One of the inputs required by daily decision support tools for scheduling irrigation is the amount of water supplied by rainfall. In-field measurements of daily precipitation are expensive or laborious, while measurements from gauges within a few kilometers are frequently not representative due to the high spatiotemporal variability of precipitation. Online radarbased precipitation analyses from NOAA’s National Weather Service (NWS) have obvious potential to provide the needed data, but are known to have varying degrees of accuracy with location and conditions. The NWS precipitation analysis is computed on a 4 km × 4 km grid, so differences should also be expected between the product and individual gauge measurements under each grid cell. In order to test the utility of the NWS precipitation analysis in a daily irrigation scheduler, daily data were gathered in July 2012 from 18 weather stations under 2 NWS precapitation analysis grid cells across instru-mented research and production fields in the Mississippi Delta. Differences between individual station measurements and the NWS precipitation analysis are examined, and root-zone daily soil water deficits computed using both station data and the NWS precipitation analysis. Sub-grid spatial variability between gauge locations, and differences in precipitation between gauges and the gridded NWS analysis, are found to be similar to those reported elsewhere. Differences between time series of soil water deficit computed using the two different precipitation data sources are noted, but prove to be of limited impact on the decision to irrigate or not to irrigate. It is also noted that profile-filling rainfalls limit the impact of accumulating error, resetting the modeled soil water to “full”. Given the Delta-local practice of irrigating to replace full evapotranspirational water used, use of the NWS daily precipitation analysis data as input for a daily irrigation scheduler is judged not only acceptable, but also preferable to other sources of daily precipitation data.
Mississippi currently enjoys plentiful ground water resources and rainfall in excess of 1 m per year. However, producers are increasingly reliant on supplemental irrigation to improve yields and profits [
Water conservation efforts in Mississippi are nascent, as producers become increasingly aware of the need for good water management. Increasing volatility in recent weather patterns has resulted in little change in overall rainfall amounts, but a decrease in the number of events and a concomitant increase in intensity [
While less efficient than sprinkler application, surface (furrow) irrigation accounts for nearly 70% of the irrigated acres in the Delta region [
Precipitation is a key input for the water balance equation. Rainfall data is available from a wide range of sources in Mississippi, although the accuracy, spatial representativeness, and accessibility of the data are extremely limited [
Management of large farms in the Mississippi Delta limits the time that producers have for data collection and processing. Field-specific precipitation measurements for use in any irrigation-scheduling tool such as MIST would be clearly preferred. However, producers in the Delta have indicated an aversion to having a rain gauge in every field for measuring field-specific rainfall, and to any requirement to gather and enter such information into an irrigation scheduling tool. As a result, the MIST was designed to run on automated inputs from national and regional databases of soil and weather information [7,8,10-12]. This automated downloading significantly reduces the time spent by farmers in data collection and importation to the model, and is expected to enhance adoption and use of MIST.
The National Weather Service (NWS) has developed rainfall estimates based on radar and rain gauge data that are readily available and increasingly accurate [11-14]. This gridded precipitation analysis (termed 1-Day Observed Precipitation Analysis on their web site, hereafter NWS-PA) is used extensively for hydrologic and modeling studies in both the operational and research communities, and is essentially an estimate of the average precipitation falling across each grid cell. The accuracy of NWS-PA in comparison to ground-based measurements of rainfall has been tested in several circumstances, both within the grid scale of the precipitation analysis (nominally 4 km × 4 km) [
Beyond the question of accuracy of the NWS-PA product, each NWS-PA 4 km × 4 km grid cell covers an area of approximately 1600 ha. With an estimated average field size of 60 ha in Mississippi, each grid cell then covers approximately 27 fields. Given the spatially variable nature of rainfall, there will necessarily be differences between the NWS-PA and the amount that actually falls on any specific field. The question addressed in this research is one of degree: how severely will the net differences (due both to inaccuracy and spatial variability) between the NWS-PA and in-field gauge-measured precipitation impact the calculated soil water balance, and the subsequent “irrigate” or “not-irrigate” decision? In short, will use of the NWS-PA in MIST produce incorrect guidance?
The results of this analysis of the use of NWS-PA in an irrigation scheduler for the Mississippi Delta are expected to be relevant to irrigation scheduling in all humid, high-rainfall areas in the US during the summer growing season.
The most critical weather period for crop producers in the Mississippi Delta is May through mid-August. July is both in a critical period of the irrigation season in Mississippi, and a month when precipitation is primarily convective in origin (adding to the expected error [
locations were established within each grid for collection of rainfall data for a total of 18 sampling locations. StratusTM RG202 cm rain gauges (Fergus Falls, MN) were placed on 0.6 m posts at locations evenly dispersed throughout each grid, with placement chosen to give a good representative coverage of the study area. Gauge positions were adjusted as needed to accommodate land use, crops and structures, and to ensure access following a heavy rainfall. The resulting 9 gauges per NWS-PA grid cell was denser sampling than that used by [
Daily soil water balance was calculated using MIST [7, 8,10]. Water use is determined in MIST using the standard Penman-Monteith equation to calculate reference evapotranspiration (ETo) from daily measurements of maximum and minimum temperature, solar radiation, wind speed, and relative humidity. The daily crop evapotranspiration (ETc) is then calculated by multiplying ETo by a crop-specific coefficient [
(1)
where WBy is the soil water balance yesterday; ETc is the daily crop evapotranspiration; Rainu is crop-usable fraction of rainfall; and Irr is the amount of irrigation water applied.
After assessing the data quality of the field observations, and frequency of rain events during July 2012, it was decided to continue with subsequent analysis. This single month of data is not sufficient to reach any conclusions relative to the quality or representativeness of the NWS-PA data. Instead, our focus is on checking whether the gauge and NWS-PA data are consistent with results from more substantive studies in the region, then subsequently on assessing the impacts of using the NWSPA data as a substitute for daily in-field rain gauge data in MIST.
As expected, substantial spatial variability in precipitation was observed between rain gauge locations.
Note that while some of the gauge-measured values were centered around the NWS-PA value, other rain events were markedly offset in magnitude. The amount of variation between gauge measurements was not consistent between events, and did not depend on amount of rain received.
To determine if any obvious spatial or directional bias existed during July 2012 between the gauge-measured and NWS-PA rainfall amounts within these two grid cells, linear distance from the grid centroid and angular direction from the grid centroid (north = 0) were determined for each rain gauge location. Differences between gaugemeasured and NWS-PA rainfall amounts for each of seven consecutive days (including those shown in
In a few cases, precipitation on the following day would compensate for a discrepancy between NWS-PA and gauge-measured precipitation in one-day totals. To check whether the differences were exacerbated or canceled over several weeks, the cumulative rainfall differences were determined for the entire month (
NWS-PA totals tended to be smaller than the gauge totals, a result that is again consistent with previous results [
To assess the net impact of using NWS-PA as input for MIST, the model was run for each rain gauge location, adjusted for specific crop (e.g., corn or soybeans) and planting date using both the gauge measurements of precipitation and the NWS-PA for the encompassing grid cell. Daily time series of soil water deficit for each of the 18 locations were generated; examples of the best and worst match between modeled soil water deficits are presented for each of the two grid cells in
Somewhat surprisingly, although a substantial error existed between gauge-measured and NWS-PA precipitation on some days at some sites, there was limited impact on the MIST-modeled soil water deficit. The worst modeled single-day difference was on July 11 at gauge II C1, with a difference of 2.6 cm in soil water deficit (fourth panel in
Despite these differences, at only one of the locations (I C2) did the NWS-PA generated MIST output change
the irrigation decision (on July 29), with a net result of only a one-day delay in irrigation. This lack of impact indicates that the precipitation errors were not sufficiently large to dominate the daily MIST calculations when the soil water column is partially filled. But more importantly, this study makes it clear that cumulative errors are removed from the running MIST calculations whenever sufficient rainfall refills the soil water column, effectively resetting the soil water balance to “full”.
As a final check on the impacts of differences in the soil water deficit predictions by MIST due to use of NWS-PA instead of field-specific rain gauge data, the month-total cumulative soil water depletion was calculated for each site, for a variety of irrigated crops and crop stages prevalent in July in the Mississippi Delta. Corn typically has high transpiration during July, making significant demands on water within the rooting zone. Mid-(April) and late-planted (May) soybeans have relatively high transpiration rates, although not as great as corn. Early-planted soybeans are beginning to senesce in July, so exhibit sharply lower transpirational water demand. MIST predictions were also calculated using the daily average of all rain gauge measurements within each grid, a value expected to be close to the NWS-PA.
The monthly grid-averaged gauge totals closely track the NWS-PA totals. The month-totals for the two individual locations suggest that month-cumulative errors of less than 1.0 cm can be expected while using NWS-PA as input for MIST during summer months, less than 1/3 of
the corresponding month-total precipitation differences.
Substitution of NWS-PA for in-field rain gauge data has less of an impact than expected on the daily soil water deficit calculations, as well as the associated irrigate/ no-irrigate guidance provided by MIST. Sub-grid spatiotemporal variability in July precipitation in Mississippi has a negative impact on the accuracy of NWS-PA relative to rain gauge readings, and on subsequent daily soil water deficit calculations, but the magnitude of the differences during the test period only changed the irrigate/ no-irrigate decision at one site (out of 18) on one day in the month of testing. It is also apparent that any rainfall or irrigation that refills the soil water profile (typical irrigation practice in Mississippi) resets the MIST calculations of soil water deficit to zero, removing any cumulative impact of use of NWS-PA instead of in-field rain gauge data.
Given the scarcity of readily accessible, quality-controlled, in-field rain gauge measurements, the NWS-PA is judged to be a viable, even preferable, alternative for use in MIST. This result should apply in all humid, highrainfall areas of the US.
However, note that this result does not necessarily apply for farmers practicing deficit irrigation, in which the soil water profile is not completely replenished. Consideration of the size of the daily error in soil water deficit, due to use of NWS-PA, would need to be factored into the decision-making process of a deficit-irrigation scheduler. Consequently, use of MIST with NWS-PA input is not recommended for those practicing deficit irrigation until and unless further evaluation across several summers is conducted, similar to that reported here, and MIST is explicitly modified to include the results.
Given the success noted here, NWS-PA should be considered and tested as input for daily irrigation schedulers wherever NWS radar coverage is relatively complete (east of the Rocky Mountains, and western CA and WA), and in-field rain gauges are scarce. Simple adjustments in the algorithm for deciding whether or not to irrigate may be possible to limit any negative impacts from use of the NWS-PA data.
We thank the Mississippi Soybean Promotion Board and the Mississippi Corn Promotion Board for financial support for this research.
Contribution No. 14-109-J from the Kansas Agricultural Experiment Station.