Using the Bureau of Land Management Site Characterization Reports
The video embedded below and the accompanying article were developed to support BLM Staff in using the BLM Site Characterization Reports available through the BLM Climate & Remote Sensing Data Reports at reports.climateengine.org.
Tutorial Outline
Introduction
Assessments of vegetation condition and trends at the scale of Bureau of Land Management land units are essential for evaluating land health and informing the land use planning process. However, the size and variability of many BLM land units makes conducting assessments of vegetation using traditional field methods difficult to impossible in many locations. To help overcome these challenges, the BLM Site Characterization reports provide advanced satellite-based vegetation and climate data for every BLM state office, district office, field office, and grazing allotment in easy-to-access PDF and PNG reports. In this tutorial, we’ll demonstrate each of the maps, figures, and summaries so that you can start using the reports to inform decisions.
Motivation
Ecosystems across the western US are diverse, vast, and dynamic and humans have been involved with the conservation and management of these systems for millennia. The Bureau of Land Management manages 245 million acres of shrubland, grassland, woodland, and riparian ecosystems across the western US and monitoring these vast vegetation and hydrologic resources at broad scales is a monumental challenge.
However, in recent years, satellite data availability, standardized field data, and machine learning modeling have advanced dramatically and, as a result, satellite-based vegetation datasets have become powerful and essential tools for monitoring vegetation condition and change across broad areas. In 2022, Andy Kleinhesselink and co-authors published Long-Term Trends in Vegetation on Bureau of Land Management Rangelands in the Western United States using these datasets, demonstrating their usefulness for multi-scale resource monitoring on BLM-managed lands.
These vegetation datasets are as vast as the western ecosystems, which can make them difficult for resource managers to use.
Site Characterization Report
To simplify and streamline the availability of satellite-based vegetation data, our team at ClimateEngine.org has partnered with the BLM to provide Site Characterization reports for all BLM state offices, district offices, field offices, and grazing allotments in the conterminous 48 states.
The Site Characterization reports are provided alongside the Drought reports — for which we have produced a separate tutorial video and tutorial article. While the Drought reports are updated every five days and reflect real-time drought conditions, the Site Characterization reports provide a longer-term history of vegetation and climate conditions and are updated yearly.
It is worth emphasizing at the outset that these reports are produced using modeled vegetation and climate datasets and that — while these datasets are relevant and useful in management contexts — they should be used alongside additional lines of evidence whenever possible, including field data, ecological site descriptions, photo points, and other local sources of information.
Datasets
The Site Characterization reports combine three datasets: the Rangeland Analysis Platform, gridMET, and gridMET drought.
Rangeland Analysis Platform (RAP)
The vegetation data in the Site Characterization reports come from a satellite-based vegetation dataset called the Rangeland Analysis Platform (RAP), which provides spatially and temporally comprehensive data from 1986 to present day. The RAP is also a great success story for the BLM’s Assessment, Inventory, and Monitoring strategy, as the RAP relies on AIM field plots for model training. The Rangeland Analysis Platform consists of two datasets, Vegetation Cover and Production and we provide both in the Site Characterization reports.
The Vegetation Cover data are provided as yearly maps of vegetation types from 1986-2023 and are updated every year. Five vegetation cover types are provided — annual forbs and grasses, perennial forbs and grasses, shrubs, trees, and bare ground — and the maps are provided at a 30-meter spatial resolution. That means that each pixel in the image represents roughly the area of the infield of a baseball diamond. For each pixel, a value between 0-100 is provided, which is the percent cover of that vegetation type for that location.
The Vegetation Production data are also provided as yearly maps of vegetation types from 1986-2023 and are provided at a 30-meter spatial resolution as well. However, for the Vegetation Production data, there are only two vegetation types provided — annual forbs and grasses and perennial forbs and grasses — and each pixel represents production in terms of lbs/acre.
As we will see, when used together these two datasets provide valuable information about ecosystems across the West. For more information on the RAP datasets visit rangelands.app and support.rangelands.app which provides support articles and YouTube videos. We also have information on the RAP at support.climateengine.org
GridMET
GridMET is a climate dataset that provides daily maps of many meteorological and hydrologic variables such as temperature, precipitation, and windspeed in near real-time. As such, gridMET is an important dataset for assessing drought and its components. As discussed in our video on the drought reports, precipitation and evaporative demand are two primary variables used to describe drought conditions, representing both sides of the water balance.
Both precipitation and evaporative demand data are provided in the Site Characterization reports to provide context about whether BLM land units are becoming wetter or dryer through time. For example, even if precipitation is steady through time, a region may become more arid overall if evaporative demand is increasing due to higher temperatures.
We also have information on gridMET at support.climateengine.org
GridMET Drought
GridMET Drought provides dozens of drought indicators using the gridMET dataset that we just discussed. Two important indicators are the multi-indicator drought blends, specifically the Long-term and Short-term Drought Blends. These are the drought indicators that we use in the Drought reports provided at reports.climateengine.org and are important because they represent two different time periods of drought, which capture drought impacts to different types of resources. We provide these two drought indicators in the Site Characterization reports to ensure consistency with the Drought reports and to provide context about drought as it pertains to vegetation data presented in the reports. We produced a similar tutorial video on the Drought videos that you can access by clicking the card in the video.
We also have information on gridMET Drought at support.climateengine.org
Demonstrations
Demo #1: Invasive Annual Grass
To begin, let’s look an example that demonstrates how the reports can be used to analyze one of the most challenging resource management issues in the West, invasive annual grasses.
Maps
In the left-hand column of the reports you will see maps of ‘Current vegetation cover’ in terms for annual forbs and grasses (AFG), perennial forbs and grasses (PFG), shrubs (SHR), trees (TRE), and bare ground (BGR) for the year of the report, which is 2023 in this case.
These maps show spatial patterns of vegetation cover across the land unit for each cover type. We can see that for this example land unit there is very little tree cover and that much of the land unit has high annual forb and grass cover — which is likely cheatgrass in this location — with upwards of 50% cover in some areas. In the far NW corner we can see that there is a pocket of higher shrub cover and higher bare ground cover.
Table
The ‘Vegetation conditions and trends’ table provides estimates of average vegetation cover and vegetation production across the land unit for the current year (2023) in the ‘Current conditions’ portion of the table and the Theil-sen’s slope trend during the period of record (1986-2023) in the ‘Trends’ portion of the table.
In terms of “Current Conditions”, we can see that in 2023 there was an average of 650 lbs/acre of herbaceous vegetation production across the land unit, with about half of that coming from annual forbs and grasses. Additionally, the “Trends” show that annual forbs and grasses have been strongly increasing during the period of record while shrubs have been declining.
It’s important to highlight two important considerations when interpreting the trend values and trend lines in this report. The first detail to be aware of is that the trend statistics are calculated based on the spatial averages across the land unit. So, in cases where the trend is near 0 for a given vegetation type, there will often still be areas within the land unit that are trending upward and others that are trending downward even if, on average, there is no trend. Similarly, even if a land unit shows a positive trend in, say, shrub cover, there still may be areas within that land unit that are trending downward. The second detail is that we do not calculate significance in terms of trend, so care should be taken when interpreting trend values.
Further investigation can be conducted in the Climate Engine web application at app.climateengine.org, which enables users to produce trend maps using the Rangeland Analysis Platform. These maps can help to account for variability in trends over land units and there is functionality to mask parts of the map that don’t meet a significance threshold. We encourage you to use the Climate Engine app for advanced investigation.
Timeseries Figures
The ‘Vegetation cover timeseries’ and ‘Vegetation production timeseries’ charts provide largely similar information to one another. Let’s look closely at these figures since there is a lot to discuss. As you’ll see, annual forbs and grasses, Perennial forbs and grasses, shrubs, trees, and bare ground are all represented in the ‘Vegetation cover timeseries’ and annual forbs and grasses, perennial forbs and grasses, and total herbaceous production are presented in the ‘Vegetation production timeseries.’
You’ll see that multiple lines are represented on the figures, which provide information on averages, trends, and spatial variability across the land unit. This is much more information than you’ll receive from most other tools and soon we’ll unpack why this is useful in management contexts.
In the annual forbs and grasses chart in the ‘Vegetation cover timeseries’ figure, the red solid line represents the spatial mean and the red dashed line represents the spatial median. These statistics indicate the average, but when they diverge from one another they can provide useful information. Notice that in the years between 1995-2010 the mean value is somewhat higher than the median. The mean is more sensitive to outliers, so this indicates that there were parts of the land unit that had very high cheatgrass cover that were dragging up the mean relative to the median. In the years since 2010, the mean and median have been relatively similar, possibly indicating that cheatgrass has infilled more homogenously across the land unit.
Additionally, in the AFG chart, you’ll see the red shaded area displaying the interquartile range. This is a useful way of visualizing spatial variability of the land unit through time. The interquartile range shows the spatial 25th-75th percentiles and the IQR in this land unit is pretty narrow in terms of annual forb and grass cover that’s because the annual forb and grass cover is relatively homogenous across this land unit. Notice that the 25th percentile of annual forb and grass cover isn’t very different from the 75th percentile.
If we compare that with another example land unit, we can see that although these two land units had similar mean annual forb and grass cover of around 32.5% in 2023, the IQR for the second example is much broader, indicating more variability across the land unit. To drive this point home, we can look at the maps from 2023 and timeseries figures of Annual Forb and Grass cover for our two examples and we can see in each case that there is more uniform cover in Example #1 and more variable cover in Example #2.
This is crucial information for management, as you will likely manage cheatgrass invasion much differently in our second example — where there are pockets of both low and high cheatgrass cover — from our first example — where there is consistently relatively high cheatgrass cover. Where the mean and median lines can help us to account for how a land unit is changing on average, the IQR area can help to account for how spatial variability is changing through time.
The last line you will see on the annual forb and grass figure is the dashed black line representing the trend. Here, we use a Theil-Sen slope trend calculation based on the spatial mean, which is consistent with how we calculate trends throughout the reports. The Theil-Sen slope trend is a non-parametric calculation, meaning it is robust against outliers. The line displayed in these figures has the same slope as is quantified in the Trends portion of the ‘Vegetation Condition and Trend’ table.
Together, these different statistics provide a wealth of information about each land unit in terms of averages, trends, and spatial variability that will help you to interpret vegetation condition and trend over nearly four decades.
Climate & Drought
The ‘Climate and drought’ section provides three different summaries.
The Summary table and Water balance timeseries figure both provide information on precipitation and evaporative demand. These two variables comprise the water balance of our land unit. It’s important to note that it isn’t ‘bad’ if there is more evaporative demand than precipitation. In fact, most western ecosystems are adapted to this kind of aridity and seasonal weather patterns. However, if precipitation or evaporative demand has been trending in a certain direction, that could have significant ecological and resource implications.
In our first example, we can see in the Summary table that there is an average of 9.16 inches of precipitation and 48.78 inches of evaporative demand during the water year. Importantly, both precipitation and evaporative demand have been increasing over time.
The water balance timeseries figure displays annual summaries of precipitation and evaporative demand for each water year from 1986-2023. The solid lines show the annual data and the dashed lines show the trends, with the slope of the trend in the figure being equal to the trend value displayed in the Summary table.
The Drought timeseries shows the Long-term and short-term drought blend for each year from 1986-2023 and we can see that for this land unit there have been regular fluctuations between drought years and wet years.
Let’s look at an example land unit from a much more arid environment. Notice that in this land unit there has been a downward trend in precipitation and an increase in evaporative demand between 1986-2023 and that the Drought timeseries figure indicates that drought conditions have been fairly consistent since around 2006. Perhaps unsurprisingly, both perennial forbs and grasses and shrubs have been decreasing during the same time period.
Let’s look at one more set of reports to investigate another important resource issue and to develop further understanding about how the Site Characterization reports can be used to account for vegetation change through time.
Demo #2: Tree Cover Change
In this example we’ll look at how tree cover has changed over nearly forty years in two large land units that have very different spatial patterns in terms of tree cover. Satellite data is especially useful for assessing changes in tree cover through time because tree cover can change substantially over the period of decades, but can be difficult to measure systematically through time using field methods alone. We’ll first look at a location with high variability in tree cover across the land unit.
Looking at the map of tree cover and shrub cover we can see that tree cover dominates the southwestern half of the land unit while shrub cover dominates the northeastern half. Because tree cover has such high spatial variability, this is a great example for looking at the timeseries indicators of spatial variability.
When we look at the tree cover timeseries figure, we can see immediately that the interquartile range is large. The 25th percentile has nearly 0% tree cover while the 75th percentile approaches 30% tree cover. We can also see that the spatial mean, median, and 25th and 75th percentiles all appear to be increasing through time, with the tree cover trend indicating a 4% increase in tree cover averaged over the entire land unit. Notice also that the 25th percentile of tree cover is beginning to increase above 0% toward the end of the time period — if this continued it could be an indicator of trees growing in areas that have not had tree cover during the satellite record.
Conversely, we can look at another example with relatively low variability in tree cover. Notice that in the map the land unit is covered fairly uniformly by around 30% tree cover. Similarly, the tree cover timeseries plot has a narrower interquartile range, indicating fairly uniform tree cover. Tree cover also appears to be increasing in this land unit.
Accounting for spatial patterns in tree cover and how patterns have changed through time is essential for managers across much of the west and satellite-based vegetation datasets have an important role to play in those assessments.
Wrap-Up
The Site Characterization reports provide a wealth of information on vegetation and climate that can be useful for informing management. Even still, care should be taken when using climate and satellite-based vegetation datasets, as ultimately they are models of complex systems. These reports will be most useful when used alongside additional lines of evidence to inform decisions, such as field-based vegetation, ecological sites, photopoints, local climate stations, and other locally available datasets. Together, these different information sources can help to form a more comprehensive picture of a land unit and its history, providing a firmer footing for informing decisions.
For more customized analysis of the datasets presented in the Site Characterization reports you can use the Climate Engine app or API, both of which you can learn more about at climateengine.org. The Climate Engine app has been supported continuously by the Bureau of Land Management since 2015 and is built on Google’s cloud computing platform Google Earth Engine, allowing for rapid analysis of decades of climate and satellite-based vegetation datasets.