Literature Review

Biomass is considered a useful indicator of structural and functional attributes of forest ecosystems across a wide range of environmental conditions [7].  It is therefore important to accurately estimate biomass to assess the role of forests in meeting the region’s energy demand or in the global carbon (C) cycle, particularly when defining its contribution toward sequestering carbon. For an assessment of forest biomass, forest inventory is most commonly used and it differs depending on scope and purpose. Inventories are being designed to obtain information on other uses of the forest like recreation, grazing, wildlife and water conservation. It is designed to measure forest biomass rather than or in addition to traditional volume. The forest inventory area is usually one or more management units, each ranging in size from a few hundred to many thousand hectares. Each unit may be divided into forest-based strata or administrative subpopulations for which separate estimates are required. The attribute of primary interest is merchantable wood volume, with stem frequency. Basal area data are of secondary importance. These attributes are usually given by tree size classes and by a number of forest and administrative classes that are described in a classification system. Above ground standing biomass of trees is the weight of trees above ground, in a given area, if harvested at a given time. The change in standing biomass over a period of time is called productivity. The standing biomass helps to estimate the productivity of an area and also gives information on the carrying capacity of land. It also helps in estimating the biomass that can be continuously extracted. Various studies have estimated forest biomass at regional and national levels based on forest inventory data, such as forest species, area, stem volume, age class, and site class [15,16,17,18,19]. The standing biomass is computed agroclimate zonewsie for a federal state in India using the remote sensing data along with the field data [15]. Archer  [20] provides a sound sampling frame for energy analysis at the national level using Advanced Very High Resolution Radiometer (AVHRR) data.  Agro ecological zonation  was done to create the link between supply and demand and provided a valid basis for extrapolating the result of the supply survey to the national level in Pakistan using GIS and remote sensing. Automated selection of sampling units from digitally classified remote sensing data was effective, with overall 93% of woody biomass primary sampling units containing measurable woody vegetation. The results indicated a reasonable consistency between zones, the forest / highlands indicate an expected wood surplus and the semi arid zone indicates a large deficit [20].

Remote sensing methods have proved to be successful in mapping and monitoring forest health and distribution when a sufficiently small ground resolution is used. Supervised, unsupervised and hybrid classification methods were evaluated for their accuracy in discriminating dead and dying tree crowns from bare areas and the surrounding forest mosaic utilising 1-m resolution remote sensing data and the hybrid classifier significantly outperformed the other methods, producing low omission and commission errors among information classes [21]. Spectral class differences are related to leaf nitrogen, soil water content, soil organic matter and plant biomass evident from the effort to identify corn and soybean crops at various growth stages using high spatial resolution (1-m) data. Maximum distinction between corn and soybeans was achieved using the near-infrared bands when the crops were mature, while the visible bands were more useful when the soybeans were senescing [22]. High spatial (≤1 m) but low spectral resolution remote sensing data were used for mapping with accuracies > 95% of Chinese tallow trees in dominant environments found in coastal and adjacent upland landscapes. Airborne colour-infrared photography (CIR) (1:12,000 scale) was used to map localised occurrences of the widespread and aggressive Chinese tallow, an invasive species [23]. Mapping multiple vegetation types at large scale, determining appropriate plot size and spatial resolution is very difficult because of spectral mixtures, low correlation of remote sensing and field data, and high cost to collect field data at a high density. The within-support and regional semi-variograms  modelling technique was adopted based on field data and geostatistics theory accounting simultaneously for within support and regional spatial variability. The range parameters of the within-support semi-variograms implied the maximum range of the appropriate plot sizes. Using the regional semi-variograms the support size was considered appropriate when the ratio of the nugget variance to sill variance stabilised [24]. The distribution of Melaleuca quinquenervia, an aggressive exotic species targeted for eradication at a site in the East Everglades was mapped using 1:7,000-scale colour-infrared (CIR) aerial photographs,  GPS and GIS, with an overall accuracy of 94% [25]. A pre field forest cover map was generated for two seasons based on field knowledge, experience, and the standard visual interpretation of Landsat TM data with an overall accuracy of 91%. The forest cover type boundaries were delineated based on various elements of interpretation.  Accuracy assessment was done using bivariate matrix between map and ground observation [26]. The hybrid approach using Decision Tree (DT) and Adaptive Resonance Theory MAP (ARTMAP) neural network with confidence or uncertainty information via majority voting and other rules seems suitable to tackle a variety of classification problems in remote sensing and may ultimately aid map users in making more informed decisions. The classifier was tested with Advanced Very High Resolution Radiometer (AVHRR) data to mapping land cover of North America. [27]. Knowledge-based, post-classification processing considerably improves the accuracy of mapping evident from the incorporation of spatial knowledge with spectral knowledge of mangroves in the interpretation of SPOT data that has enhanced the accuracy level to 96.7% from 83.3%. Earlier parametric classification approaches had failed due to the spectral similarity of mangroves to other coastal vegetation despite their habitat being inside coastal waters [28].

Remotely sensed data, with synoptic coverage, allows for a means to collect data over an entire landscape. Its applications in the estimation of biomass for large-area forest inventory at the stand level have been successfully reported using remote sensing data sources [29,30,31]. The relative error of volume estimates using visual aerial photo interpretation has been reported to be 14–45% and the corresponding error using satellite image interpretation has been reported between 20-65% [32, 33, 34]. The approach presented in this paper has been developed to aid in the generation of biomass values from available or currently collected data.