRR data remains challenging. The 1 km spatial resolution with two primary spectral bands is not adequate for development of field level canopy parameters. MODIS imagery offers an opportunity for daily coverage and adequate resolution required in operational applications. The objective of this research is to investigate the applicability of the 8-day MODIS composite data in the operational programs for crop condition and yield assessment. A field study was conducted in the predominantly corn and soybean area of Iowa in the Upper Midwest U.S. Crop yields were simulated at 250 m resolution and results were mapped for the study area.

Comparison of different MODIS data product collections over an agricultural area

Standard data products from NASA’s moderate resolution imaging spectroradiometer (MODIS) were available at launch (collection 3) and have undergone two revisions (collections 4 and 5) during the continuing Terra and Aqua missions. In 2000, a research project was conducted in large fields of corn and soybean to evaluate MODIS leaf area index (LAI) and land cover type (MOD15 and MOD12 data products, respectively) as input to crop yield models. Our objective was to compare collections 3, 4 and 5 with the ground data to determine data product improvement. Classification of land cover type for collections 3, 4 and 5 were similar to the USDA-NASS Cropland Data Layer. The collection 5 MOD15 LAI was considerably improved over earlier collections when the quality assurance flags indicated good LAI retrievals. Land surface reflectances (MOD09) of MODIS band 2 (near-infrared, 250-m) were used as inputs to an inversion of the scattering by arbitrarily inclined leaves (SAIL) model. Compared to the collection 5 MOD15 product, SAIL-derived LAI had approximately equal agreement with the field data and had less systematic bias in root mean square error (RMSE). MODIS data products were designed to address global scientific questions without reliance on ground data, whereas the SAIL model inversions required ground data for model inputs. The accuracy of MODIS observations was not limiting LAI accuracy for broadleaf crops; additional information from non-MODIS sources may be required for improved MOD15 LAI.

Monitoring of crop rotation and weed infestation estimation using MODIS

This research was conducted to develop the method of weed infestation assessment in the predominantly wheat areas in Northern Kazakhstan. Standard classification procedures were not applicable to separate the weed population, therefore indirect methods were investigated. Multiyear land use inventory was used and the correlation between the number of sowing years in fallow-crop rotation and the weed population was determined. Historical and current MODIS reflectance data (band 2, 250 m resolution) was used for spring wheat classification and for delineation of fallow areas for the last years. According to the number of sowing years after fallow four classes as 1st, 2nd, 3rd, 4th and others culture after fallow masks were created. Ground surveys were conducted to define the weed population in cereal fields for these classes. It was demonstrated that this approach was the perspective tool to estimate the weed infestation level for croplands in Northern Kazakhstan. The result of this research is important for grain production forecast