Alexander J. Hernandez

Land cover dynamics monitoring with Landsat data in Kunming, China: a cost-effective sampling and modelling scheme using Google Earth imagery and random forests

Changes in forest composition impact ecological services, and are considered important factors driving global climate change. A hybrid sampling method along with a modelling approach to map current and past land cover in Kunming, China is reported. MODIS land cover (2001–2011) data-sets were used to detect pixels with no apparent change. Around 3000 ‘no change points’ were systematically selected and sampled using Google Earth’s high-resolution imagery. Thirty-five per cent of these points were verified and used for training and validation. We used Random forests to classify multi-temporal Landsat imagery. Results show that forest cover has had a net decrease of 14385 ha (1.3% of forest area), which was primary converted to shrublands (11%), urban and barren land (2.7%) and agriculture (2.5%). Our validation indicates an overall accuracy (Kappa) of 82%. Our methodology can be used to consistently map the dynamics of land cover change in similar areas with minimum costs.

A Landscape Similarity Index: Multitemporal Remote Sensing to Track Changes in Big Sagebrush Ecological Sites

Abstract A similarity index for big sagebrush ecological sites was developed in northern Utah. In contrast to field measurements used to calculate similarity to reference states, our approach relies on the utilization of historic archives of satellite imagery to measure the ecological distance to benchmarks of undesired conditions such as invasion by exotic annuals and woodland encroachment. Our benchmarks consisted of locations for which there are field data collected for monitoring and evaluation purposes for several time periods. We utilized a temporal series of Landsat thematic mapper (TM) imagery that spanned 1984 to 2008 from which the soil-adjusted vegetation index (SAVI) and other transformations were extracted. Topographic and climatic variables were also included as ancillary data. Multidimensional scaling (MDS) was used to obtain scores in reduced ordination space for two periods of interest: 1984–1996 and 1997–2008. Interannual SAVI mean-variance plots provided evidence that the benchmarks and ecological sites have a distinct temporal response that allows an objective comparison. Our MDS results also show that natural clusters can be identified in the reduced statistical space for ecological sites that are a dominant component of a soil map unit. The two MDS solutions allowed the ordination of ecological sites in two gradients of productivity and bare ground. Interpretations of the transitions and trajectories of mountain, Wyoming, and basin big sagebrush sites correlated well with the ecological expectation. We anticipate that range conservationists and others actively working in rangeland evaluation can use this application to develop and update ecological site descriptions and state-and-transition models from a remotely sensed perspective.

Forest sector carbon analyses support land management planning and projects: assessing the influence of anthropogenic and natural factors

Management of forest carbon stocks on public lands is critical to maintaining or enhancing carbon dioxide removal from the atmosphere. Acknowledging this, an array of federal regulations and policies have emerged that requires US National Forests to report baseline carbon stocks and changes due to disturbance and management and assess how management activities and forest plans affect carbon stocks. To address these requirements with the best-available science, we compiled empirical and remotely sensed data covering the National Forests (one fifth of the area of US forest land) and analyzed this information using a carbon modeling framework. We demonstrate how integration of various data and models provides a comprehensive evaluation of key drivers of observed carbon trends, for individual National Forests. The models in this framework complement each other with different strengths: the Carbon Calculation Tool uses inventory data to report baseline carbon stocks; the Forest Carbon Management Framework integrates inventory data, disturbance histories, and growth and yield trajectories to report relative effects of disturbances on carbon stocks; and the Integrated Terrestrial Ecosystem Carbon Model incorporates disturbance, climate, and atmospheric data to determine their relative impacts on forest carbon accumulation and loss. We report results for several National Forests across the USA and compare their carbon dynamics. Results show that recent disturbances are causing some forests to transition from carbon sinks to sources, particularly in the West. Meanwhile, elevated atmospheric carbon dioxide and nitrogen deposition are consistently increasing carbon stocks, partially offsetting declines due to disturbances and aging. Climate variability introduces concomitant interannual variability in net carbon uptake or release. Targeting forest disturbance and post-disturbance regrowth is critical to management objectives that involve maintaining or enhancing future carbon sequestration.

A comparison between cluster busting technique and a classification tree algorithm of a moderate resolution imaging spectrometer (MODIS) land cover map of Honduras

A national land cover map derived from moderate resolution imaging spectrometer (MODIS) imagery products was developed for Honduras, Central America. We compared two methods of image classification: a cluster busting (CB) classification technique and a classification and regression tree (CART) algorithm. Field data samples were used to validate the resulting classifications. Inthe classification process, we used: a Google Earth™ sampling scheme, a time series of MODISs Enhanced Vegetation Index (EVI) and digital elevation data(shuttle radar topography mission, SRTM). The CART classification method provided a more accurate classification (Kappa coefficient, K = 74%, overall model accuracy = 79.6%) while compared to the CB classification (Kappa coefficient, K = 9%, overall model accuracy = 25.1%). The findings are useful to design more accurate MODIS classification protocols in tropical countries.

Likelihood of occurrence of bark beetle attacks on conifer forests in Honduras under normal and climate change scenarios

Conifer forests cover approximately 27% of Honduras (∼3 million ha), and have been traditionally affected by bark beetle (Dendroctonus frontalis) outbreaks. These outbreaks impact ecosystem health and predispose more attacks. We developed a logistic model to assess the forests’ susceptibility to a beetle attack. Models were fitted using climatic, topographic and remote sensing variables. We show a method to generate pseudo-absences based on a multi-temporal wetness index. Our optimized threshold to convert the continuum of probabilities into 0 and 1 value was 0.65. Our overall accuracy was 68.7%. We also developed models that integrate climate change scenarios. Our predictions signal an increase in the overall susceptibility for an attack when including climate change scenarios. To the best of our knowledge this is the first effort to develop a spatially explicit model of the probability of beetle outbreaks in a Central American country.