Hypotheses A-F assume that vegetation is transpiring. As noted above the June data and the widespread use of irrigation in the Central Valley means that this will often be the case. We do not have direct measurements of crop ET. However we utilize Land Surface Temperature as an indicator of plant activity. Fields with lower LST are assumed to consist of healthier and less stressed plants than those with higher LST because plants that have adequate water will transpire and cool themselves. Thus fields with lower leaf temperatures should have steeper water slopes . Results will identify opportunities and limitations of using water vapor imagery to study ET at the ground surface in this important agricultural region.This research focuses on a transect of the California Central Valley that was flown as part of the Hyperspectral Thermal Imager Airborne Preparatory Campaign . The study area includes portions of Fresno, Tulare, Kern, and Kings Counties, three of the top four leading agricultural counties in California. The study area is part of the Tulare Lake Hydrologic Region, which comprises the southern third of the Central Valley and is the largest agricultural region in California with about 1.2 million irrigated hectares out of its 4.4 million hectares total. The region is prosperous for agriculture partially because of its long growing season, with moist winters often blanketed by fog and dry summers. Tulare Lake is the driest region of the Central Valley, receiving an average of less than 25.4 cm of precipitation a year.The Surface Biology and Geology mission was identified in the 2017 Decadal Survey as a designated program element prioritized for development as a means to enhance our ability to monitor ecosystems, natural hazards, and land use over time. The proposed sensor will capture a large number of spectral bands in the visible and shortwave infrared at a 30 m resolution,large pot with drainage as well as multiple bands in the thermal infrared at a 60 m resolution. The HyspIRI Airborne Campaign offers the potential of evaluating many of the capabilities of SBG in advance of the mission.
It flew the AVIRIS and MODIS-ASTER Simulator instruments throughout California seasonally from 2013 to 2017 at an altitude of 20 km. AVIRIS is a 224 band imaging spectrometer that captures wavelengths from 350 nm—2500 nm at approximately 10 nm increments. MASTER captures 8 bands in the TIR region from 4–12 μm. NASA’s Jet Propulsion Lab preprocessed the imagery and produced orthorectified, high spectral resolution radiance and atmospherically-corrected reflectance imagery from AVIRIS at an 18 m spatial resolution and a LST product from MASTER at a 36 m spatial resolution. AVIRIS imagery was resampled to 36 m by pixel aggregation for consistency with LST imagery. Two dates of imagery from HAC were analyzed: June 6, 2013, and June 2, 2015 collected at 18:25, 21:41, and 18:59 UTC, respectively.Pixel-level water vapor estimates were calculated from the radiance imagery using Atmospheric COR rection Now 6.0 atmospheric correction software. ACORN 6 models atmospheric gas absorptions and scattering effects using nonlinear least-squares spectral fitting with look-up tables of water column densities generated from MODTRAN 4 radiative transfer runs. The result is pixel-level water vapor and liquid water estimates over the entire study scene. ACORN 6 was run in mode 1.5 for atmospheric correction of hyper spectral data. ACORN 6, mode 1.5 inverts AVIRIS radiance to solve for surface reflectance while simultaneously estimating column water vapor. Chosen parameters include the use of both 0.94 μm and 1.14 μm to derive water vapor, a mid-latitude summer model, an average surface elevation of 100 meters, automated estimate of visibility and artifact suppression.The studied agricultural landscape was characterized using both the reflectance imagery and a geographic information system layer of field data. Multiple Endmember Spectra Mixture Analysis was applied to AVIRIS reflectance imagery to estimate pixel-level fractions of green vegetation , soil, and non-photosynthetic vegetation . Spectral Mixture Analysis is a technique that estimates fractional cover of key components within a scene as linear combinations of the pure spectra of those components, called end members. MESMA is a variant of standard SMA, in that it allows the number and types of end members to vary on a per pixel basis, thus accounting for within class end member variability. End members can be derived from field-based spectral libraries, laboratory measurements or from imagery.
In this study, spectral libraries for NPV, GV and Soil were derived from AVIRIS imagery from densely vegetated , senesced and bare soil fields . MESMA was run on each of the image dates using one spectral library of 40 image-selected end members that were collected from all dates. In addition to cover fractions, we obtained field size and crop species data. This information was obtained from GIS crop data layers provided by Tulare, Kings, Kern, and Fresno Counties. Field boundaries and crop information are gathered as part of a California’s required registration and permitting of agricultural fields that use pesticides. Field sizes ranged from <100 m2 to 2.7 km2 . Using the MESMA results and the crop map, which delineates field boundaries, a mean estimate of GV, Soil, and NPV was calculated for each field.Advected water vapor should vary locally depending on wind speed and direction. Therefore, we interpolated a map of wind over the study area during the study times and dates in order to create an estimate of wind activity against which to compare water vapor patterns. To calculate wind speed and wind direction, we relied on weather station data from 13 meteorological stations in or within close proximity to the study area . Weather stations used in the analysis are managed by various sources including government agencies, private firms, and educational institutions. Meteorological data were downloaded from the MesoWest and California Irrigation Management Information System networks. For each station, the wind speed and direction that most closely matched the flight time for each date was recorded. All observations were within 30 minutes of the flight time. Using the 13 data points for wind magnitude and direction, data were spatially interpolated across the study area using an inverse weighted distance formula. IDW relies on the idea that each estimated data point will be influenced by the known data surrounding it, and this influence will diminish with distance. IDW has been widely used in climatic studies for interpolating data such as temperature, rainfall, and wind and is computationally efficient, but has been found to have low accuracy when data are sparse or unevenly distributed.
As our study area has low relief and stations are somewhat evenly disbursed, we felt IDW would be an appropriate and efficient method for wind interpolation. At the extreme north and south ends of the line, elevations range from 900 m to 600 m . However,snap clamps ABS pvc pipe clip along the densely vegetated portions of the line, elevations only range from 53 to 120 m with no significant relief. The result is a pixel-level estimation of wind speed and wind direction across the study area for each of the study dates.Water vapor analysis was performed at the field scale to reduce the impact of pixel-level variation and allow for sufficient fetch to retrieve water vapor gradients from individual fields. Fields were defined from the GIS crop data layers. To evaluate whether a gradient was present over a field, pixel-based estimates of column water vapor were derived for each field, then a 3D linear trend surface was fitted to the water vapor data. We acknowledge that non-linear trends are possible, but restrict our analysis here to only those fields were the linear trend was significant. This was done for every field defined by the GIS drop data for 2013 and 2015, . Using the fitted surface, an intercept, slope, and trajectory were calculated for each field to explore the magnitude, rate of change, and directionality of the water vapor above it. The slope acts as a measure of moisture advection as a factor of wind at the field-level. The intercept is an important measure of water vapor at both the field and scene levels. At the scale of an individual field, the intercept quantifies the build-up of moisture over a field, while at the scale of the entire study site, the spatial pattern of intercepts highlight advection of moisture across the scene. The trajectory is equivalent to the azimuth of the water vapor trend at the field level. To assess the statistical significance of the modeled, fitted surface, r-squared and p-values were also calculated. Only fields that had statistically significant linear trends were analyzed . The water vapor occurring above an example field and its corresponding fitted plane are shown in Fig 4.At the field level, we analyzed gradients of water vapor as they vary over agricultural fields as conceptualized in Fig 1 and as explained through Hypotheses B through F in Section 1. As such, we tested Hypotheses B and C by evaluating the relationship between wind speed and direction with the slope of water vapor. Even if pixel or scene-level trends were not identified in an image, we included both dates of imagery in the field-level analysis as we hypothesize that trends may occur at variable scales such that a trend can be significant at one scale but non-significant at a different scale.To test hypothesis B and C we plotted wind magnitude against water vapor slope in each year. We calculated the difference between the estimated wind direction and the trajectory of the water vapor above each field as the directional difference.
For those fields that had directional differences of less than 30˚ and a statistically significant slope of vapor, we analyzed their characteristics such as crop type and GV fraction to understand what types of fields our set of hypotheses holds for. Second, we tested the impact of field size on water vapor slope in fields of >50% GV to examine Hypothesis D. We plotted field size against water vapor gradient. Third, we examined the relationship between GV fraction and water vapor slope in order to test hypothesis E. We separated fields into groups of similar field size to control for the impact of field size and then estimated the correlation between green vegetation cover and water vapor slope and intercept within each of those groups. We compared water vapor slope and intercept in fields with lower vegetation cover with fields containing a majority GV fraction . We used a 50% GV threshold as was set in Shivers et al.. Fieldlevel correlations between GV and intercept would occur in situations with low winds and higher build-up of water vapor whereas strong correlations between GV and slope would occur if consistent, moderate winds created advection of moisture across fields. A higher concentration of water vapor would be confirmed though a positive correlation with water vapor slope if winds are consistent and moderate, or an increase in intercept if winds are faint and/or variable. Fourth, this study evaluated Hypothesis F by evaluating the slopes and intercepts of the fitted water vapor surfaces over fields of different irrigated crop species. These intercepts indicate the magnitude of water vapor above a field while the slope is indicative of the trend of vaporover a field. A one-way ANOVA was performed to assess differences in slopes between the crop species, and results were evaluated based on published water demand by species. Crop water use was calculated using the crop ET coefficient for irrigated crops for June in the Southern San Joaquin Valley of California in a dry year.Water vapor concentrations, calculated using all pixels in a flight line, varied significantly between dates with mean values of 21.1 mm, and 17.0 mm in 2013, and 2015 respectively . While 2015 had a normal distribution, 2013 showed a bimodal distribution. The image from 2015 had the least variance in water vapor values with a standard deviation of 0.99 mm as compared to 1.46 mm in 2013. Calculating linear surface trends of water vapor over fields, in 2013, 84% of the 2,591 fields had statistically significant water vapor surface trends, and 92% of the 3,186 fields in 2015 did.