CLASlite Projects

Satellite data obtained through remote sensing has been a common and widespread source of information for forest monitoring. This is, in part, due to its ability to cover large tracts of forest lands, and its accessibility, availability, and cost. However, its spatial resolution can be coarse and may not be adequate to meet certain level of accuracy and uncertainty. Airborne-generated data sets have higher spatial resolutions and better accuracy, but can be very costly to obtain. To monitor large forest areas for various purposes, including forest carbon assessment under the Reduced Emission from Deforestation and forest Degradation (REDD+) program of the UN Framework Convention for Climate Change (UNFCCC), it is imperative to develop tools and methodologies that are cost efficient and meets acceptable level of accuracy. This paper explores different tools and methodologies that combines data obtained from medium resolution imagery obtained from remote sensing and high resolution data sets obtained from airborne sensors such as LiDAR and radar. These tools will be applied and tested using data obtained from pilot REDD+ areas.

  • USAID B+WISER Program and Department of Natural Resources, University of Illinois, USA
  • USAID Office of Environment, Energy and Climate Change, Manila
  • Fauna & Flora International Philippines, Philippines

Makira Natural Protected Area in northeastern Madagascar houses high levels of biodiversity, but is currently threated by encroachment of agriculture and the illegal logging of hardwoods. The Wildlife Conservation Society currently works with local communities in running a Reducing Emissions from Deforestation and Forest Degradation (REDD+) project in Makira. High cloud cover in humid forests and seasonal variations in vegetation make temporally consistent classifications difficult, affecting the estimates of deforestation needed for REDD+. Carnegie Landsat Analysis System lite, or CLASlite, is a free software tool designed to pre-process and analyze remotely sensed data, creating forest and deforestation maps, between other outputs. This study was performed in collaboration with WCS using fifteen Landsat 5, 7, and 8 images to determine the performance of CLASlite in comparison to previous mapping work using a conventional mapping approach. A combination of CLASlite and expert interpretation created four forest vs. non-forest maps from 1994, 2001, 2010, and 2014 with 92% overall accuracy in the 26,956 km² study area. Over the twenty year period, 7.9% of the 11, 464 km² of original forest was lost. Visual comparisons with the previous mapping work revealed that CLASlite was better at detecting small patches of forest as well as connectivity in riparian areas. Various cloud masking settings within CLASlite allowed for adjustments specific to cloud and sensor types. Cloud contaminations were minimized by utilizing Landsat tiles from the dry season. CLASlite is well designed for conservation practitioners and performed well in this forest type.

  • Clark University, United States
  • Wildlife Conservation Society, United States

The Carnegie Landsat Analysis System-lite (CLASlite) was used to map and monitor tropical forest change in two large tropical watersheds in Peru: Greater Marañón and Ucayali. CLASlite uses radiometric and atmospheric correction algorithms as well as an Automated Monte Carlo Unmixing (AutoMCU) to obtain consistent fractional land cover per-pixel at high spatial resolution. Fractional land cover is automatically extracted from universal spectral libraries which allow for a differentiation between live photosynthetic vegetation (PV), non-photosynthetic vegetation (NPV) and bare substrate (S). Fractional cover information is directly translated to maps of forest cover based in the physical characteristics of the forest canopy. Rates of deforestation and disturbance are estimated through analysis of change in fractional land cover over time. The Greater Marañón and Ucayali watersheds were studied over the period 1985 to 2012, through analysis of 1900 multi-spectral images from Landsat 4, 5 and 7. These images were processed and analyzed using CLASlite to obtain fractional cover and forest cover information for each year within the period. Annualization of the collected maps provided detailed information on the gross rates of disturbance and deforestation throughout the region. Further, net deforestation and disturbance maps were used to show the general forest change in these watersheds over the past 25 years. We found that deforestation accounts for just ~50% of the total forest losses, and that forest disturbance (degradation) is critically important to consider when making forest change estimates associated with losses in habitat and carbon in the region. These results also provide spatially-detailed, temporally-specific information on forest change for nearly three decades. Information provided by this study will assist decision-makers in Peru to improve their regional environmental management. The results, unprecedented in spatial and temporal scope, are another example showing the fidelity of tropical deforestation and forest degradation monitoring made routine using the CLASlite system.

  • Global Ecology, Carnegie Institution of Science, USA

Recently, several remote sensing methods have been developed to quantify the degradation of tropical forests. However, it still lacks finest spatial and temporal analysis to define trajectories of forest degradation i.e. a temporal analysis of the impacts on forest integrity. This communication aims to explore this issue and proposes a set of operational indicators to monitor forest degradation, which can constitutes a decision tool to support forestry managers and policy makers. We studied the trajectories of forest degradation in the municipality of Paragominas - PA in the eastern Brazilian Amazon between 1995 and 2009, with a focus on the forestry company Cikel (400 000 ha certified by FSC since 2001). First, we developed a semi-automatic remote sensing methodology to detect forest degradation using multi-temporal Landsat images (spatial resolution of 30m) covering the 1995-2009 period. This method included two steps: 1) Identification of logging tracks and log landings using an algorithm of Bourbier et al. (2013). This algorithm uses spectral indices and morphological filters to strengthen the spectral contrasts between bare soil and forest cover. 2) Identification of logging gaps - which are characterised by senescent vegetation due to trees fall - using a Spectral Mixture Analysis carried out in CLASlite (Asner et al., 2009) and a fraction index (Souza et al., 2013). So, we obtained annual maps identifying these three major impacts. Secondly, we calculated annual landscape metrics of forest degradation using the R package "SpatialEco". Then, we calculated indicators which synthetize information about logging impacts and logging frequencies over the period from these annual degradation metrics. Finally, we selected a set of 6 indicators and statistically analysed the trajectories of degradation occurring in Paragominas using ACP and CAH. Our results emphasize four major degradation trajectories from well managed forests to highly-logged forests. They clearly show a difference between legal and illegal logging in terms of forest degradation. Moreover, they indicate that impacts of FSC certification on forest degradation was positive. Degradation was statistically lower in the certified logged plots compared to the uncertified plots. These set of indicators are adequate to monitor forest degradation through space and provide guidance to policy-makers for a better management of forest resources.

  • Tritsch Isabelle, CIRAD-ES-UPR BSe, France
  • Blanc Lilian, CIRAD-ES-UPR BSef, France
  • Gond Valéry, CIRAD-ES-UPR BSef, France
  • Bourgoin Clément, France
  • Cornu Guillaume, CIRAD-ES-UPR BSef, France
  • Sist Plinio, CIRAD-ES-UPR BSef, France

The anthropogenic use of natural resources has become a major cause ofbiodiversity loss and habitat degradation throughout the world. Deforestation - theconversion of forests to alternative land covers - has led to a decrease in localbiodiversity directly through a decrease in habitat, and indirectly through habitatfragmentation. Likewise, defaunation - the loss of animals both directly through huntingand indirectly through deforestation - has led to the empty forest syndrome andsubsequent deterioration of forest ecosystems. In many cases, areas where anthropogenicuse of natural resources is high overlap with areas of high biodiversity value. Therefore,the present series of studies aims to better understand the impacts that different types ofnatural resources use and habitat degradation have on biodiversity. This dissertationdetails the results of five studies, which aimed to: 1) examine the effects of habitatdegradation on plant-frugivore networks; 2), understand the live capture and extent ofownership of lemurs in Madagascar; 3) understand the micro- and macro-level drivers ofwild meat consumption in Madagascar; 4) describe the capture, movement, and trade ofwild meat in Madagascar; and 5) the impacts of habitat changes on the diets and verticalstratification of frugivorous bats.

  • Temple University, Pennsylvania

Monitoring Asia’s changing forests in a consistent, repeatable manner is of great importance to understanding the carbon balance. As part of the United States Agency for International Development (USAID) Lowering Emissions in Asia’s Forests (LEAF) program’s effort to facilitate the development of practical monitoring systems, this study evaluated three automated forest loss detection algorithms in the four core USAID LEAF study areas using Landsat data spanning 2001 to 2013.

The evaluated algorithms were CLASlite, Hansen’s Global Forest Loss product, and a new algorithm called Multiple Linear Trend Analysis (MLTA). Google Earth Engine (GEE) was used throughout this project as an efficient cloud-based computation platform. GEE provides massively parallel computing infrastructure available to many areas with limited information technology infrastructure. CLASlite is freely available, but rather difficult to properly calibrate to the changing forest dynamics of the four study areas. Hansen’s loss product is readily available, spatially explicit, and updated annually, but lacks any ability to refine the definition of deforestation. Additionally, it has no separate forest degradation category. MLTA can be customized to meet specific definitions of deforestation and forest degradation but proved difficult to properly calibrate to characterize deforestation and forest degradation without sufficient on-the-ground knowledge.

Results indicate no clear trends of annual forest loss rates throughout the four study areas from 2001 to 2013. Also, there is a large variance in the amount of forest loss detected by each algorithm. A quantitative accuracy assessment was conducted using the Timesync Landsat visualization tool across a total of 2,000 30- by 30-m sample pixels. Results indicate that the Hansen product, while only identifying forest cover loss, overlaps with much of what the accuracy assessment characterized as forest degradation and degradation, thus combining the two in a single class. The CLASlite products generally had the lowest accuracies. The MLTA product had high accuracies in some areas, which indicated that with better calibration the method could potentially meet monitoring needs.

Remote sensing-based methods have the potential to provide practical automated estimates of forest change in Asia. Currently, methods are being actively developed to meet these growing needs. Results from this study indicate that currently available methods may be sufficient for first-order estimates of deforestation and degradation, but further refinement may be necessary for more precise needs.

  • RedCastle Resources at the Remote Sensing Applications Center (RSAC) in Salt Lake City, Utah.
  • Chulalongkorn University in Bangkok Thailand
  • USAID LEAF Bangkok, Thailand.

CLASlite Course support: (

CLASlite Technical support: (

Carnegie Institution for Science

260 Panama Street • Stanford, CA 94305, USA