Suggested Searches

Abstract

Author: Christopher S. Potter, Bioshpheric Science Branch (SGE), NASA Ames Research Center
Source: Journal of Biodiversity, Management, & Forestry [Full PDF]

Extreme drought from 2013 to 2015 has been linked to extensive tree dieback in the Sierra-Nevada region of California. Landsat satellite imagery was analysed for the region from Lake Tahoe to the southern Sequoia National Forest with the objective of understanding the patterns of tree mortality in the years of 2013 to 2015 and into the near-normal precipitation year of 2016. The main mapping results for Landsat moisture index differences from year-to-year showed that the highest coverage of tree dieback was located in the Sierra and Sequoia National Forests, at four to five times greater area each year than within any other National Park or National Forest unit. Since 2013, over 50% of the Sierra Nevada forest dieback area was detected in the mid elevation zone of 1000- 2000 m. The total area of tree mortality in the lower elevation zone of 500-1000 m did not grow notably from 2015 to 2016. Within the largest California river drainages in the Sierra region, new tree mortality in 2015 was detected mainly below 1200 m elevation, whereas new tree mortality in 2016 was detected mainly at higher elevations, up to about 2200 m. In three out of the four years studied, results showed that about 60% of all new tree mortality areas were located on north-facing hill slopes.
Keywords: Landsat; Forest; NDMI; Drought; Sierra Nevada; California
 

Methods

Satellite image processing: Near cloud-free imagery from the Landsat 8 sensor was selected from the year 2011 and every year from 2013 to 2016. Image data from Landsat path/rows 41/35, 42/34, 42/35, and 43/33 were acquired between July 15 and August 30 each year, around the peak of the Sierra summer growing season [30,31]. All images used in this study were geometrically registered (UTM Zone 11) using terrain correction algorithms (Level 1T) applied by the U. S. Geological Survey EROS Data Centre [32]. Landsat 8 surface reflectance products were generated from the L8SR algorithm [33] using a method that uses the scene centre for the sun angle calculation and then hard-codes the view zenith angle to 0. The solar zenith and view zenith angles are used for calculations as part of the atmospheric correction.

NDMI calculations: The severity of drought stress across the study region was determined using the Landsat NDMI [17], computed by the equation: NDMI = (NIR – SWIR) / (NIR + SWIR) where NIR is the near infrared (0.85 to 0.88 μm) band and SWIR is the short wave infrared (1.57 to 1.65 µm) band for the Landsat 8 sensor. NDMI (scaled 0 to 10000) is comparable to the Landsat normalized burn ratio (NBR) used for wildfire severity mapping [34]. The reduction of reflectance of the SWIR as compared to the NIR is due to the absorption of water in leaf tissues [35]. Hais et al. [36] reported that NDMI for conifer canopies decreased as leaf cover is lost and SWIR band values increase with exposure of soils under the stressed trees. High NDMI values represent relatively high vegetation canopy moisture and lower drought stress, while near-zero values would represent relatively low vegetation canopy moisture content and higher drought stress [23]. Potter [25] reported that a negative change of at least 2000 NDMI between two image dates corresponded closely to sites identified by USFS aerial surveys with high tree mortality in the southern Sierra region.
 

Other data sets used: Elevation, slope, and aspect at 1 arc-second resolution were determined from the United States Geological Survey (USGS) National Elevation Dataset (NED). Wildfire boundaries and years since fire (YSF) were compiled from the California Department of Forestry, Fire and Resource Assessment Program, with additions from the National Park Service. River drainage basin boundaries were delineated by the U. S. Geological Survey [37] and downloaded from the National Hydrography Database.
 

Data analysis: Following on the validation of methodology by Potter [25] in closely matching forest stand mortality from aircraft surveys with Landsat imagery, each Landsat NDMI data layer from the years 2013 to 2016 was subtracted individually from the Landsat NDMI data layer for the year 2011, which was the most recent non-drought year in the region with above average annual precipitation and snow pack levels [30]. The NDMI differences between years greater than 2000 units were labeled as areas of highest probable tree mortality (Potter, 2016) [25] and saved for further spatial analysis. All areas burned by wildfires from the years 2013, 2014, and 2015 were masked out of the resulting dNDMI > 2000 map layers.
To generate the dNDMI layer for the year 2014, locations of dNDMI > 2000 from the 2013 layer were masked out to prevent double-counting areas of probable tree mortality. The same masking out of previous year’s dNDMI layers was performed for the 2015 and 2016 dNDMI layers, such that only new areas of probable tree mortality were included in these years post-2013. 
Zonal statistics for the dNDMI>2000 layers from 2013 to 2016 were computed using river basin delineations, elevation zones, National Forest (NF) and National Park (NP) boundaries, and predominant aspect class (north or south) in the ArcGIS software extension. Areas of probable tree mortality were summed in units of hectares within each different topographic or administrative zone.

Results

Yearly patterns of tree mortality were mapped (for ease of viewing) into three sections of the Sierra Nevada study region, namely the Lake Tahoe Basin, Yosemite to Kings Canyon NPs, and Sequoia NF and NP (Figure 2a-c). Most of the tree mortality detected by the dNDMI method in the Lake Tahoe section occurred in 2013 and 2014. In contrast, the areas to the west and south of Yosemite NP and Kings Canyon NP showed extensive new tree dieback locations in 2015 and 2016. This included the western sections of the Sierra NF. In Sequoia NP and NF in the southern Sierra range, tree mortality locations were abundant at the lower elevations below 2000 m during all years from 2013 to 2016.
Areas of new tree mortality totaled by NF and NP administrative units for 2015 and 2016 showed that the highest coverage was in the Sierra and Sequoia NFs (Figure 3), at four to five times greater area than in the next highest unit, the Stanislaus NF. All three NPs, Yosemite, Kings Canyon, and Sequoia, each had less than 2000 ha of new tree mortality detected in either 2015 or 2016.
Areas totaled within major drainage basins showed that the Upper sections of the following California river basins, the San Joaquin, King, Kern, Merced, and Stanislaus, each with more than 20,000 ha of new tree mortality detected (Table 1). The San Joaquin, King, Stanislaus, and south fork of the Kern River basins all showed a ratio greater than 1.3 of 2016-to-2015 tree mortality area. The same pattern of increased tree mortality area in 2016 compared to 2015 was detected in the Truckee, Feather, Yuba and Tuolumne River drainages. 
 

 Landsat tree mortality from dNDMI for August 2013 to August 2016 (versus August 2011, pre-drought). (a) Lake Tahoe Bas
Figure 3: New tree mortality areas from dNDMI summed for the years 2015 and 2016 for NF and NP administrative units in the Sierr
Table 1: Annual area of new tree mortality totaled over major river basin detected boundaries using Landsat dNMDI methodology fo
Table 1: Annual area of new tree mortality totaled over major river basin detected boundaries using Landsat dNMDI methodology for Sierra Nevada study region

More detailed views of new tree mortality patterns in both 2015 and 2016 within four major California river basins were mapped in Figure 4. In the Lake Tahoe basin, three times more tree dieback was detected in 2016 compared to 2015. In one of the largest drainages in the Sierra region, the Upper San Joaquin in the Sierra NF, new tree mortality in 2015 was detected mainly below 1200 m elevation, whereas new tree mortality in 2016 was detected mainly at higher elevations up to 2200 m. The Upper King River drainage just south of the Upper San Joaquin basin showed the same kind of shift in tree mortality patterns from lower to higher elevations from 2015 to 2016. In the South Fork of the Kern River draining much of the Sequoia NF, nearly five times more tree dieback was detected in 2016 compared to 2015. Portions of the Upper Deer, Poso, and Kelso Creek sub-basins were heavily impacted in 2016 by tree dieback largely above 1500 m.
Across the entire Sierra Nevada study region, changes from year-to-year in three elevation zones (Figure 5) showed that new tree mortality increased steadily from 2013 to 2015 in the lower zone (500-1000 m) to just over 68,000 ha then added only about 3,500 ha in 2016. In the mid elevation zone (>1000-2000 m), which made up over 50% of the total Sierra Nevada forest dieback area since 2013, new tree mortality increased steadily from 2013 to 2016 to over 210,000 ha. In the high elevation zone (>2000-3000 m), which made up 22% of the total Sierra Nevada forest dieback area since 2013, new tree mortality was highest in 2013 and increased at consistent lower rates from 2014 to 2016 to nearly 84,000 ha. 
 

Figure 4: Maps of new tree mortality areas from dNDMI for the years 2015 and 2016 within major California river drainages of the

The influence of topography was further examined across the entire Sierra Nevada study region by dividing areas of tree mortality area each year into north- and south-facing aspects. Results showed that about 60% of all new tree mortality areas were located on northfacing hill slopes in 2013, 2014 and 2016 (Figure 6). In 2015, the proportion of all new tree mortality areas on south-facing slopes was 53%, which represented an exception to the overall predominance of tree dieback on north-facing slopes.

Source: Journal of Biodiversity, Management, & Forestry [Full PDF]