Skip to main content
ARS Home » Plains Area » Fort Collins, Colorado » Center for Agricultural Resources Research » Rangeland Resources & Systems Research » Research » Publications at this Location » Publication #315888
Title: Quantifying characteristic growth dynamics in a semiarid grassland ecosystem by predicting short-term NDVI phenology from daily rainfall: a simple 4 parameter coupled-reservoir model

Author
item HERMANCE, JOHN - Brown University
item Augustine, David
item Derner, Justin

Submitted to: International Journal of Remote Sensing
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 7/24/2015
Publication Date: 11/30/2015
Citation: Hermance, J., Augustine, D.J., Derner, J.D. 2015. Quantifying characteristic growth dynamics in a semiarid grassland ecosystem by predicting short-term NDVI phenology from daily rainfall: a simple 4 parameter coupled-reservoir model. International Journal of Remote Sensing. 36(22):5637-5663.

Interpretive Summary: The seasonal pattern of growth in grasslands around the world can be measured from satellites using an index called the Normalized Difference Vegetation Index (NDVI). Changes in NDVI over the growing season provide an index of total plant production, which often varies dramatically from year to year in semi-arid grasslands. Predicting these yearly variations from annual or seasonal rainfall (e.g. amount of rainfall in spring months, or amount of rainfall in spring and summer months) is difficult because the timing and size of individual storm events can affect plant growth substantially, but is not reflected adequately in total seasonal or annual rainfall amounts. We develop a simple 4-parameter model that takes daily rainfall measurements over an entire growing season, and predicts grassland growth patterns as measured by NDVI. We show that this model provides a useful way of summarizing daily rainfall in a manner that accounts for the amount, timing, and size of individual storm events to accurately predict plant growth patterns in a semi-arid grassland of northeastern Colorado over 14 years. Our model also revealed how droughts in one year can affect plant growth in the subsequent year, which we suggest is related to effects of drought on soil nutrient availability.

Technical Abstract: Predicting impacts of the magnitude and seasonal timing of rainfall pulses in water-limited grassland ecosystems concerns ecologists, climate scientists, hydrologists, and a variety of stakeholders. This report describes a simple, effective procedure to emulate the seasonal response of grassland biomass, represented by satellite-based NDVI, to daily rainfall. The application is a straightforward adaptation of a staged linear reservoir that simulates the pulselike entry of rainwater into the soil and its redistribution as soil moisture, followed by the uptake of water by plant roots, short-term storage and subsequent transpiration through foliage. The algorithm precludes the need for detailed, site specific information on soil moisture dynamics, plant species, and the local hydroclimate, while providing a direct link between discrete rainfall events and consequential biomass responses throughout the growing season. We applied the algorithm using rainfall data from the Central Plains Experimental Range to predict vegetation growth dynamics in the semiarid shortgrass steppe of North America. Mean annual rainfall is 342 mm, which is strongly bifurcated into a dominantly 'wet' season, whereduring the three wettest onths (MJJ) the mean monthly rainfall is ~55 mm mo-1; and a 'dry' season, where during the three driest months (DJF), mean monthly rainfall is ~7 mm mo-1. NDVI data (the MODIS MOD13Q1 16 d, 250 × 250 m product) were used as a proxy for grassland phenology for the period-of-record 2000-2013. Allowing for temporal changes in basic parameters of the response function over the growing season, the predicted response of the model tracks the observed NDVI metric with correlation coefficients exceeding 0.92. A twostage reservoir is preferred, whereby the characteristic time for transfer of a rainfall event to the peak response of NDVI decreases from 24 days (early growing season) to 12 days (late growing season), while the efficiency of a given volume of rainfall to produce a correspondingly similar amount of aboveground biomass decreases by a factor of 40% from April to October. Behaviors of the characteristic time of greenup and loss of rainfall efficiency with progression of the growing season are consistent with physiological traits of cool-season C3 grasses versus warmseason C4 grasses, and prior research that early season production by C3 grasses is more responsive to a given amount of precipitation than mid-summer growth of C4 shortgrasses. Our model explains > 90% of seasonal biomass dynamics. We ascribe a systematic underprediction of observed early season greenup following drought years to a lagged or 'legacy' effect as accumulated soil inorganic N during drought becomes available for future plant uptake.