Abstract:

The average rice (Oryza sativa L.) farmer in sub-Saharan Africa (SSA) harvests less than 1.5 t ha’
1 of grain yield in uplands production systems, while the yield potential is substantially higher.
There are large yield potential gaps due to nutrient (especially nitrogen [N]) and water stress
(WS). The main study objectives were to determine the effects of WS imposed at different
growth stages (GS) on upland rice performance, to evaluate and compare nitrogen and water
use efficiencies, to calibrate and test a crop simulation model for predicting water uptake and
/ield under a wide range of agroecological conditions and to quantify rice yields, and yield gap
for the equatorial climate in Uganda and propose adaptive management strategies for
mproving yields.
=ield experiments were conducted between 2.013 and 2016 using two upland rice varieties
tommonly grown in SSA. Crop parameters were estimated from measured data for modelling
Purposes. The SWB-Sci model was parameterised, calibrated and tested using independent
fata from two seasons and secondary data from Ugandan research sites. Simulation studies
were then performed for diverse rice growing areas along the equator (lying 0.10°S-3.28° N
(nd 31.13° E-34.160 E) over the period 2008-2012.
Grain yield measured under well-watered, adequately fertilised conditions for the mediumduration
variety (Nerica 4) was 7.2 t ha'1, and 4.5 t ha'1 for the short-duration variety (Nerica
10). When water was withheld during tillering (Ti), anthesis and grainfilling for Nerica 4 it
resulted in severe WS, but yield penalties were minimal (<25%), compared to a 75% yield loss
with stress during panicle initiation. Considerable water savings (176-245 mm) are possible with WS during the non-sensitive GS.
Increasing N level altered tiller development, reduced thermal time to key GS, increased water
use by 17-33% and grain N uptake per unit water used of Nerica 10, compared to the zero-N
treatment. Use efficiencies for input resources declined with N rates above 120 kg N ha'1.
The calibrated SWB-Sci model generally predicted water uptake, growth and yield of both
varieties under different treatment conditions well, with little error and bias, for both Hatfield
of Nerica 10 (0.37-0.98) were generally higher than for Nerica 4 (0.08-0.86) across
agroecological zones (AEZs).
Yield gaps of upland rice varieties were variable and specific to AEZs. Inter-seasonal differences
xxi
1
and Ugandan research sites. The attainable (Yt)/potential yield (Yp) ratio ranged from 0.04 to
0.59 between locations. If N limitations are alleviated, water-limited yield (Yw) /Yp ratio values were very apparent in the bimodal and transition rainfall zones. The gaps were small for the
Eastern Savannah Moist (Yw/Yp - 0.78 ± 0.03) and the Northern Moist Farming Systems (0.75 ±0.03). Adaptive cropping tactics to increase yield and annual rice production for WS-prone
zones were identified. The use of the model should be useful in future studies to identify
specific agronomic practices to increase WUE. Information can be used to drive policy on
upland rice, for instance government initiatives to intensify rice production.