Remote Sensing Vegetation
Remote sensing entails acquiring information about a given object or phenomenon without making physical contact with that object. Over time the application of data obtained from remote sensing has widen ranging from engineering, disaster management, agriculture, wildlife management, business, forestry among others. Remote sensing has been widely applied in mapping vegetation (Jensen, 2007). Globally remote sensing has been applied in forestry to detect level of forest destruction, types of vegetation, the health status of vegetation among others (Campbell, 2002).
This essay critically examines the concept of mapping vegetation using remotely sensed data, IKONOS and the type of sensor it uses and how it works. Additionally the concept of normalized difference vegetation index is succinctly covered, the differences between supervised and unsupervised classification and how it works, its importance in vegetation mapping and how global positioning system works forms part of this paper.
Mapping vegetation using remotely sensed satellite sensors
The ability to understand and estimate vegetation mass as well as productivity has become an important aspect particularly with the inception of remote sensing (Treitz & Howarth, 1999). This has made it possible for scientists to quantify as well as qualify amount and health of vegetation on the earth’s surface. It is worth noting that vegetation classification and vegetation mapping are important tasks in managing natural resources. Mapping vegetation also provides important information in understanding the natural and man-made environment by quantifying vegetation cover not only locally but also globally (Xie, Sha & Yu, 2008).
As suggested by Xie, Sha & Yu, 2008 with remote sensing, all of the major limitations of traditional methods of mapping vegetation are addressed. Remote sensing technology provides researchers with practical and economical means to study vegetation cover particularly over a long period of time and large areas. The technology also can be applied to map vegetation submerged in aquatic environment Al-Bakri & Taylor, 2003).
To map vegetation using remote sensing there are a number of steps to be followed. They include;Selecting the images to be used Carrying out radiometric calibration where necessary Carrying out geometric correction Enhancing the image Image classification which depends on the objective and expertise of the analyst and can either be supervised, unsupervised or hybrid classification Post-classification editing the follows This is then followed by classification accuracy assessment Detecting change if it is necessary The last step is creating maps either in soft copies or hard copies
It is worth mentioning that recently there is effort directed towards using hyper-spectral imagery to map vegetation. Here the imagery include hundreds of spectral bands making it suitable in vegetation studies since the absorption and reflection signatures from individual species as well as a mixed vegetation can be easily distinguished. An example of sensor which yield such imagery include AVIRIS which provide image through 224 channels or bands with wavelengths of between 400 and 2500 nm (Xie, Sha & Yu, 2008). Another approach used in vegetation mapping is through image fusion where image from different sensors are fused although they have multiple spatial resolutions. This has been shown to be effective due to its accuracy (Treitz & Howarth, 1999).
IKONOS sensors and how it works
IKONOS was launched back in 24 September 1999 in California. It is sun synchronous and it was the first to offer the public with high resolution images at 1.0-4.0 m (Space Imaging. 2001). The satellite has two sensors multispectral and panchromatic. The former collects images with spatial resolution of 1.0 m. On the other hand, multispectral sensor includes blue, green, red and near-infrared which collect information at 4.0m spatial resolution (Richards & Jia, 2006). These sensors have a revisit time of between 3 and 5 days and a ground swath of 11m. the satellite is at an altitude of 681km. it is worth noting that IKONOS can be set to acquire stereo IKONOS satellite image data useful in production of digital surface models or digital elevation models. Images are usually collected with 11- bit sensitivity which is later delivered in an unsigned 16-bit data format (Space Imaging. 2001). However to save the storage space, the collected data are rescaled to 8-bit which results in loss of information (Richards & Jia, 2006).
In terms of application, images acquired from sensors onboard IKONOS can be utilized in the following fields; agriculture, archaeology, coastal management, defense mapping, forestry, global warming, hurricane mitigation, land development, law enforcement, mining, cadastre and land records as well as environmental monitoring among others. In respect to this assignment, images acquired by IKONOS can be used in mapping vegetation cover at local scale as well as validating vegetation cover classified from other images such as Landsat (Space Imaging. 2001).
Normalized Difference Vegetation Index
Multi-spectral transformation involves using two bands with the intention of distinguishing subtle features or characteristics of a given feature that would otherwise be difficult. In this process data redundancy is reduced. One of the approaches under band ratioing is normalized difference vegetation index. According to Pinty & Verstraete, 1992 NDVI has been used as a graphical indicator employed in analyzing data obtained from satellites aimed at assessing whether the area covered contain live green plants or otherwise. The science behind determining the density of green vegetation on the earth’s surface is by observing distinct colors visible and near-infrared sunlight reflected by vegetations. Plants when stricken by sun energy absorb some energy as well as reflect some (Richards & Jia, 2006). The former is due to presence of chlorophyll which absorbs the visible light 0.4 to 0.7 µm used to manufacture food. However energy from 0.7 to 1.1 µm is reflected by the cells of vegetations. It has been shown that the number of leaves a plant has directly influences how visible and infrared energy are absorbed and reflected respectively (Pinty & Verstraete, 1992).
When considering the appearance of vegetation in both visible and near-infrared, they appear different in these two situations. In the former plants are darker while bare land is bright. In situation of near-infrared both vegetation and bare land are almost similar, bright in color. This thus offers the opportunity of comparing them hence making it possible for scientists to measure relative amount of vegetations (Richards & Jia, 2006). Generally more reflection of near-infrared wavelength than in visible portion suggests that the region has dense vegetation, presence of a forest. Very little differences in the intensity of visible and near-infrared wavelengths reflection means that the vegetation is sparsely distributed. Vegetations that are healthy higher percent of visible light reaching the leaves while unhealthy plants reflects more visible light and less near-infrared energy (Pinty & Verstraete, 1992).
Mathematically NDVI is “near-infrared radiation minus visible radiation divided by near-infrared radiation plus visible radiation” (Campbell, 2002). From the computation the value usually ranges between +1 and -1. Values close to 0 means sparsely distributed vegetations while values towards +1 indicate dense vegetation that is healthy. Apart from helping scientists to distinguish between areas with vegetation, NDVI has the potential of compressing the size of the data to be manipulated by a factor of 2. This helps in saving storage capacity as it reduces data redundancy. NDVI helps in monitoring vegetation conditions which is useful in providing early warning on occurrence of drought and famines. Similarly it has also been used to approximate evapotranspiration as well as identifying areas that are suitable for locust development (Richards & Jia, 2006).
Although NDVI is an important ration in mapping vegetation, there are a set of factors that negatively impact on its accurate use. These include atmospheric effects, clouds, soils effects, anisotropic effects as well as spectral effects. Due to the factors, there is need to exercise caution when using NDVI (Pinty & Verstraete, 1992).
Supervised and unsupervised classification
It is worth remembering that one of the main purposes of satellite remote sensing is to interpret the data as well as classify features. Having data that is not analyzed and presented in a manner that can help end users make decision is a waste of resources (Jensen, 2007). The field of remote sensing has in place mechanisms in which classification of data obtained this is a step towards deriving information from remotely sensed data. There are two main categories of image classification; supervised and unsupervised (Campbell, 2002). Ideally image classification entails efforts directed towards modifying data into information such as land cover among others.
When talking of supervised classification, it is worth noting that it is used to extract quantitative information from remotely sensed data. In this approach the analyst has at his or her disposal a set of known pixels which are then used to develop representative parameters for every class of interest. In other words, the analyst has a prior knowledge of the main themes in the area of interest before carrying out classification. For instance he or she is aware of let’s say water body and the exact geographic location (Richards & Jia, 2006). This is what has been termed at training zones. Ideally the analyst defines representative training areas for every class in the image. Later the computer is used to combine each class’ training areas and then calculates the necessary statistics such as mean, standard deviation covariance matrix among others. It is from these statistics that all the pixels with the exception of those which do not meet the predefined criteria are categorized into one of the classes. It is important to note that a pixel used in a certain class training area may end up being classified in another class (Campbell, 2002).
There are three main methods used to accomplish supervised classification; minimum-distance-to-mean method in which every pixel is assigned to the class to whose mean it’s closest. In parallelepiped methods, a pixel falls in a certain class if it’s within a specified number of standard deviations from the mean of the class. Lastly maximum likelihood method assumes that the training area pixels are normally distributed. To classify pixels, then their probability of occurring in each class is calculated. Then the pixel is assigned to the class with the highest probability or likelihood of it being a member. This approach is deemed to be more accurate compared to the other two (Richards & Jia, 2006).
Unsupervised classification is where one classifies the image without having advance (prior) information about classes of interest; rather the images were examined and broken into the most prevalent natural groupings present in the data. From this, the clusters then identified as land cover classes through a combination of visual interpretation and the topographic map provided or any others source of data as well as analysts experience.
The major steps in this approach include; even distribution of cluster centers within the data space, then every pixel is assigned to the cluster to whose mean it’s nearest, a new mean is then calculated for every cluster from the new group of pixels then the preceding step is repeated, in situations where the two means in each class differ by less than a predefined distance then the new clusters are adopted as final otherwise steps 2-4 are repeated a predefined number of times which is usually set by the analysts (Amarsaikhan & Douglas,2004).
Two main approaches used to accomplish unsupervised classification are K-means and ISODATA. In these two approaches pixels may not be classified if their distances to the respective means in the final iteration are over a predefined value. The major difference between these techniques is ISODATA allows for a range of classes and the K-means only a specific number of classes (Richards & Jia, 2006).
There is no doubt that the two techniques play a significant role when it comes to mapping vegetations. They make it possible for end user to clearly establish the different vegetations types, the boundaries between different vegetation, the extent of vegetation cover among others. This information is very vital to policy makers as well as other primary stakeholders. Additionally if data from previous years can be compared it is possible to detect change through such methods as post-classification comparison, multi-date composite image classification, change vector analysis or temporal image ratioing (Xie, Sha & Yu, 2008).
How GPS works
Global positioning system is a navigation satellite system which is in the space and offers location as well as time information regardless of weather conditions near or on the earth’s surface (Mendizabal, Berenguer & Melendez, 2009). It operates where there is no obstruction. This project came to being back in 1973 to help address the limitations of the previous navigation system. It became operational in 1994. Ideally there are a total of 24 satellites and additional 5 which are in space and receiver stations on the earth. The 5 satellites were launched so that they can provide data when one of the primary satellites malfunctions (Mendizabal, Berenguer & Melendez, 2009).
The GPS receivers usually calculate the position by precisely timing the signals received from GPS satellites that are in the orbit. It is worth noting that every satellite rotates the earth twice per day and continuously transmit information that include time of the message transmission and satellite position at the time the information is transmitted. From this information the receiver can approximate the transit time of information it receives and calculates the distance to every satellite. The distances as well as location of satellites are accomplished through trilateration usually based on the algorithm used to calculate the location of the receiver. It is worth noting that the position is then shown more often with a moving map or latitude and longitude may be included (Mendizabal, Berenguer & Melendez, 2009). It has been shown that a minimum of 3 satellites can successfully help in determining the position of objects on the earth’s surface. Nonetheless, there are some issues such as clock errors leads to large positional errors when the same is multiplied by the speed of light, for this reason, four satellites are used to calculate position and time of an object.
ETrex Vista is a devise used in mapping that combines the major features of eTrex summit having a full base map of Americas, the Atlantic or Pacific, a barometric altimeter, electronic compass as well as 24MB internal memory (Mendizabal, Berenguer & Melendez, 2009). With the ability of the device to use GPS to collect waypoints, this can be used in vegetation mapping. For instance with the potential to help user assign a name to a waypoint coupled with selecting a symbol or an icon to identifying waypoints on geographical map, it is possible for analysts to use this information during classification of vegetation. In other words, this can help in accurately determining training areas useful during supervised classification (Mendizabal, Berenguer & Melendez, 2009).
From the essay, it is evident that a number of issues with regards to vegetation mapping have been succinctly brought forth. The first section has addressed how remote sensing data can be used to map vegetation. The necessary steps in successfully doing so have been brought out. Additionally other types of vegetation mapping such as image fusion and using hyper-spectral imagery are mentioned. The next section talked about Normalized Difference Vegetation Index, the term was defined in terms of a mathematical function. Additionally the challenges facing its application as well as it use have been covered. The paper also addressed the two main methods used to classify data from remote sensing; supervised and unsupervised classification. The major types of each approach are also brought out clearly. The importances of the two methods which include classifying vegetation, detecting change among others are covered. The last section of the paper is about global positioning systems and how it works, the importance of eTrex having GPS and its application in vegetation mapping.