Land Cover and Land Use: Description of Vegetation Cover during the Growth Period and Crop Classification with Multitemporal High Resolution Spot Images

Remote sensing data provide required inputs for environmental modelling at regional to global scales and play an important role in monitoring agricultural land use with regard to EU agricultural subsidy and sustainable watershed management. To characterise the dynamics of agricultural parcels, it is important to have information about the crop type and the vegetation cover (VC) during the year. In this study, the potential of multitemporal high resolution satellite-images (SPOT) for monitoring agricultural land use was evaluated. The study area is located in the Eifel region (49°8’N, 6°5’E), in the South-West of Germany. Data sets used in this study cover the growing season between April and September 2003 and comprise five satellite scenes from SPOT 4 and SPOT 5 systems. The data was geometrically and radiometrically corrected using standard procedures. To ensure which crop types can be separated with the available image data, a Spectral Mixture Analysis (SMA) was executed for every SPOT image by using the image endmembers vegetation, soil, and shadow. With the vegetation abundance image, vegetation cover time profiles identification of the crop types rape, maize, winter and summer corn appeared feasible. For crop classification we applied Isodata, Maximum Likelihood (ML) and the Spectral Angle Mapper (SAM) and compared their classification accuracy. The study shows that with the used multitemporal high resolution satellite-images (SPOT) different crop types can be identify. Based on these results, crop and vegetation cover maps for different crop types in the study area were produced for the year 2003. INTRODUCTION For environmental modelling and to support the EU subsidy policy timely and accurate information on agriculture land cover/use are required (i, ii). The vegetation on agricultural land shows a high temporal dynamic. The condition of the crops changes throughout the growing season because crops pass through a series of phenological stages in which the plant’s physiology, reflected by its biochemistry and structural characteristics, constantly changes (iii). For monitoring this dynamic remote sensing provides an alternative technique to expensive, time consuming ground surveys. Different image classification techniques for getting information on the crop type exist. For the successful classification of crop types the use of multitemporal data is recommended (iv). The imagery must be acquired during key stages of crop growth development (v). The aim of the study was to evaluate the potential of multitemporal high resolution images for monitoring agricultural crops. For this aim we compared different classification methods and mono-/multitemporal datasets. STUDY AREA AND DATA SETS STUDY AREA The study area (468 km2) is located in the region Eifel, SW Germany (49°8’N, 6°5’E). The topography is characterised by highlands with an average elevation of about 300 metres. The Center for Remote Sensing of Land Surfaces, Bonn, 28-30 September 2006 81 mean annual temperature is about 7° to 9°C and the mean annual precipitation is about 800 mm. The main land use is agriculture and forestry. In the study area the typical crop types are: wheat, maize, rape, barley, triticale and forage crops. The crops are cultivated in typical crop rotation systems. The agricultural parcels are small (about 100x100m2) compared to other agricultural areas in Europe. DATA SETS For this study five SPOT scenes with less than 5% cloud cover have been acquired during the growing season between April and September 2003 (Table 1). The geometric resolution of the SPOT 4 and SPOT 5 system is best suited to analyse the different land cover types of the small agriculture parcels in the study area. SPOT 4 and SPOT 5 collect data in 4 spectral band covering VIS and near-IR domains. Table 1: Used Sat-Images Sensor Image Date Θv sensor view angle geometric resolution SPOT 4 21-04-2003 19.5° 20m SPOT 4 07-05-2003 9.0° 20m SPOT 4 22-06-2003 28.9° 20m SPOT 5 03-08-2003 16.4° 10m SPOT 5 14-10-2003 3.2 ° 10m Ground truth data from 107 test fields were collected based on interviews with farmers. For these parcels the information about the crops (maize, summer barley and winter barley, summer wheat and winter wheat, rape and grassland) that were cultivated 2003 was available. This ground truth data was split into training fields and validation fields (Table 2). Phenology data from 2003 was provided by the National Meteorological Service (DWD). Other crop types cover only limited areas and were not considered for the classification. Table 2: Ground truth, validation and training data Training data Validation data Name Pixel number of fields Name Pixel number of fields summer corn 4753 8 summer corn 1164 25 winter corn 1683 4 winter corn 702 11 maize 3494 5 maize 836 15 rape 4108 5 rape 1397 15 grassland 2437 5 grassland 1413 14 TOTAL 16475 27 TOTAL 5512 80 DATA PRE-PROCESSING The SPOT data were georectified and atmospherically corrected. To make the various multi temporal scenes spatially comparable, first step in data processing was the geometric correction of each image by using vector data and a 20 metre digital elevation model (DEM) from Rhineland Palatinate’s mapping agency. Proceedings of the 2 Workshop of the EARSeL SIG on Land Use and Land Cover 82 For atmospheric correction first a Landsat 5 image (4 Aug. 2003) was radiometrically corrected by using a 5S-based parametric model (vi). After correction of this Landsat image the SPOT data were radiometrically corrected using an empirical-line approach (Figure 1). Figure 1: Radiometric and Geometric Correction ASSESSING THE VEGETATION COVER DYNAMICS Methods To ensure if a separation of the crop types would be feasible with the available image data, a profound analysis of the temporal crop growth dynamics was mandatory. To analyse this growth dynamic the vegetation cover (VC) for each SPOT image was calculated. The vegetation cover is defined as the vertical projected fraction of the green photosynthetic active vegetation. We applied a Linear Spectral Mixture Analysis (SMA) for every image to get the temporal profile of the vegetation cover during the growing season. These profiles helped us interpreting the classification results and to find the best day and best input combination for crop type mapping in the study area. LINEAR SPECTRAL MIXTURE ANALYSIS (SMA) SMA assumes that the overall signal of a pixel is the sum of signals (endmembers) from different surfaces (vii, viii, ix, x). The SMA separates the spectral information of each pixel into its constituent’s surface abundances (1).