Significance

This study shows multiangle imaging can be used to map forest aboveground biomass (AGB) in dryland regions with good precision and accuracy with respect to reference estimates from NASA’s Carbon Monitoring System program based on extensive airborne lidar surveys. The multiangle-derived AGB has almost the same accuracy as NASA’s GEDI lidar flying on ISS.

Chopping, M., Wang, Z., Schaaf, C.B., and Bull, M. (2026), Forest biomass in the southwestern U.S. from MISR and GEDI: assessment with NASA Carbon Monitoring System data, Remote Sensing of Environment, 333, 115117, 15 January 2026, .

Related recent presentation: Chopping, M., Wang, Z., Schaaf, C.B., and Bull, M. (2025), Forest Biomass in the Southwestern U.S. from MISR and GEDI: Assessment with NASA Carbon Monitoring System Data, American Geophysical Union Fall Meeting, Session B34C: Spatiotemporal Dynamics of Forest Disturbance and Recovery II Oral, abstract ID: 1989798, December 17, 2025, .

Abstract

Forest aboveground biomass (AGB) density mapping initiatives use one of three remote sensing approaches: lidar, radar, or near-nadir multispectral imaging leveraging machine learning methods, or a combination thereof. However, the active instrument record is limited and near-nadir multispectral imaging data are relatively insensitive to canopy physical structure. Multiangle imaging enables annual wall-to-wall mapping with a global record that extends back to 2000 as these data are highly sensitive to forest AGB. This paper describes work to validate estimates in a published annual, wall-to-wall record of forest AGB on a 250 m grid, derived using 672 nm imagery from the NASA, Jet Propulsion Laboratory’s Multiangle Imaging Spectro-Radiometer (MISR) for 2000–2021, covering the southwestern United States. Estimates in the published MISR-derived annual forest AGB map series for the southwestern United States and the Global Ecosystem Dynamics Investigation (GEDI) L4B Gridded 1 km AGB product were both found to be highly consistent with NASA Carbon Monitoring System (CMS) airborne lidar survey (ALS) AGB data. MISR and GEDI v.2 (v.2.1) estimates yielded similar coefficients of determination (~0.7) and Root Mean Square Error (RMSE) (~60 Mg ha−1) for all ALS data used. For the large CMS Sonoma County Improved AGB dataset, MISR and GEDI v.2 (v.2.1) estimates yielded R2 = 0.88, 0.88 (0.91); RMSE = 58, 40 (37) Mg ha−1. Estimates from MISR thus have an accuracy similar to that of the GEDI L4B gridded AGB product, with some limitations (e.g., topographic shading, tall, dense canopies). However the published MISR maps are on a 250 m grid, wall-to-wall, and cover the period 2000 – 2021. These results suggest MISR is able to provide a means to investigate trajectories of forest AGB change in the southwestern U.S. from 2000 onwards –over a substantial period of accelerating environmental and human- and climate-driven change– with reasonable precision.

View Presentation Slide

However, the climate is also drying. Warming and drying, together with decades of fire suppression and widespread bark beetle outbreaks from the mid-2000s, have made these forests increasingly prone to wildfires – and those fires are now burning hotter and more destructively. Pick any month and year in recent history and check the U.S. Drought Monitor and you will see that the monthly maps are not an aberration: the western US is drying up. Warmer springs mean the mountain snowpack that used to keep the forests moist until early summer now melt away in spring. Drier conditions mean that fires now burn hotter and tend to be “stand destroyers”: instead of clearing out the undergrowth and leaving the established trees alive, the forest is now burned to a crisp. The trees are killed and soils are seared, making it harder for saplings to take root. This does not bode well for the future of these forests.

Why do we care? Well, interactions between people, forests, and the climate do not always result in slow changes: if tipping points are reached, the system can flip suddenly to a different stable state, one that is far less friendly to people; and what happens in this region has impacts much further afield. For example, in recent years, smoke from western wildfires has started to impact the air in our region; see this July 2021 MSU news article. We thus have to monitor at scale: we need to zoom out and see the big picture – and this is what remote sensing from orbiting satellites allows. Scientists and forest managers also want to know how much carbon is lost (or gained) in forests, but losses cannot be measured after the forest has burned, or died from bark beetle infestations or disease, or been harvested (collectively: “disturbance”).

You might think that we could use the kind of satellite images seen in Google Earth to do this, or perhaps imagery from the long-running Landsat series of multispectral sensors. Both kinds of imagers provide pictures of the surface using sunlight reflected at several wavelengths, including in the blue, green, red, and near-infrared. However, clouds and infrequent viewing mean that, until recently, Google Earth imagery does not provide wall-to-wall coverage; and the imagery provided by both kinds of sensors is very sensitive to the condition of leaves: in a wet year, leaves will be more green, making it difficult to estimate forest carbon storage from year to year reliably. Recently we have started to use radar (radio detection and ranging) and lidar (light detection and ranging) to map and monitor forest carbon stocks. These advanced, active sensing technologies are excellent for this application; however, their records to date are patchy and too short.

There is another way to estimate forest carbon: exploit observations of reflected sunlight made at different angles relative to the sun direction. Because 3-D objects like trees cast shadows and scatter light in different directions, their presence, size, and spacing affects satellite readings made at different, oblique viewing angles, so we can use the difference to infer surface properties. The Google Earth and Landsat sensors can’t help here: they view almost straight down, towards the center of the Earth. However, in 2000 NASA launched a satellite called Terra that carries five instruments, including the Multi-angle Imaging Spectro-Radiometer (MISR). MISR was developed by the Jet Propulsion Laboratory that you may associate with the Mars rover missions. The MISR acronym is a typical JPL riff: the “miserly” satellite imager has only four spectral bands, vs the 224 of JPL’s “greedy” AVIRIS (Advanced Visible Infra-Red Imaging Spectrometer) airborne imager (that is clearly avaricious!). Here’s a picture showing how the nine MISR multispectral cameras view in front of (forward), straight down (nadir), and behind (aftward of) the satellite.

MISR was designed to measure properties of the atmosphere such as aerosols, clouds, and winds – it gathers measurements from its 9 cameras in about 7 minutes, before air masses move – but its observations can also be used to estimate ice sheet surface roughness and forest woody aboveground biomass (AGB; that is, the mass of wood in the aboveground parts of trees that is about 50% carbon). Tests showed that combining measurements of reflected red light (the kind that plants like to use for photosynthesis) from three of MISR’s cameras in a ratio provides a metric – a multi-angle index – that is strongly related to forest biomass. If we try to use the more traditional remote sensing approach of combining single-angle red and near-infrared (NIR) measurements into a multi-spectral index, such as the Normalized Difference Vegetation Index (NDVI), or use those measurements separately, estimates of AGB vs reference data are likely to be quite poor, with R2 < 0.4, vs 0.9 for the multi-angle index (see plots).

What about those pesky clouds? MISR can deal with clouds because it acquires imagery over a ~400 km wide swath, providing repeat viewing of any location within 8 days. This is sufficiently frequent for the construction of cloud-free composites.

AGB maps made using the multi-angle index match those available from satellite radar (2000) and lidar (2005) very well (right), allowing us to create annual time series from 2000 onwards. MISR AGB estimates were also consistent with highly reliable NASA Carbon Monitoring System airborne lidar-derived estimates for the Rim Fire area in California. The distribution of forest losses from fire and beetle disturbance in 2015 over 2000 also matched historical data in published sources.

These maps allow the quantification of AGB losses from fire, beetle, harvest, and disease at a time when climate warming and drying is increasingly stressing forests across the southwest. The subset at right is derived from the map series and shows the net change in forest AGB density (units: Mg ha-1) for part of northern California in 2015 over 2000. The method and results for 2000 – 2015 are reported in (Impact Factor: 10.2). The dataset has also been extended to 2021 and published at the .

We hope these map series will help ecologists and foresters to improve our understanding of the changes we are seeing in southwestern forests; and how the progression of these changes might affect people living in this region in the coming decades as the regional climate warms and dries.

Citation: Chopping, M., Wang, Z., Schaaf, C., Bull, M.A., and Duchesne, R.R. (2022), Forest aboveground biomass in the southwestern United States from a MISR multi-angle index, 2000–2015, Remote Sensing of Environment, 275, 112964, ISSN 0034-4257, .

My research is focusing on the effects of climate change on tall shrubs in the Alaskan and Canadian regions by combining Dr. Mark Chopping’s CANAPI, CANAPAMI and CANAPANI tall shrub detection software with high resolution satellite imagery from the National Geospatial-Intelligence Agency as part of NASA’s Arctic-Boreal Vulnerability Experiment (ABoVE). The research intends to evaluate the changes in shrubs cover and height over a period of 10 – 20 years. The results of this study can be used to determine the impact on summer albedo — an important factor affecting the region’s hydrology, microbial respiration, and rates of permafrost thaw — and will thus be useful in ecological models, and to validate lower spatial resolution ABoVE remote sensing data products.

During his stay, Bao Gang will advance research that provides the most recent information on dynamics in vegetation NPP and its responses to climate variables in arid and semiarid ecosystems. This is important as the satellite record lengthens and becomes more informative on decadal time scales. It will help us to understand what recent climate-driven vegetation changes have occurred in the arid and semiarid areas of Eurasia, and why. Climate-driven changes in the geographic distribution of precipitation are of particular concern in this region – as in California and the western US, though both climatic and human drivers differ – and important differences between Eurasian and North American NPP trends have been observed that remain to be understood.

Bao Gang’s selection of the Department of Earth and Environmental Studies at �ǿ����޴�ý University was owing to the recognized expertise in satellite-based vegetation mapping, with NASA-supported projects on community type mapping and tracking carbon pools in semi-arid environments. Mark Chopping has long experience in Inner Mongolia Autonomous Region, having carried out fieldwork in the Xilingol grasslands in 1993, 1996, and 2012, including surveys for his doctoral research. During the 2012 trip meetings were held in the IMNU Remote Sensing Laboratory that led to the discovery of research topics of mutual interest. His stay at �ǿ����޴�ý University is fully supported by the Chinese Scholarship Council, the Chinese Ministry of Education’s non-profit organization that provides student financial aid to Chinese citizens and foreigners to study abroad or to study in China, respectively. It is hoped that this project will help to establish a solid and ongoing collaboration between researchers and institutions in this area, as well as enriching the experiences of �ǿ����޴�ý University students who take remote sensing classes, providing them with broader perspectives.

Phillip is studying Geography and Government, with a focus in GIS. He brings experience in ArcGIS, QGIS, Computer Aided Design, and Python and also has outstanding achievements in Track & Field. He is assisting with the evaluation of results from over 600 CANAPI runs on large images, with millions of trees mapped. The results will be submitted in an upcoming report to NASA; they will also shed light on techniques to be used in a new NASA-funded project at MSU: "Changes in Shrub Abundance in Arctic Tundra from the Satellite High Resolution Record for the Arctic-Boreal Vulnerability Experiment and Impacts on Albedo”, part of NASA’s ABoVE project (http://above.nasa.gov). This research is supported by NASA’s Carbon Cycle Science and Land Cover Land Use Change programs.