1. Regional analyses of aboveground net primary production (ANPP):
using long-term data to understand the past and to forecast future dynamics
Debra Peters, Jin Yao, Dylan Burruss, Osvaldo Sala,
Kris Havstad, and Laureano Gherardi

2. What about precipitation?
S.L. Collins et al. (2009)
Non-stationarity in future drivers
How can we use historic long-term data to predict consequences of an uncertain future?
What about precipitation?
LTER
USDA

3. What can we say about precipitation?
At any given location in the grasslands region, long-term mean annual precipitation could increase or decrease in the future.,
Our ecological predictions need to account for both a directional increase or decrease in precipitation.

4. Site-based climate data show periods of drought and wet years (≥ 4y).
Jornada LTER-LTAR site
H: Responses in dry (wet) periods are different than in individual dry (wet) years and different from long-term mean response. We found this to be the case at the Jornada: what about other grasslands in North America?
Ecosystem response data (primary production; ANPP) in dry (decrease) and wet periods (increase) can be used as “natural experiment” to provide insight to future dynamics under alternative climatic regimes.

5. Does the second wet (dry) year in a row also fit this relationship?
Hypotheses for site-based responses in drier or wetter climate
2. Alternative model: biotic contingencies modify the amount of water available to plants AND affect ability of plants to respond to rainfall in dry/wet periods.
In grasslands during multi-year drought, little grass response may be expected with multiple years of low rainfall, low grass cover, high soil and water erosion, and few meristems or seeds in soil.
These legacies of previous year’s production and biomass are also expected to be important with a shift to a drier climate.
In wet periods, larger primary production (ANPP) than expected based on rainfall (higher Rain Use Efficiency; RUE) may occur as biomass and litter accumulate in a series of wet years that reduce losses of soil water to evaporation and erosion, and lead to greater plant available water.
These positive plant-soil water feedbacks are also expected to be important with a shift to a wetter climate.
Is this true for periods of years or directional  or in precipitation?
Does the second wet (dry) year in a row also fit this relationship?
How about the third wet (dry) year in a row?
1. Climate-driven model:
The more it rains this year, the more plants grow this year.
ANPP(t) = f( PRECIPITATION[t])
Hypothesis: the same equation works for all types of years (individual wet or dry years, wetter periods, drought, wetter or drier future).
ANPP
PRECIPITATION
Wet period/ wetter climate
Current climate
Drought or drier climate

6. Wet period
Long-term
No trend
Dry
Wet
ANPP data from Jornada desert grasslands and shrublands showed importance of biotic contingencies in wet periods but not dry periods.
Wet period
ANPP = 1.32 x PPT
R2 = 0.13

7. Year
Year
ANPP increased non-linearly through time in both grasslands and shrublands.

8. How general are these results for other grasslands in the central US
How general are these results for other grasslands in the central US? Regional variation in primary production (ANPP) of grasslands and shrublands, and its response to precipitation.
PRECIPITATION 
TEMPERATURE
tallgrass prairie
northern mixedgrass prairie
southern mixedgrass prairie
shiortgrass steppe
desert grassland and shrubland
Plant production (g/m2)
FORESTS
GRASSLANDS
DESERTS
Huxman et al Nature
Precipitation (cm/y)
Küchler Potential Vegetation 50
150

9. LTAR sites (ANPP) LTER sites (ANPP) 32 cm (9oC) 87 cm (15oC) 46 cm
Mandan (677)
Cheyenne (144)
CPER (95)
El Reno (300)
MAP
(MAT)
PRECIPITATION 
TEMPERATURE
25 cm
(17oC)
26 cm
(14oC)
90 cm
(13oC)
LTER sites (ANPP)
Konza (507)
Sevilleta (87)
Jornada (139)*
59 cm
(12oC)
Hays (290)

10. Net primary production (g/m2)
Long term precipitation and primary production (ANPP) data from each site.
Mandan
Precipitation (mm)
Net primary production (g/m2)
wet period
dry period

11. Separate equations in dry periods were significant at four sites; only the desert grassland had a separate wet period equation.
These equations provide better predictions than long –term ANPP-PPT equations under directional decrease/increase in PPT.
dry
no trend
wet

12. All 8 sites exhibited one of these two biotic contingency patterns.
The number of sequential years in dry (2 sites) or wet periods (4 sites) was more important than amount of precipitation to variability in ANPP. All of the southern sites, except Hays, showed this pattern.
All 8 sites exhibited one of these two biotic contingency patterns.
How to use these equations
1. Predictive equations for future climate scenarios
2. Data mining from site-based studies for process-based understanding.
3. Hindcasting – providing insight to past mysteries.
dry
no trend
wet

13. 1930s drought and ecological consequences
Hindcasting
1930s drought and ecological consequences
Plowing individual fields
Cultivation of marginal land
Drought, plant mortality, strong winds
PDSI 1934
Soil erosion
What led to the transition between loss of grass cover (IA) and total devastation (NE)?
Can we use our drought equations to answer this question?
Dust storms

14. Predicted ANPP using grassland-specific drought equations and 1934 PPT.
Vegetation transition occurs over gradient in precipitation in 1934 (4th dry yr) and change from tallgrass to mixedgrass prairie.

15. Arid west: cascading events, cross-scale interactions
Mesic east: nonlinear dynamics (long-term soil x multi-year drought)
1. To understand Dust Bowl, we need to account for spatial variation across the region.
2. Historic data invaluable in providing insight to future responses under alternative climate scenarios.
3. The long-term ANPP-PPT equations are typically not the best predictors for any grassland site under future increase or decrease in precipitation.
Dust source