Upgrade numpy and install via pip

[caseycesari]

Overview

Undos work done in #3090 to get around an issue with numpy, and installs an updated version of numpy that is compatible with ulmo and other dependencies.

Connects #3093

Testing Instructions

• Analyze and run models for an area, and confirm that things work as expected.
• Follow the instructions in Circumvent GWLF-E SegFaults due to NumPy #3090 to test that the segfault issue has been addressed.

Checklist

• All JavaScript tests pass ./scripts/testem.sh

[rajadain]

Using the Darby Creek-Mantua Creek, HUC-10 Watershed boundary, I'm seeing different Hydrology, Water Quality, and Summary Load numbers on Local (running on this branch) and Staging:

diff -u staging/mapshed_hydrology.csv local/mapshed_hydrology.csv
--- staging/mapshed_hydrology.csv	2019-08-14 14:37:30.000000000 -0400
+++ local/mapshed_hydrology.csv	2019-08-14 14:37:33.000000000 -0400
@@ -1,14 +1,14 @@
 Month,"Stream Flow (cm)","Surface Runoff (cm)","Subsurface Flow (cm)","Point Src Flow (cm)","ET (cm)","Precip (cm)"
-Jan,8.58,1.88,3.46,3.24,0.59,7.92
-Feb,8.83,1.83,4.06,2.93,0.91,7.25
-Mar,10.03,1.73,5.06,3.24,2.63,8.75
-Apr,8.60,0.44,5.03,3.14,5.09,8.91
-May,7.46,0.49,3.73,3.24,9.07,9.63
-Jun,5.79,0.55,2.10,3.14,12.06,9.26
-Jul,5.22,0.92,1.06,3.24,11.88,10.81
-Aug,4.29,0.61,0.44,3.24,9.00,9.15
-Sep,4.14,0.69,0.31,3.14,5.63,8.69
-Oct,4.47,0.74,0.48,3.24,4.03,6.98
-Nov,5.01,1.21,0.66,3.14,2.25,8.40
-Dec,6.88,1.61,2.02,3.24,1.13,8.71
-Total,79.30,12.70,28.41,38.17,64.27,104.46
+Jan,9.15,1.88,4.04,3.24,0.59,7.92
+Feb,9.10,1.83,4.34,2.93,0.91,7.25
+Mar,10.16,1.73,5.20,3.24,2.63,8.75
+Apr,8.65,0.44,5.07,3.14,5.09,8.91
+May,7.48,0.49,3.75,3.24,9.07,9.63
+Jun,5.80,0.55,2.11,3.14,11.27,9.26
+Jul,5.22,0.92,1.06,3.24,10.70,10.81
+Aug,4.30,0.61,0.46,3.24,8.74,9.15
+Sep,4.22,0.69,0.40,3.14,5.46,8.69
+Oct,4.58,0.74,0.60,3.24,4.03,6.98
+Nov,5.39,1.21,1.05,3.14,2.25,8.40
+Dec,7.63,1.61,2.78,3.24,1.13,8.71
+Total,81.68,12.70,30.86,38.17,61.87,104.46
diff -u staging/mapshed_water_quality_loads.csv local/mapshed_water_quality_loads.csv
--- staging/mapshed_water_quality_loads.csv	2019-08-14 14:36:36.000000000 -0400
+++ local/mapshed_water_quality_loads.csv	2019-08-14 14:36:46.000000000 -0400
@@ -10,7 +10,7 @@
 "High-Density Mixed","242,440.8","5,323.9",546.6
 "Low-Density Open Space","205,543.5","5,469.9",581.6
 "Farm Animals",0.0,"6,833.7","1,663.0"
-"Stream Bank Erosion","253,144,182.0","136,021.0","46,096.0"
-"Subsurface Flow",0.0,"312,448.0","4,044.5"
+"Stream Bank Erosion","257,710,534.0","138,474.0","46,927.0"
+"Subsurface Flow",0.0,"338,819.4","4,385.9"
 "Point Sources",0.0,"1,752,465.0","77,164.0"
 "Septic Systems",0.0,"26,609.9",0.0
diff -u staging/mapshed_water_quality_summary_loads.csv local/mapshed_water_quality_summary_loads.csv
--- staging/mapshed_water_quality_summary_loads.csv	2019-08-14 14:36:32.000000000 -0400
+++ local/mapshed_water_quality_summary_loads.csv	2019-08-14 14:36:44.000000000 -0400
@@ -1,5 +1,5 @@
 Sources,Sediment,"Total Nitrogen","Total Phosphorus"
-"Total Loads (kg)","256,032,527.2","2,267,038.5","133,262.6"
-"Loading Rates (kg/ha)","6,137.73",54.35,3.19
-"Mean Annual Concentration (mg/L)",774.03,6.85,0.40
-"Mean Low-Flow Concentration (mg/L)",954.55,9.62,0.61
+"Total Loads (kg)","260,598,879.2","2,295,862.9","134,435.0"
+"Loading Rates (kg/ha)","6,247.20",55.04,3.22
+"Mean Annual Concentration (mg/L)",764.69,6.74,0.39
+"Mean Low-Flow Concentration (mg/L)",944.49,9.49,0.60
diff -u staging/project.gms local/project.gms
--- staging/project.gms	2019-08-14 14:56:31.000000000 -0400
+++ local/project.gms	2019-08-14 14:55:57.000000000 -0400
@@ -1,5 +1,5 @@
 10,6,0
-1.4.0,0.03228491379310344,0,10,0,0,0.06723457566537072,12.065378195901392,381204,3188.046345500479,378015.9536544995,619.7222697814921,0,2.9980752361369207,0.0045102962476675475,30,1961,1990,0.003255958295489718,59397.940628333665,0,0,0,0.18325275269809418
+1.4.0,0.03228491379310344,0,10,0,0,0.06723457566537072,9.418960164181,381204,3188.046345500479,378015.9536544995,619.7222697814921,0,2.9980752361369207,0.0045102962476675475,30,1961,1990,0.003255958295489718,59397.940628333665,0,0,0,0.18325275269809418
 0
 0
 0

Relevant files are in here: numpy-upgrade.zip.

I'll try and take a look to see if I can figure out the source of this discrepancy.

[rajadain]

As can be seen in the above diff, there is for some reason a difference in their Avg AWC value (Line 2 Comma 8 in the GMS file), which is causing the difference in numbers. That value is sourced from GeoTrellis rasters, and does not go through NumPy. It may be cached from an earlier run, either on my local or on staging.

I tried other shapes and the numbers match up there.

[rajadain]

+1 tested, this is working correctly. Numbers from MapShed match up (except in the case when geoprocessing results are different, as expected).

[caseycesari]

Thanks!