Stephen Hagen

Baseline map of carbon emissions from deforestation in tropical regions.

Tropical Carbon Loss Accurate and precise measures of tropical deforestation and the resulting carbon emissions are needed in order to formulate climate policy. Harris et al. (p. 1573; see the Perspective by Zarin) used satellite observations of deforestation within the tropics of three continents to estimate that gross annual carbon emissions were approximately 0.8 Pg Cyr−2 (Pg = 1015 g) for the years 2000 to 2005, from the loss of 43 million hectares of forest. This result, which is about one-third of some previous estimates, should serve as a baseline for future assessments of changes in the rate of loss of tropical forests. Tropical deforestation and degradation across three continents led to ~0.8 petagrams of yearly carbon emissions from 2000 to 2005. Policies to reduce emissions from deforestation would benefit from clearly derived, spatially explicit, statistically bounded estimates of carbon emissions. Existing efforts derive carbon impacts of land-use change using broad assumptions, unreliable data, or both. We improve on this approach using satellite observations of gross forest cover loss and a map of forest carbon stocks to estimate gross carbon emissions across tropical regions between 2000 and 2005 as 0.81 petagram of carbon per year, with a 90% prediction interval of 0.57 to 1.22 petagrams of carbon per year. This estimate is 25 to 50% of recently published estimates. By systematically matching areas of forest loss with their carbon stocks before clearing, these results serve as a more accurate benchmark for monitoring global progress on reducing emissions from deforestation.

Data association methods with applications to law enforcement

Associating records in a large database that are related but not exact matches has importance in a variety of applications. In law enforcement, this task enables crime analysts to associate incidents possibly resulting from the same individual or group of individuals. In practice, most crime analysts perform this task manually by searching through incident reports looking for similarities. This paper describes automated approaches to data association. We report tests showing that our data association methods significantly reduced the time required by manual methods with accuracy comparable to experienced crime analysts. In comparison to analysis using the structured query language (SQL), our methods were both faster and more accurate.

Mapping inland lake water quality across the Lower Peninsula of Michigan using Landsat TM imagery

The number, size, and distribution of inland freshwater lakes present a challenge for traditional water-quality assessment due to the time, cost, and logistical constraints of field sampling and laboratory analyses. To overcome this challenge, Landsat imagery has been used as an effective tool to assess basic water-quality indicators, such as Secchi depth (SD), over a large region or to map more advanced lake attributes, such as cyanobacteria, for a single waterbody. The overarching objective of this research application was to evaluate Landsat Thematic Mapper (TM) for mapping nine water-quality metrics over a large region and to identify hot spots of potential risk. The second objective was to evaluate the addition of landscape pattern metrics to test potential improvements in mapping lake attributes and to understand drivers of lake water quality in this region. Field-level in situ water-quality measurements were collected across diverse lakes (n = 42) within the Lower Peninsula of Michigan. A multicriteria statistical approach was executed to map lake water quality that considered variable importance, model complexity, and uncertainty. Overall, band ratio radiance models performed well (R2 = 0.65–0.81) for mapping SD, chlorophyll-a, green biovolume, total phosphorus (TP), and total nitrogen (TN) with weaker (R2 = 0.37) ability to map total suspended solids (TSS) and cyanobacteria levels. In this application, Landsat TM and pattern metrics showed poor ability to accurately map non-purgable organic carbon (NPOC) and diatom biovolume, likely due to a combination of gaps in temporal overpass and field sampling and lack of signal sensitivity within broad spectral channels of Landsat TM. The composition and configuration of croplands, urban, and wetland patches across the landscape were found to be moderate predictors of lake water quality that can complement lake remote-sensing data. Of the 4071 lakes, over 4 ha in the Lower Peninsula, approximately two-thirds, were identified as mesotrophic (n = 2715). This application highlights how an operational tool might support lake decision-making or assessment protocols to identify hot spots of potential risk.

Monitoring Rice Agriculture in the Sacramento Valley, USA With Multitemporal PALSAR and MODIS Imagery

Rice agriculture is an important crop that influences land-atmosphere interactions and requires substantial resources for flood management. Multitemporal acquisition strategies provide an opportunity to improve rice mapping and monitoring of hydroperiod. The objectives of this study were to 1) delineate rice paddies with Phased Array L-band Synthetic Aperture Radar (PALSAR) fine-beam single/dual (FBS/D) mode measurements and 2) integrate multitemporal, ScanSAR Wide-Beam 1 (WB1)- and Moderate Resolution Imaging Spectroradiometer (MODIS)- observations for flood frequency mapping. Multitemporal and multiscale PALSAR and MODIS imagery were collected over the study region in the Sacramento Valley, California, USA. A decision-tree approach utilized multitemporal FBS (HH polarization) data to classify rice fields and WB1 measurements to assess paddy flood status. High temporal frequency MODIS products further characterized hydroperiod for each individual rice paddy using a relationship between the Enhanced Vegetation Index (EVI) and the Land Surface Water Index (LSWI). Validation found the PALSAR-derived rice paddy extent maps and hydroperiod products to possess very high overall accuracies (95% overall accuracy). Agreement between MODIS and PALSAR flood products was strong with agreement between 85-94% at four comparison dates. By using complementing products and the strengths of each instrument, image acquisition strategies and monitoring protocol can be enhanced. The results highlight how the integration of multitemporal PALSAR and MODIS can be used to generate valuable agro-ecological information products in an operational context.

Mapping Total Vegetation Cover Across Western Rangelands With Moderate-Resolution Imaging Spectroradiometer Data

Abstract Remotely sensed observations of rangelands provide a synoptic view of vegetation condition unavailable from other means. Multiple satellite platforms in operation today (e.g. Landsat, moderate-resolution imaging spectroradiometer [MODIS]) offer opportunities for regional monitoring of rangelands. However, the spatial and temporal variability of rangelands pose challenges to consistent and accurate mapping of vegetation condition. For instance, soil properties can have a large impact on the reflectance registered at the satellite sensor. Additionally, senescent vegetation, which is often abundant on rangeland, is dynamic and its physical and photochemical properties can change rapidly along with moisture availability. Remote sensing has been successfully used to map local rangeland conditions. However, regional and frequently updated maps of vegetation cover in rangelands are not currently available. In this research, we compare ground measurements of total vegetation cover, including both green and senescent cover, to reflectance observed by the satellite and develop a robust method for estimating total vegetation canopy cover over diverse regions of the western United States. We test the effects of scaling from ground observations up to the Landsat 30-m scale, then to the MODIS 500-m scale, and quantify sources of noise. The soil-adjusted total vegetation index (SATVI) captures 55% of the variability in ground measured total vegetation cover from diverse sites in New Mexico, Arizona, Wyoming, and Nevada. Scaling from the Landsat to MODIS scale introduces noise and loss of spatial detail, but offers inexpensive and frequent observations and the ability to track trends in cover over large regions. Resumen Observaciones de pastizales con sensores remotos proporcionan una vista sinóptica de la condición de la vegetación que no está disponible usando otros medios. Múltiples plataformas satelitales en operación hoy en día (e.g. Landsat, MODIS) proporcionan oportunidades para un monitoreo regional de los pastizales. Sin embargo, la variabilidad espacial y temporal de los pastizales posee retos relacionados con el mapeo de la condición de la vegetación. Por ejemplo, las propiedades del suelo pueden tener gran impacto en la reflectancia registrada por el sensor del satélite. Adicionalmente, la vegetación senescente, la cual es a menudo abundante en los pastizales, es dinámica y sus propiedades físicas y fotoquímicas pueden cambiar rápidamente debido al contenido de humedad disponible. Los sensores remotos han sido utilizados con éxito para mapear las condiciones locales de los pastizales. Sin embargo, mapas regionales y frecuentemente actualizados de la cobertura de la vegetación en pastizales no están disponibles en la actualidad. En esta investigación, se compararon medidas del suelo del total de la cobertura, incluyendo ambas coberturas la verde y la senescente, contra la observada por el satélite para desarrollar un método robusto con la finalidad de estimar el total de la cobertura de la copa de la vegetación sobre la diversa región del Oeste de estado Unidos. Se evaluaron los efectos de escala desde observaciones al ras de suelo hasta aquellas usando Landsat a una escala de 30 m, entonces a la escala de 500 m en MODIS y se cuantificaron las fuentes de variación. El índice ajustado total de vegetación (SATV) captura 55% de la variabilidad en la estimación del total de la cobertura vegetal de diversos sitios en Nuevo México, Arizona, Wyoming, y Nevada. La conversión de escala de Landsat a MODIS introduce cierto margen de error y pérdida de detalle espacial, pero ofrece observaciones baratas y frecuentes así como la capacidad de rastrear las tendencias en cobertura sobre extensas regiones.

Detection of Large-Scale Forest Canopy Change in Pan-Tropical Humid Forests 2000–2009 With the SeaWinds Ku-Band Scatterometer

We analyzed the 10-year record (1999-2009) of SeaWinds Ku-band microwave backscatter from humid tropical forest regions in South America, Africa, and Indonesia/Malaysia. While backscatter was relatively stable across much of the region, it declined by 1-2 dB in areas of known large-scale deforestation, and increased by up to 1-2 dB in areas of secondary forest or plantation forest growth and in major metropolitan areas. The reduction in backscatter over 142 18.5 km × 18.5 km blocks of tropical forest was correlated with gross forest cover loss (as determined from Landsat data analysis) (R = -0.78); this correlation improved when restricted to humid tropical forest blocks in South America with high initial forest cover (R = -0.93, n = 22). This study shows that scatterometer-based analyses can provide an important geophysical data record leading to robust identification of the spatial patterns and timing of large-scale change in tropical forests. The coarse spatial resolution of SeaWinds ( ~ 10 km) makes it unsuitable for mapping deforestation at the scale of land-use activity. However, due to a combination of instrument stability, sensitivity to canopy change and insensitivity to atmospheric effects, and straight-forward data processing, Ku-band scatterometery can provide a fully independent assessment of large-scale tropical forest canopy dynamics which may complement the interpretation of higher resolution optical remote sensing.

High Resolution Mapping of Peatland Hydroperiod at a High-Latitude Swedish Mire

Monitoring high latitude wetlands is required to understand feedbacks between terrestrial carbon pools and climate change. Hydrological variability is a key factor driving biogeochemical processes in these ecosystems and effective assessment tools are critical for accurate characterization of surface hydrology, soil moisture, and water table fluctuations. Operational satellite platforms provide opportunities to systematically monitor hydrological variability in high latitude wetlands. The objective of this research application was to integrate high temporal frequency Synthetic Aperture Radar (SAR) and high spatial resolution Light Detection and Ranging (LiDAR) observations to assess hydroperiod at a mire in northern Sweden. Geostatistical and polarimetric (PLR) techniques were applied to determine spatial structure of the wetland and imagery at respective scales (0.5 m to 25 m). Variogram, spatial regression, and decomposition approaches characterized the sensitivity of the two platforms (SAR and LiDAR) to wetland hydrogeomorphology, scattering mechanisms, and data interrelationships. A Classification and Regression Tree (CART), based on random forest, fused multi-mode (fine-beam single, dual, quad pol) Phased Array L-band Synthetic Aperture Radar (PALSAR) and LiDAR-derived elevation to effectively map hydroperiod attributes at the Swedish mire across an aggregated warm season (May-September, 2006-2010). Image derived estimates of water and peat moisture were sensitive (R-2 = 0.86) to field measurements of water table depth (cm). Peat areas that are underlain by permafrost were observed as areas with fluctuating soil moisture and water table changes.

Monitoring the brazilian pasturelands: A new mapping approach based on the landsat 8 spectral and temporal domains

Abstract In a world marked by a rapid population expansion and an unprecedented increase in per capita income and consumption, sustainable food production is certainly the most pressing issue affecting mankind. Within this context, the brazilian pasturelands, the main land-use form in the country, constitute a particularly important asset as a land reserve, which, through improved land-use strategies and intensification, can meet food security goals and contribute to the mitigation of greenhouse gas emissions. In this study, we utilized the entire set of Landsat 8 images available for Brazil in 2015, from which dozens of seasonal metrics were derived, to produce, through objective criteria and automated classification strategies, a new pasture map for the country. Based on the Random Forest algorithm, individually modelled and applied to each one of the 380 Landsat scenes covering the Brazilian territory, our map showed an overall accuracy of 87%. Another result of this study was the thorough spatial and temporal assessment of Landsat 8 data availability in Brazil, which indicated that about 80% of the country had 12 or fewer observations free of clouds or cloud shadows in 2015.