The average slopes from 2013 and 2015 were 0.86 and 0.70 μm of water vapor per meter of field respectively.The two dates of imagery showed slightly different spatial trends of water vapor. In 2013, water vapor showed a clear increasing trend from southwest to northeast. This trend was significant enough that it outweighed the modest elevational trend, in which elevation increases from 58 m in the central portion of the flight line, to 114 m in the northern portion. This increase in elevation, which translates to a decrease in the distance between the sensor and ground, should result in a decrease in column water vapor. This trend was also present in 2015, but was not as pronounced. Four areas in the northern part of the flight line, with low intercepts in both dates are sparsely vegetated hills, with elevations ranging between 510 m to 600 m . Very low water vapor in the farthest south portion of the flight lines is also likely a product of surface elevation, which rises to 900 m. However, in densely vegetated portions of the flight line , elevations only range between a low of 53 to a high of 120 m. When observing the imagery at a larger scale, the water vapor from 2013 and 2015 shows strong coupling with the ground surface below with agricultural field boundaries clearly defined. This result is consistent with surface-atmosphere interactions with individual fields. Computation of water vapor intercepts and interpolation of wind directionality allowed for comparison between water vapor abundance and patterns of wind as laid out in Hypothesis A. Fig 7 shows the directionality of the wind and the water vapor intercept maps side-by-side for comparison.

The 2013 imagery shows the most clear pattern of advected moisture that generally agrees with the wind map,square pot especially in the northern portion of the study area. The intercept map shows water vapor concentration increasing from south to north while the wind direction map shows a south to north trend of wind in the northern part of the study area. As crops transpire and water vapor advects, the intercepts above fields should show increasing moisture. The southern portion of the study area shows less agreement with winds, indicating winds coming from the northeast but a water vapor gradient increasing from west to east. The 2015 image shows water vapor that is not as clear in its trends. The 2015 water vapor intercepts show patterns that are somewhat similar to 2013 with a general south to north increase in moisture, except for in the most northern portion of the flight line. In principle, there should be a relationship between wind direction and a field-based intercept. Progressing downwind, the intercept over a field will be the product of moisture from that field and any additional moisture added to the atmosphere from fields located upwind, as was found by Ogunjemiyo et al.. This appears to be the case in 2013, but is not as strongly evident in 2015. In part, this may be a product of the modest elevational gradients present along the flight line. However, it also may be a product or poor representation of winds at the field scale. The wind map is a snapshot at the time of flight, while the intercept is the product of ET and wind-driven advection that are not instantaneous.Hypotheses B and C anticipated relationships between the directionality of water vapor and its slope with both wind magnitude and wind direction. When calculating field-scale trends, 84% of the fields in 2013 showed statistically significant water vapor surface trends, and 92% of fields showed significant trends in 2015, indicating that water vapor gradients were detectable in the imagery.

When examining fields that were predominantly covered in green vegetation , we found patterns that showed relatively high variability but were consistent with our hypotheses that a moderate wind speed would show higher slopes than very low wind speeds or high wind speeds . This was particularly true for 2015, which had the highest range in wind speeds, highest r 2 and a clear peak near 2 m s –1. Although r-squared values were low, both years showed significant quadratic relationships between water vapor slope and wind magnitude .We also hypothesized that wind direction and water vapor trajectory would align if fields were actively transpiring . Sub-selecting fields that met our 30˚ degree criteria for alignment only 787 fields out of 5779 met the criteria. Of those, 275 were in 2013 and 512 in 2015, which accounted for roughly 11% and 16% of fields in those years respectively. Fig 9 highlights two fields from each 2013 and 2015 that showed vapor patterns that were consistent with wind direction, as hypothesized. Analyzing all directionally aligned fields by GV cover and crop type in each year, we found no significant characteristics related to GV when these sub-selected fields were compared to all fields in the study. When examining histograms of GV fraction within the fields that showed directional agreement, no discernable pattern was found.In fact, the mean GV of the selected fields was 0.45and 0.43 for 2013 and 2015, in comparison to 0.47 for the average of all fields in the study. In contrast, grouping the fields by crop type showed significant differences in the proportion of fields in which the water vapor slope was aligned with wind direction . Almost a quarter of the pistachio and almond fields showed water vapor trends that were in agreement with the wind direction while walnut, cherry and grape fields had roughly half that amount. These differences may be attributable to differences in the crops themselves such as rate of ET or health of the plants, or these differences could be attributable to their position within the study scene as the crops are not evenly disbursed throughout the area.

We assessed the average slope of each crop type with its expected water use to test Hypothesis F, and expected to find a positive correlation with crops that have higher water use also having higher slopes of water vapor above them. However, we did not find a significant linear trend when plotting expected water use against average slope either year . Given the results of the field size study, we tested the relationship between average slope and average field size and found a significant negative linear relationship in each year . Fig 12 shows 2013 as an example. Therefore, the differences in water vapor slope by crop type, as found by the one-way ANOVA, may be attributable to the average field size of these crops instead of their water use as hypothesized. Examining patterns of water vapor over fields within a single crop species, we plotted LST, which we use here as an indicator of plant water stress or reduced transpiration rate, against water vapor slope for fields of alfalfa, almonds, and cherries . As would be expected for actively transpiring crops, LST was lowest in the fields where GV fraction was highest. However, we also expected to find a negative relationship between LST and water vapor slope as the cooler fields should have higher rates of evapotranspiration leading to an increase in water vapor above a field, detectable through slopes . However,greenhouse snap clamps in contrast to Hypothesis F and G, LST and water vapor slope showed a positive correlation that was significant for alfalfa in both years and almonds in 2015 . This finding suggests that fields with more healthy, green vegetation have lower slopes, on average, than less vegetated or stressed fields. The reasons for this finding will be explored further in the discussion.Water vapor imagery contains information regarding the land surface, the atmosphere, and their interactions, and thus offers a valuable dataset with which to observe and quantify key fluxes between the two. In this study, we expanded upon the work of Ogunjemiyo et al. by testing several hypotheses to investigate how AVIRIS estimates of water vapor vary with the surface properties and atmospheric conditions in the Central Valley of California. In this section, the key findings will be interpreted, the challenges of this methodology will be discussed, and thoughts on how to move this work forward will be presented.The results supported that water vapor imagery will show coupling with the land surface at the pixel-level and advection at the scene-scale, under certain atmospheric conditions . Further, results supported hypotheses that water vapor gradients will form over fields as a function of wind speed and direction .

Crops with higher water demand showed higher agreement between the water vapor trajectory and interpolated wind direction but not water vapor slope . Hypotheses that water vapor slope will increase with field size , would correlate with green vegetation fraction and would be higher for lower LST crops were not supported.By calculating field-level water vapor intercepts and evaluating them over the scene, we found evidence that supported Hypothesis A, particularly in the 2013 scene. In that image, small-scale trends of vapor across the study scene showed patterns that were consistent with advection of moisture, as shown in Fig 7. This finding suggests that, as air moves across the Central Valley, crop ET adds water vapor to the atmospheric column, which builds up in the downwind direction and is detectable in regional scale imagery. The presence of a regional gradient in 2013 that runs counter to an expected decrease in water vapor with increasing elevation is particularly compelling. Hypotheses B and C were supported by significant quadratic relationships between wind magnitude and slope that suggest that water vapor slopes only occur when the wind is strong enough to create such trends but weak enough that the water vapor slope is not too shallow. Work by Ogunjemiyo et al. found their conceptual model held best when winds were at 1.17 to 1.24 ms-1. The one case that showed no water vapor patterns was with an August image with 3.91 m/s winds. The winds in our study were between 1.5 and 3 m/s, in between Ogunjemiyo’s values of 1.24 to 3.91 ms-1. The quadratic result of our wind speed vs. slope curves suggest that relationships were strongest at intermediate speeds from 2.2 to 2.5 ms-1, and that the wind speed in Ogunjemiyo were lighter than their optimal speed for creating water vapor slopes. However, there are also significant differences between the two sites, including high levels of irrigation and high levels of ET in the clonal Populus studied by Ogunjemiyo. Some fields showed water vapor trends consistent with our hypotheses regarding water vapor trajectory and wind direction, supporting Hypothesis B. Within pure GV pixels there existed a large range of temperatures that suggests that not all green crop fields were transpiring, possibly due to water shortages mid-summer during a severe drought. Therefore, we hypothesize that fields that show directional agreement between water vapor patterns and wind could be indicative of those fields that were actively transpiring at the time of flight. In agreement with this hypothesis, we found the proportion of fields that showed aligned with interpolated wind direction varied by crop type . The three crop types that had the highest rates of water application, namely alfalfa, pistachio and almond, also showed the highest proportion of fields in which water vapor slope aligned with wind direction . This result suggests that high water-use crops more frequently show water vapor patterns that align with the wind direction, which is consistent with inputs of crop transpiration the local column water vapor. Hypotheses E, F and G were not supported. These hypotheses all emanated from a larger idea that water vapor slope would correlate with ET. We did find that water vapor slope correlated with field size , but the relationship was negative, with larger fields showing lower water vapor slopes. One possible explanation is that the signal of water vapor movement becomes more diffuse and less concentrated over larger fields due to variation in wind speed and direction. For small fields, the length of time required to move a parcel of air from one end of the field to the other will be low, and thus the likelihood that the wind speed and direction will be consistent will be higher, resulting in a more pronounced gradient.