Define Parameters

Step 1: Define AOI.

On the AOI selection tab, Initiate ECCwBD, select United States or Global, then Predefined or User Drawn. Once Main Panel is activated select continental U.S. or global landcover and elevation data sources. Choose from national 30m land cover or global 500m land cover sources (LP DAAC). Also select from national 10m or global 30m elevation datasets (NED, ALOS). Cross-reference national datasets with global datasets to verify results in desired. Next, select AOI using "Select AOI type" dropdown menu below. If you have a CSV or SHP (.shp, .dbf, .prj, & .shx) already uploaded  to your Earth Engine (EE) account , then from the dropdown menu select "Your own (below)" and type in the path to your EE Asset in the textbox below.  Use the GUI polygon drawing tool for boundary reference only. To draw and save your AOI, please draw within EE Code Editor, specify as Feature Collection and save as EE Asset. Sign-up for an EE account using your Gmail account. If you choose to draw your own temporary AOI for on-the-fly analysis, it is necessary to redraw it each time. Or use "Select AOI type" to select a predefined AOI, e.g., "Select US Native American reservation (below)". Then use corresponding dropdown menu to select AOI. Select predefined AOI for a grassland conservation area, wildlife refuge, county boundary, etc... Please note, larger areas require longer processing times, e.g., depending on hardware and internet speed processing time for Yellowstone National Park, USA is less than 3 minutes.

Step 2: Set timeframes for geospatial datasets.

NPP, PsnNet, & Phenology datasets (NASA 2020). After initial run get average Greenup and senescence day from Preliminary Results Panel & adjust date accordingly; works well for typical growing season Greenup and senescence day. For best results, when priming ECCwBD set the start date as close to the average MODIS Greenup day as possible. Winter predictor variables are summed from January 1st through average AOI Greenup day, hence a start date post average Greenup day increases predictive power. Run ECCwBD to get the average AOI Greenup day and senescence day (NASA).

ECCwBD R Shiny EE application; data upload.

Click "View data" to generate a complete sample of all 500m pixel centers, 5000 pixels or less, of all MODIS productivity layers used to estimate ungulate carrying capacity in these analyses (Running et al. 2004). Included are 41 environmental variables; mean MODIS cycle 1 Greenup and senescence 2001-2018, winter mean MODIS leaf area index, fraction of photosynthetically active radiation, snow cover and albedo from January 1st through Greenup, OpenLandMap 250m soil organic carbon content, CHIRPS 5566m  winter and growing season total precipitation, NLCD 30m national land cover data, USGS 10m national elevation, ALOS 30m global elevation data, and ERA5 11.1km level 2 winter mean soil temperature and snow depth, and ERA5 11.1km mean volumetric soil water content from January 1st through Greenup and for Greenup through senescence. (Hall et al. 2016, Funk et al. 2015,  LP DAAC, Muñoz 2019, Tomislav and Wheeler 2018, USGS, please see NOTICE). Also available are EE reduced 500m elevation along with 500m elevation derived slope and cosine aspect to account for landscape variation at the MODIS scale. This extraction method only works for 5000 pixels or less. For large extents, refer to ECCwBD Earth Engine application results panels where ungulate capacity estimates are based on all AOI pixels.

On click, a chart will appear at the bottom of the preliminary results panel with a chart enlarge arrow. Save ee-chart as CSV. Next, click "ECCwBD analyses // R Shiny App", upload the ee-chart to view and interact with data in R Shiny Server. Interaction options are built into the following tabs; Data Upload, Correlation & Covariance, Boxplots & Histograms, Pairs Plots, GLMM Model Fitting, GLMM Model: PsnNet Prediction, and Carrying Capacity Estimation. 

ECCwBD analyses herein are based on statistical measures of covariate & multi-covariate relationships and interactions displayed via R Shiny graphical options. Here Survivingenious offers interpretation of statistical summaries in reference to the graphical elements of this page. We offer spatiotemporally relevant prediction of season sum photosynthesis (PsnNet) per 500m^2 within AOI. We offer this insight conditioned on Apache 2.0 License and Terms.

~

ECCwBD analyses, Copyright 2021 Survivingenious LLC. Licensed under the Apache License, Version 2.0;  you may not use this file except in compliance with the License. You may obtain a copy of the License at

 http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an AS IS BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.

Step 3: Fit GLMM to predict forage.

Interactive modeling to predict prior year's seasonal net photosynthesis (PsnNet).

Prediction of the prior year's seasonal sum PsnNet per 500m pixel within AOI grassland & shrubland on slopes < %15 provides a basis for calculating change in productivity per variation in environmental predictor variables. Selected model type for these analyses is the generalized linear mixed model (GLMM) with a gaussian family, identity link function. A normal distribution in the response and, in this case, for most predictor variables means the GLMM is deprecated to a linear mixed model, though here we continue referencing GLMM. Fixed predictor variables and a random effect are interactive with options to select from 9 variable predictors and 12 static predictors for prior year's PsnNet prediction. In the example below, a fitted generalized linear model (GLMM) with a random effect from average growing season PsnNet 2001-2020 (U.S.) and additive fixed effects from mean CHIRPS winter precipitation, mean MODIS albedo Jan.-Greenup day, and a direct interaction term between MODIS surface temperature multiplied by IRA5 soil temperature. In a GLMM an intercept is calculated per random effect group, which accounts for within group variation leading to increased approximation where the local intercept is applied with generalized slope coefficients per point (Bolker et al. 2009). This GLMM yields an Akaike information criterion (AIC) of approximately 49972. This estimate of error is important when comparing models because an AIC closer to zero means a lower residual deviance, i.e., less model error. If we plan to select for the most significant model referencing AIC, then we will theoretically see greater correlation between actual PsnNet and predicted PsnNet. Spatial temporal relevant variation leads to varying degrees of AOI predictive accuracy of PsnNet and/or combinations of variables per different models hypothetically relevant to heterogeneity or extents of homogeneity across the AOI landscape. 

Basically, select combinations of variables according to the model's predictive accuracy. Here we use all data to fit our GLMM, which gives us the proximate location of greatest residual deviance and better definition of variation across the landscape when we apply the coefficients on the next tab. We can now proceed to the next tab, where we will use our fitted GLMM parameters and coefficients to predict PsnNet for the current year, or rather the timeframe of interest.

Step 4: Interpret GLMM forage prediction.

Prediction of current year's seasonal net photosynthesis (PsnNet).

Here we predict current seasonal sum PsnNet per 500m pixel within AOI grassland & shrub land on slopes < %15 based on the fitted GLMM coefficients from previous tab. The fitted GLMM on the previous tab is autosaved and the GLMM coefficients are applied to the current year predictor variables (Jan.-Greenup day). Match the predictor variables and random effect as set on GLMM Fitting tab.

Compare summaries of predicted season PsnNet with actual season PsnNet as below. In our example outputs below, high correlation (Multiple R-squared = 0.8524) exists between the AOI sum GLMM predicted growing season PsnNet (kg*C/500m^2/time) to actual growing season PsnNet (kg*C/500m^2/time) for the northern range of Yellowstone National Park, USA. Our predictive power increases when growing season conditions between years are more alike. Since our GLMM was developed for this AOI we have included the 2020 data as the Data Upload placeholder. This 2020 data can be verified by running ECCwBD for the 'Northern Range within Yellowstone National Park' within average MODIS Greenup and senescence days; NLCD 30m landcover and USGS 10m elevation selected. Eventual research of how well the prediction works across multiple landscapes under varying climatic and animal utilization conditions will provide more insight into estimating carrying capacity with big data.

Step 5: Input forage composition by weight and crude protein content.

On the Carrying Capacity Estimation tab describe the landscape by entering forage types with respective composition by weight and crude protein content for up to 4 forage types. This leads us to an AOI relevant required forage quantity to optimize fitness while limiting ecosystem disturbance. Together with the total number of days of all ungulate occupancy, estimated necessary bulk consumption of timeframe forage necessary to fulfill fitness requirements per ungulate specifications is set. Forage abundance and quality is represented by percent cover of AOI rangeland per floral species multiplied by species' respective percent crude protein content. Resultant product is summed for all designated rangeland forage species.

Step 6: Specify ungulate intake requirements and numbers.

On the Carrying Capacity Estimation tab, enter number of ungulates per age group per species for up to 5 species. Ungulate numbers along with bulk daily intake (kg) and crude protein content (proportion) requirements are used to estimate carrying capacity based on AOI proportion of available forage (kg*C/AOI m^2/time) to ungulate intake (kg/AOI m^2/time). Variable ungulate occupancy timeframes per species per age group is adjustable. Options also exits to select from using the predicted PsnNet, actual PsnNet, or average PsnNet 2001-2020 among others. Estimation of carrying capacity for a single ungulate species can be achieved by entering 1 for number of females, e.g., enter 1 bison cow to see how many animals to add. Carrying capacity estimates are responsive to user specified inputs. Below carrying capacity estimates for the northern range of Yellowstone National Park, USA were calculated with the number of bison cows set to 1 and all other inputs set to default. In general, this estimate is where the landscape supports a single bison cow either with or without calf and the respective nutritional intake requirements. Wherein there is a surplus of available forage and the carrying capacity estimate is positive then animals may be added. 

Each year other ungulates may also be present, such as wintering elk on Yellowstone's northern range. Differing combinations of ungulate species with their respective intraspecific numbers (numbers within age groups) all share space and seek to utilize a limited AOI seasonal net photosynthesis (PsnNet). Thus, the results below are likely biased high in terms of actual available forage per total ungulate intake of all species. Where inter and/or intra annual variation in AOI productivity limits stocking rates and carrying capacity estimates are negative given an expected number of ungulates, ungulate adjustment is necessary. Ungulate over stocking can lead to intraspecific and/or interspecific competition and alter the landscape primary productivity potential of preferred forage types (Bork et al. 2012, Connel 1983, Schoener 1973, Veen et al. 2008). Please reference the 'Why care?' tab for more information on ungulate disturbance and ecosystem processes.

Step 7: Interpret implications for broad carrying capacity estimates.

Season predicted PsnNet, actual season PsnNet, or slope-intercept adjusted PsnNet (predicted NPP) can be used to estimate ungulate carrying capacity. Predicted NPP is important where MODIS NPP is unavailable. Both the end of season PsnNet based carrying capacity and that based on predicted NPP can assist conservation initiatives and strategic management decisions. ECCwBD is strengthened with ground observations, such as fine-scale dominant cover-types other than rangeland, preferred grazing areas and/or proximity to water. Locale information regarding historical stocking rates and grazing patterns can bolster estimation of carrying capacity by confirmation of remotely sensed estimates with ground observations of herd fitness and vegetation community dynamics and deviance from United States’ NRCS ecological site descriptions (or relative standard reference) per AOI.  Specification and adjustment of forage crude protein content and composition by weight for up to four forage types can be used to compensate for differences in productivity due to heterogeneous cover and/or disturbances manifesting in deviance from historical vegetative community types. Further, herbivore and/or omnivore species within AOI can be any specified species who utilize and are limited by the same resources. If you know the daily crude protein and bulk intake requirements for such species, AOI carrying capacity then can be estimated.

Below are 2022, ECCwBD Earth Engine application, carrying capacity estimates for Montana Native American reservations, MT, USA where all inputs are set to default. Be aware spatio-temporal differences in MODIS Greenup and senescence introduces error over large, separate AOIs. Hence, carrying capacity estimates for all MT reservations are improved if done separately per MT reservation and then collectively summed. Earth engine uses all pixels to estimate carrying capacity, whereas estimates for rangelands 5000 pixels or less can be estimated via R Shiny portion of R Shiny Earth Engine Application.

© 2020 Survivingenious LLC. All Rights Reserved.  

Library of Congress Registration Number PAu 4-060-238